Innovation Snapshot

"SaaS business products are basically databases with some business logic. When business logic is all going to agents which can work with multiple possible such databases, those will be possible to swap much more easily than before."

— Satya Nadella, CEO, Microsoft

In today's lineup: 

Robots

• CEOs are pushing AI agents. Only 11% of tech leaders are ready.

• Real ROI from Gen AI

• The real-time AI budget most CIOs still don't have

People

• Jena Holtberg-Benge turns AI pilots into production systems at AGCO

Love

• The Challenger structure hiding inside every AI go-live

• Antifragile by Nassim Nicholas Taleb

Reading time: 3 min. 

Top Insights

• SpaceX's $1.77T IPO sets a new record, marking what AI-era infrastructure commands on public markets.

• KPMG retracts AI report over hallucinated citations, raising board-level doubt about every AI governance briefing.

• Mistral seeks €3B to challenge US model lock-in, giving enterprises a credible European AI alternative.

• Google sues Chinese network for weaponizing Gemini, making AI misuse prevention a board-level security priority.

• Pokémon Go data powered military drone AI, exposing data provenance risks in enterprise AI vendor contracts.

CEOs are pushing AI agents. Only 11% of tech leaders are ready.

A new IBM Institute for Business Value study of 2,000 technology executives finds just 11% ready for the AI agent surge their CEOs are demanding, even as 80% report that mandate from the top.

The problem is the foundation. Organizations that engineer governance into their infrastructure deploy 16x more agents at 4x lower AI spend with 18% higher margins.

Three moves the 11% make:

• Map lock-in exposure across infrastructure, data, and models before approving new deployments

• Require every agent to be observable, stoppable, and owned before it reaches production

• Separate AI spend: TCO rigor for operational AI, portfolio logic for strategic bets

Real ROI from Gen AI

How human-centric AI delivers business results

• Commonwealth Bank built an agentic fraud detection AI on Snowflake that autonomously cut fraud losses by over 20% in the first half of 2026 while generating 75% of its card fraud detection rules without human prompting.

• TotalEnergies piloted Honeywell's Experion Operations Assistant at its Port Arthur refinery to predict control-room alarms 5-10 minutes before they occurred, preventing unplanned downtime without adding headcount.

• Houston Methodist rolled out an ambient AI documentation platform enterprise-wide, cutting clinician charting time by 40% and returning 27% more face time with patients across every specialty.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

The real-time AI budget most CIOs still don't have

CloudZero

Earlier this month, CloudZero launched its Financial Control Plane for AI: call-by-call attribution that maps every dollar of AI spend to a customer, product, or feature. Only 14% of CFOs can currently demonstrate clear AI ROI, and 80% of companies miss AI spend forecasts by 25% or more.

Best for: Finance and IT leaders who need to show cost-per-outcome at the use-case level, not just the department level.

Choose if you...

• Need to attribute AI spend to specific customers, products, or revenue streams.

• Are managing AI pilots across multiple teams and need a single ROI dashboard.

• Have a CFO demanding proof of AI return before approving the next budget cycle.

Harness

Best for: Engineering and DevOps leaders who need to connect AI spend to code shipped, not just infrastructure consumed.

Choose if you...

• Are running AI coding agents (Claude Code, Cursor, Copilot) and need to track cost vs code merged.

• Manage multi-cloud AI infrastructure across AWS Bedrock, Google Vertex, and Anthropic and need unified visibility.

• Need joint IT-Finance accountability on which AI bets are delivering and which should be exited.

3 other tools to explore

• Finout suits mid-market teams managing Snowflake and OpenAI costs in a single KPI view.

• Vantage is the open-source-friendly option for engineering-led FinOps.

• Kubecost covers Kubernetes-native teams that need container-level AI cost attribution.

Transformation Champion

Jena Holtberg-Benge turns AI pilots into production systems at AGCO

Over two decades in agricultural manufacturing and digital transformation. Earlier this year, AGCO Corporation promoted her to SVP & Chief Digital & Information Officer, charged with scaling AI-led transformations across one of the world's largest agricultural equipment companies. Her focus: the data and governance infrastructure that turns AI pilots into production-ready systems for AGCO's global network of dealers and farmers.

The Challenger structure hiding inside every AI go-live

Roger Boisjoly flagged the O-ring failure seven months before Challenger. The engineers knew. The launch happened anyway because the people with trench wisdom weren't in the decision room.

Every AI go-live carries this structure. Someone on the ops team knows the model will break in production. Someone in engineering has seen the edge cases. The go-live date gets locked regardless.

Before your next launch, answer two questions: who holds the trench wisdom, and do they have the authority to stop it?

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

In Antifragile: Things That Gain from Disorder, Nassim Nicholas Taleb argues that the highest-performing systems don't resist volatility. They're designed to gain from it. For CIOs scaling AI agents across unstable infrastructure: robustness is not the goal. Optionality is.

Perfect for: IT leaders redesigning infrastructure for machine-speed change, and executives managing AI investments with short model lifecycles and asymmetric risk.

In Culture

Approved. Launched. Celebrated. Governed... next quarter.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Knowledge has to be improved, challenged, and increased constantly, or it vanishes."

— Peter Drucker, Management Theorist and Author

In today's lineup: 

Robots

• The AI savings gap leaving 44% of companies in Pilot Purgatory

• Real ROI from Gen AI

• When AI starts predicting your revenue before your reps do

People

• Jonathan Godwin uses AI to design materials faster than any lab on Earth

Love

• The knowledge gap blocking your AI savings

• The Lean Startup by Eric Ries

Reading time: 3 min. 

Top Insights

• Uber's AI budget runs dry by April, as per-developer token costs surge 18.6x in nine months across the industry.

• AirTrunk pledges $30B for India AI data centers, committing to 5GW of capacity as global AI infrastructure investment accelerates.

• Shell's C3 AI agents close the full maintenance loop, automating 30,000-asset repair cycles for measurable wave-two returns.

• NSA deploys Anthropic's Mythos for offensive cyber operations, despite an active DoD supply chain ban on the company.

• Great American AI Act would preempt state AI rules, giving enterprises three years of federal compliance clarity.

The AI savings gap leaving 44% of companies in Pilot Purgatory

Nearly 40% of companies landed below 10% in AI cost savings, well below the 11–20% they targeted.

Companies that can't prove last year's AI savings are committing to next year's wave. A Bain survey of 951 companies found 40% saw cost reductions below 10%, well short of the 11–20% projected, and 44% are financing the next round on those unconfirmed returns.

Data access is the bottleneck. Most organizations can't query their own systems. Bain calls this the primary separator between top performers and everyone else. Only 4% achieved savings above 30%.

The top 4% tie every deployment to a CFO-trackable metric before scaling.

Three moves to close the gap:

• Audit data access across your AI deployments before the next budget cycle, not after it starts.

• Assign one CFO-approved savings metric to every active AI initiative and measure it quarterly.

• Refuse to sign off on wave two until wave one's savings are confirmed, not projected.

Real ROI from Gen AI

How human-centric AI delivers business results

• Airbus partnered with Mistral AI to deploy AI across four aerospace workflows, including technical documentation automation and edge AI models onboard aircraft for situational awareness.

• Kaiser Permanente deployed AI ambient scribes across its physician network, saving the equivalent of 1,794 working days of documentation time in a single year, freeing doctors to spend more time with patients.

• Citi freed 100,000 hours of weekly developer capacity through AI-powered code reviews, and is now piloting agentic AI with 5,000 colleagues for complex multistep task automation.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

When AI starts predicting your revenue before your reps do

Gong

Gong surpassed $500M ARR this past month as enterprises doubled down on revenue AI. Teams using Gong report 77% more revenue per rep compared to teams not using AI.

Best for: Revenue and GTM leaders who need a single source of truth across every customer conversation, deal, and forecast signal.

Choose if you...

• Need AI to flag at-risk deals from call patterns before reps notice the slip.

• Want coaching that fires automatically after every call without manager overhead.

• Run a complex enterprise sales cycle and need a forecast the CFO will trust.

Clari

Best for: Operations and finance leaders who want AI-generated revenue forecasts tied directly to CRM pipeline data, not gut feel.

Choose if you...

• Prioritize forecast accuracy over conversation intelligence and already have a call-recording tool.

• Need a revenue operations view that spans sales, renewals, and expansion in one model.

• Want to connect AI forecasting directly into your financial planning workflow.

3 other tools to explore

• Avoma suits mid-market teams on a tighter budget. 

• Jiminny adds European-first support with built-in coaching analytics. 

• Chorus.ai (ZoomInfo Revenue) brings contact intelligence into the conversation layer.

Transformation Champion

Jonathan Godwin uses AI to design materials faster than any lab on Earth

The materials science industry takes a decade and $100 million to design a single new industrial compound. Godwin, a former Google DeepMind researcher, built Orbital Industries on the thesis that AI can collapse that timeline to months. His team's first product, a PFAS-free GPU cooling fluid, is on track to become the first AI-designed molecule to hit commercial markets by next year. Earlier this month, Orbital closed a $50M Series B with Nvidia's venture arm as a backer.

The knowledge gap blocking your AI savings

Most AI programs generate activity without leaving a trail. When the CFO asks what six months of investment actually produced, nobody has the answer. That's the behavioral root of the savings gap. The fix: one synchronized log, every active AI initiative, one measurable outcome per sprint.

1. Log every active AI initiative this week with a single CFO-trackable metric.

2. Assign one person per sprint to capture the result, win or miss.

3. Present two logged quick wins to leadership before each budget cycle as proof.

Build the knowledge infrastructure behind every AI breakthrough in my book, The LEAP Guide.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

In The Lean Startup, Eric Ries argues that most organizations build too much before proving anything. His Build-Measure-Learn loop forces teams to validate what actually works before scaling investment. That's the discipline separating the 4% of companies that hit AI savings above 30% from everyone else funding the next wave on projections.

Perfect for: AI program leads, COOs, and CFOs deciding whether to scale an existing pilot or shut it down.

In Culture

The CFO who asked whether wave one actually paid off. Every AI kickoff, without fail.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Artificial intelligence is not a man versus machine saga; it's about humans and machines working together."

— Ginni Rometty, Former Chairman & CEO of IBM

In today's lineup: 

Robots

• 73% of AI budgets are buying productivity theater

• Real ROI from Gen AI

• The AI OS eating enterprise procurement

People

• Barbara Bachus cuts materials discovery from years to weeks

Love

• Measure AI by what it produces, not how often it runs

• The Second Machine Age by Erik Brynjolfsson & Andrew McAfee

Reading time: 3 min. 

Top Insights

• Glean hits $300M ARR selling AI budget cuts, tripling revenue in 15 months as CFOs push back on wasteful AI spend.

• Nvidia pledges $150 billion a year to Taiwan, as Vera Rubin enters full production with five times the inference power of its predecessor.

• Xcena closes $135M to put AI inference inside memory, betting its chip can collapse ten-server AI setups into one.

• Taiwan lifts its 2026 growth forecast to 9.64%, driven by AI export demand toward its strongest economic performance in five decades.

• California clears 30 AI bills at its crossover deadline, advancing pricing, workplace surveillance, and disclosure rules toward the governor.

73% of AI budgets are buying productivity theater

Across every market surveyed, the majority of executives are ready to scale back AI budgets if 2026 profit goals aren't met. US leads at 75%.

A new G-P survey of 2,850 global leaders finds 73% report underwhelming ROI from AI, and 88% believe employees are using AI to appear productive without generating real business value. G-P calls it "productivity theater."

The pattern is familiar: tools get deployed, dashboards multiply, activity metrics climb. Revenue per employee does not.

Three things keep organizations stuck:

• Adoption targets replace outcome targets, rewarding tool usage instead of measuring value produced.

• AI layers over broken processes, automating the chaos rather than fixing it first.

• No accountability owner translates AI activity into a P&L line, leaving spend invisible to the CFO.

The 69% ready to scale back spend are not giving up on AI. They are giving up on activity theater that produces no evidence.

Real ROI from Gen AI

How human-centric AI delivers business results

• UnitedHealth launched Avery, an AI member coordinator across 1,000-plus AI tools, with Q1 2026 results beating analyst forecasts and its benefit cost ratio falling to 83.9%, 1.6 points below projection.

• BNY Mellon crossed 40% AI-authored code in Q1 2026, with Eliza 2.0 and an OpenAI partnership driving 42% EPS growth across custody and settlement operations.

• General Mills deployed AI to assess more than 5,000 daily shipments, with logistics optimization generating over $20 million in supply chain savings since its 2024 fiscal year.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

The AI OS eating enterprise procurement

Zip

Recently launched 50-plus AI agents across accounting, legal, and procurement, delivering $6 billion in customer savings and a Gartner Magic Quadrant Visionary ranking.

Best for: Enterprise teams that need AI agents operating across finance, legal, IT, and procurement from a single orchestration layer.

Choose if you...

• Run a complex procure-to-pay cycle and need AI to close the loop from contract review to payment without human bottlenecks.

• Want more than 50 specialised AI agents that can act autonomously across departments.

• Need contract cycle time reductions across legal and procurement.

Pivot

Recently raised $40M to replace legacy procurement software with an AI operating system. DoorDash runs on it; the platform processes $3 billion in invoices annually across 25 countries.

Best for: Finance and procurement teams replacing a patchwork of approval workflows with one AI-native system.

Choose if you...

• Need a single platform to manage sourcing, approvals, invoicing, and payments.

• Want real-time visibility into committed spend before it hits the books, not after.

• Are onboarding vendors across multiple countries and need a compliant, automated intake process.

3 other tools to explore

Oro Labs, Lio, and Coupa for sourcing, AI-native procurement agents, or enterprise spend management.

Transformation Champion

Barbara Bachus cuts materials discovery from years to weeks

Finding a new material that performs at specification can take years. ExoMatter, co-founded by Barbara Bachus, screens thousands of candidates in days using AI and quantum chemical modeling, serving Audi, Airbus, and Infineon. The company was recently named a VivaTech 2026 Female Founder Award finalist, selected from a global field.

Measure AI by what it produces, not how often it runs

Organizations track how many people use AI. Almost none measure what it actually changed. The fix: restructure productivity metrics to reward impact and shared knowledge, not hours logged with a tool.

1. Replace AI adoption KPIs with output-per-workflow metrics: time saved, decisions accelerated.

2. Assign an output owner to each deployment, accountable for the business delta, not the usage log.

3. Run a monthly value audit: 90 days in, no P&L movement, deprioritize.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Brynjolfsson and McAfee's The Second Machine Age makes one argument: productivity gains go to organizations that redesign workflows around their tools, not those that merely adopt them. The gap between AI investment and AI return is a management problem.

Perfect for: CEOs and CIOs whose AI stack is not moving the P&L.

In Culture

The fastest product roadmap update of 2026.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Managers light fires under people; leaders light fires within them."

— John Maxwell, Author & Leadership Expert

In today's lineup: 

Robots

• 80% of CEOs say their role is at risk if AI doesn't deliver this year

• Real ROI from Gen AI

• The AI layer your CFO and board have been waiting for

People

• Matt McKinney turns logistics chaos into supply chain foresight

Love

• You build what you measure

• Prediction Machines by Ajay Agrawal, Joshua Gans & Avi Goldfarb

Reading time: 3 min. 

Top Insights

• AI leaders push governments to address mass job loss, making workforce displacement a board-level governance issue, not an HR one.

• SoftBank adds $61 billion to market cap on AI optimism, repricing what AI confidence is worth to enterprise investors overnight.

• WSJ warns a 'vibe slop' quality crisis is coming to AI, as output that looks credible now fails on depth where it matters most.

• Microsoft races to reclaim its AI lead with agentic Copilot, betting the enterprise on computer-using agents after Wall Street's SaaSpocalypse repricing.

• Trump postpones the AI executive order after last-minute objections, leaving enterprise compliance teams without a federal framework heading into Q3.

80% of CEOs say their role is at risk if AI doesn't deliver this year

A new Dataiku/Harris Poll survey of 900 CEOs across seven regions finds 80% of global leaders say their role is at risk if their company fails to deliver measurable AI gains by year-end. Seventy-seven percent believe a fellow CEO will be ousted this year due to a failed AI strategy or AI-driven crisis. The pressure has shifted from adoption to accountability: boards are no longer asking what AI can do. They're asking what it delivered.

How leaders close the gap

• Translate board FOMO into a 90-day roadmap with unit economics attached before the next capital cycle.

• Require AI proficiency for new directors and brief existing ones on the distance between hype and the actual deployment stack.

• Sequence workflow redesign ahead of every board-approved pilot so FOMO capital stops automating broken processes.

Real ROI from Gen AI

How human-centric AI delivers business results

• Waystar used Google Cloud's Gemini-powered AltitudeAI to prevent $15.5 billion in denied healthcare claims and cut denial appeal and recovery time by 90% in under a year of deployment.

• Xylem sourced $200M across 12 procurement RFPs using Coupa's agentic AI, cutting preparation time by 60% and landing 15% in savings on each deal, without adding headcount to the sourcing team.

• Bancolombia deployed GitHub Copilot across its engineering teams, generating 30% more code and 18,000 additional code changes per year, with developer productivity up 20–55% depending on task complexity.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Build your voice stack or buy it whole

Mosaic

Recently raised $18M in a Series A to automate LBO models, DCF analyses, and financial scenario planning for deal-makers.

Best for: CFOs who need board-ready AI ROI models without adding FP&A headcount.

Choose if you... 

• Need to model AI investment scenarios for the board before the next planning cycle.

• Want AI to ingest financial documents and generate models, cutting manual prep time.

• Are consolidating spreadsheets into one AI-powered source of truth for finance and ops.

Pigment

Best for: Enterprise planning teams that need one AI platform covering revenue, workforce, and capex.

Choose if you... 

• Run multi-department planning and need AI to flag forecast variances before board meetings.

• Want finance, sales, and ops on the same live model instead of swapping files.

• Need real-time scenario modelling when AI priorities shift mid-quarter.

3 other tools: Cube, Runway, and Anaplan for AI planning on existing spreadsheets, investor-ready financial modelling, or enterprise-wide connected planning at scale.

Transformation Champion

Matt McKinney turns logistics chaos into supply chain foresight

Matt McKinney watched freight billing errors compound across Uber Freight and realized the root problem was data, not payment. Loop, the AI platform he co-founded, turns fragmented logistics and operational data into supply chain intelligence that predicts disruptions before they hit. The company recently closed a $95M Series C backed by Founders Fund, J.P. Morgan, and Index Ventures, serving enterprise shippers and manufacturers globally.

You build what you measure

This one is personal. Yesterday my wife Tanyella got her TechCrunch moment, and I couldn't be prouder. She co-founded Maka Kids, a streaming app for kids ages zero to six built on well-being, not watch time. No algorithms, no autoplay, no ads.

What I find fascinating: they mapped 650 developmental indicators before building the product. You build what you measure. If you have small kids at home, this one's for you.

👉🏽 Read the TechCrunch feature

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Three economists reframe AI as a prediction technology, showing that what gets cheaper is the raw material of decisions, not decisions themselves. The framework forces leaders to identify where human judgment still carries the most value and where AI can absorb the cost of being wrong.

Perfect for: CEOs allocating AI budgets and defending every bet to their board.

In Culture

Championed the transformation. Skipped the proof points. Very immediately available.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Innovation distinguishes between a leader and a follower."

— Steve Jobs

In today's lineup: 

Robots

• Board FOMO is your biggest AI risk this year

• Real ROI from Gen AI

• Build your voice stack or buy it whole

People

• Minna Song rewires leasing before scaling AI everywhere else

Love

• The best AI is the AI your board never sees

• Measure What Matters by John Doerr

Reading time: 3 min. 

Top Insights

• AI data centers drove US electricity prices up 76%, with America's largest grid watchdog warning the consumer cost impact is not reversible.

• HSBC upgraded Cisco to Buy as its AI infrastructure orders tripled, with orders hitting $5.3B year-to-date and management raising the full-year target to $9B.

• Runway adds $40M ARR in one quarter and expands into world models, moving beyond AI video to challenge Google on generative media infrastructure.

• JPMorgan names Taiwan the world's purest AI buildout play, lifting its bull-case target 20% as AI monetization concerns ease and institutional capital rushes in.

• Cerebras surges 68% in the biggest AI IPO of 2026, raising $5.5B on Nasdaq and minting two billionaires in the AI chip race.

Board FOMO is your biggest AI risk this year

CEOs say 35% of their performance evaluation depends on AI ROI. Their boards estimate 27%. The gap lives in the corridor.

A new BCG study of 625 CEOs and board members finds 61% of CEOs say their board is rushing AI transformation. Three-quarters of those boards believe their AI knowledge matches peers. Four in ten CEOs disagree. The gap lands hardest in capital allocation and performance reviews: CEOs weight AI ROI at 35% of their scorecard; boards estimate 27%. Speed without architecture is expensive noise.

How leaders close the gap

• Translate board FOMO into a 90-day roadmap with unit economics attached before the next capital cycle.

• Require AI proficiency for new directors and brief existing ones on the distance between hype and the actual deployment stack.

• Sequence workflow redesign ahead of every board-approved pilot so FOMO capital stops automating broken processes.

Real ROI from Gen AI

How human-centric AI delivers business results

• BMW Group extended its Figure humanoid robot programme to its Leipzig plant, with the Spartanburg pilot already delivering 30x faster simulations and 25% better overall equipment effectiveness on the X3 line.

• Citi Wealth rolled out CitiScribe to every North American advisor and unveiled Citi Sky on Google Cloud's Gemini stack, turning AI into a permanent member of the wealth team handling client documentation and real-time financial guidance.

• Mass General Brigham studied ambient AI scribes across 4,000 clinicians and reported a 10% drop in EHR documentation time plus a 42% reduction in after-hours work under a hybrid human-AI scribe protocol.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Build your voice stack or buy it whole

Vapi

This past week the platform closed a $50M Series B at a $500M valuation after Amazon Ring picked it over 40 competing voice AI vendors.

Best for: Engineering teams that want full control of the voice stack and refuse single-vendor lock-in.

Choose if you...

• Want to swap LLM, TTS, STT, and telephony components independently as the stack matures around you.

• Run a structured vendor bake-off and need an iPaaS-style orchestration layer underneath every option.

• Treat voice AI as a portable workflow, not a sealed appliance, and value modularity over managed convenience.

Retell AI

Best for: Operations leaders who need unit economics on paper before the next pilot goes to the board.

Choose if you...

• Need transparent flat-rate pricing the CFO can underwrite before scaling beyond a single contact centre.

• Want sub-600ms latency out of the box without engineering a custom telephony pipeline.

• Prefer a managed stack with enterprise compliance baked in rather than gluing best-of-breed parts together.

3 other tools to explore

Bland AI is the vertically integrated low-latency option with a proprietary phone infrastructure. ElevenLabs Agents is the voice-quality leader, riding the post-Series-D platform expansion. Deepgram Voice Agent API is the builder-friendly pick if your team already runs Deepgram speech-to-text.

Transformation Champion

Minna Song rewires leasing before scaling AI everywhere else

Property management runs on a stack of legacy systems most software teams refuse to touch. EliseAI CEO Minna Song spent six years rewiring leasing, maintenance, and renewals before she let the platform expand. The MIT computer science graduate weathered 100 VC rejections and now serves 1 in 6 U.S. rental units across 75% of the top-50 property managers at a $2.2B valuation, with healthcare next. She redesigned the workflow before she scaled the AI.

The best AI is the AI your board never sees

Boards reward visibility. They love the demo, the chatbot, the screenshot in the quarterly deck. The trouble is, the highest-returning AI deployments rarely look like that. GE's flagship system monitors 7,000 turbines worldwide and surfaces failures weeks ahead of time. Its operators describe the dashboard with one word: underwhelming. The boring deployments are the ones quietly compounding capital efficiency while the flashy ones eat the press release.

Three moves:

1. Audit every pending pilot for visible-vs-invisible value and reprioritise the silent ones.

2. Reframe board metrics from new launches to fewer failures, lower latency, less rework.

3. Brief the board on the turbine standard so the next AI demo is not the gate to funding.

Build the invisible operating systems that compound efficiency in my book, 100x.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Doerr's account of OKRs is built from Andy Grove's Intel playbook and Google's earliest scaling years. He shows how boards, CEOs, and operators stop fighting over scoreboards once objectives are ladderable, time-bound, and shared. The case studies span Bono's ONE Campaign to YouTube's billion-hour goal.

Perfect for: CEOs and CIOs being asked by their board to attach hard AI ROI numbers to next quarter's roadmap.

In Culture

$297 billion in, and half the boardroom is still 'exploring AI's potential.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

“Knowledge has to be improved, challenged, and increased constantly, or it vanishes.”

— Peter Drucker, Father of Modern Management

In today's lineup: 

Robots

• Workflow redesign is the only 4x AI multiplier

• Real ROI from Gen AI

• The $11B legal AI land grab

People

• Garth Sheldon-Coulson powers AI from the open ocean

Love

• Your CAIO is a placeholder without real authority

• Team of Teams by General Stanley McChrystal

Reading time: 3 min. 

Top Insights

• Cloudflare cuts 20% to go AI-first, paying the redesign bill before the AI dividend.

• Cognizant ships AI governance as a lifecycle service, the stack that makes the new CAIO role operable.

• Anthropic's Mythos forces a White House AI rethink, moving model security review from optional to mandatory.

• JPMorgan calls the 'SaaSpocalypse' overdone, naming Palo Alto Networks, Oracle, and CrowdStrike as AI-resilient.

• Meta and Google enter the 'agentic wars', turning AI into revenue infrastructure for ad and commerce.

Workflow redesign is the only 4x AI multiplier

CEOs who have redesigned five core business areas are 4x more likely to have delivered on their business objectives.

A new IBM CEO study of 2,000 senior leaders across 33 countries finds the AI scoreboard now turns on workflow architecture, not tool selection. Companies that redesigned five core areas (technology, finance, HR, operations, and cross-functional collaboration) are 4x more likely to deliver on their business objectives. Meanwhile 76% of organizations now appoint a Chief AI Officer, up from 26% last year. Accountability is locking in fast. Without the redesign underneath, the new role has nothing to govern.

What sets leaders apart

• Pay down workflow debt across technology, finance, HR, operations, and cross-functional collaboration before scaling any AI initiative

• Lock the 90-day roadmap before naming a Chief AI Officer so the role inherits an execution architecture, not a strategy vacuum

• Underwrite each pilot with unit economics so the CFO sees capital efficiency before AI moves from pilot to production

• Treat the 4x multiplier as an architectural threshold, not a productivity stat, and budget the redesign cost into every AI program

Real ROI from Gen AI

How human-centric AI delivers business results

• Equinor wired 24,000 sensors into predictive maintenance and accelerated seismic interpretation tenfold, crediting AI with USD 130 million in 2025 savings and over 330 million since 2020.

• AIG rebuilt follow underwriting on Palantir Foundry with McGill and Partners, automating up to $1.6 billion in specialty premium decisions through agentic real-time risk ontology.

• Hisense standardised factory operations monitoring on Lenovo and NVIDIA-powered AI agents, reaching 100% coverage with 40% fewer alerts and 50% faster issue investigation across plants.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage 

Harvey

Harvey closed a $200M raise at an $11B valuation this past month, then pushed enterprise contracts further with Ansarada for M&A diligence and a global rollout through PwC's Corporate Finance Network.

Best for: Large law firms and US in-house teams running collaborative diligence on complex matters.

Choose if you:

• Run US-centric matters that lean on case law, SEC filings, and federal regulatory depth.

• Need agentic workflows for M&A diligence, litigation prep, and partner-led document review.

• Have budget for a 3-to-6 month enterprise rollout with a co-build motion alongside the vendor.

Legora

Best for: Lean teams pushing high-volume contract portfolios through structured, auditable review.

Choose if you:

• Process thousands of contracts and need structured data extraction, not summary paragraphs.

• Operate across the UK, EU, or APAC and need regional data residency on day one.

• Want onboarding measured in days and a price point that survives a CFO review.

3 other tools to explore

For lighter lifts, Spellbook drops contract drafting straight into Word, Robin AI pairs review software with a managed legal-ops team, and Ironclad runs the full contract lifecycle from first draft to obligation tracking.

Transformation Champion

Garth Sheldon-Coulson powers AI from the open ocean

AI's electricity bill is now an existential constraint on every cloud roadmap. Panthalassa CEO Garth Sheldon-Coulson built self-propelled ocean nodes that capture wave energy and run GPUs onboard, beaming results back via Starlink. The Portland team recently raised $140M led by Peter Thiel at a $1B valuation, after roughly a decade in stealth. He redesigned the energy stack from the seabed up rather than bolt AI onto a strained grid.

Your CAIO is a placeholder without real authority

A title is not authority. Most AI initiatives stall because the new role gets a name on the org chart but not the budget, the veto, or the seat in the room where bets are made.

Three moves:

1. Give the CAIO a P&L line item, not just a budget request flow.

2. Hand them veto power over any AI roadmap that touches their named outcomes.

3. Put them in the weekly executive review with the CFO and CRO, not in a quarterly steering meeting.

Build the operating model that makes innovation stick in my book, The LEAP Guide.

👉🏽 Reply to this email with "LEAP" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Team of Teams by General Stanley McChrystal turns a Special Operations command into a network. McChrystal shows how shared consciousness and empowered execution replace top-down command when the operating environment moves faster than the chain of command can.

Perfect for: CEOs and CIOs redesigning cross-functional operations for the agentic AI era.

In Culture

300 megawatts of SpaceX power went to Claude Code. Grok got the deep end.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

“Now is the most important time.”

— Jensen Huang, Founder & CEO of Nvidia

In today's lineup: 

Robots

• Growth not productivity separates AI leaders now

• Real ROI from Gen AI

• Vertical CX agents vs horizontal agent platforms

People

• Lilian Schwich closes the loop on Europe's battery supply

Love

• How AI leaders escape the pilot purgatory trap

• High Output Management by Andy Grove

Reading time: 3 min. 

Top Insights

• Microsoft's AI run rate hits $37B, turning the PwC leader-laggard gap into a P&L line.

• McKinsey hands client-team selection to AI agents, putting agentic ops at the heart of the consulting playbook.

• Caterpillar's last bear capitulates on AI power demand, naming AI infrastructure a multi-sector P&L tailwind.

• Forrester names CIOs the new agent governors, recasting the Strategy Vacuum as a role mandate.

• Nebius buys Eigen AI for $643M, consolidating the inference layer where AI unit cost compounds.

Growth not productivity separates AI leaders now

A new PwC study of 1,217 senior executives across 25 sectors shows just 20% of companies are now capturing 74% of AI's economic value, and the gap is widening fast. Leaders pull away because they point AI at growth and business reinvention, not at cost reduction alone. They report productivity gains of 25–40% inside augmented workflows. The median company sees 3–7%, still trapped in pilot purgatory.

What leaders must do

• Build the 90-day roadmap before any tool selection so an internal sponsor owns the execution architecture

• Pay down workflow debt and remediate data inputs before scaling AI across any broken process

• Underwrite every pilot with unit economics so the CFO sees capital efficiency before the second budget cycle

• Anchor each initiative to industry convergence, the single strongest predictor of AI financial returns

Real ROI from Gen AI

How human-centric AI delivers business results

• Restb ai standardised fragmented MLS listing data across 26 new partner systems, reaching 1 million+ real estate agents across North America with image-recognition workflows.

• CIMB Niaga rebuilt its banking workflows on Google Cloud and Artefact's life-centric AI agents, shifting millions of retail clients to context-aware banking across Indonesia.

• Cleveland Clinic scaled ambient AI across 80+ specialties to absorb clinical documentation, with 80% of physicians reporting reduced charting time and more direct patient face-time.

Tool Spotlight

Vertical CX agents vs horizontal agent platforms

Adobe CX Enterprise

Recently unveiled at Adobe Summit to orchestrate customer-lifecycle agents with Brand Intelligence and an MCP-compatible Coworker layer.

Best for: Marketing and CX leaders consolidating fragmented prospect, engagement, and loyalty workflows.

Choose if you…

• Need brand-consistent agents that reason against your existing creative stack.

• Want MCP endpoints to plug agents into AWS, Google Cloud, or Anthropic.

• Prefer a CX-native platform over retrofitting horizontal agent tools.

Google Gemini Enterprise

Best for: Technology and innovation leaders running hundreds of agents from different vendors.

Choose if you…

• Need a single runtime and observability layer across departments.

• Want low-code Agent Studio plus a kit for multi-agent orchestration.

• Prefer an open partner marketplace over building every connector in-house.

3 other tools to explore

ServiceNow AI Orchestrator, LangGraph, and Tray.io are strong alternatives for ITSM-rooted digital workers, open-source multi-agent orchestration, or an iPaaS spine connecting best-of-breed components.

Transformation Champion

Lilian Schwich closes the loop on Europe's battery supply

Europe's EV transition produces vast volumes of spent batteries with no clean recycling path. Cylib CEO Dr. Lilian Schwich built a water-based process to recover lithium, cobalt, and nickel at industrial scale. The Aachen team recently joined the SIB:DE sodium-ion recycling consortium after securing a €63.4M EU grant, closing the loop on critical minerals rather than chasing the next battery.

How AI leaders escape the pilot purgatory trap

Well-funded initiatives often stall out because the systems built to protect your core business naturally suffocate radical ideas before they reach a prototype.

Vision drift trips most AI strategies, with failure rates climbing above 70%. The fix: stay stubborn on vision, flexible on execution.

Three moves:

1. Lock the outcome metric for the next twelve months and refuse to relitigate it.

2. Treat every quarter's plan as a draft, not a contract.

3. Reassign budget weekly toward whichever pilot is closest to that fixed metric.

Build the operating system for the next decade in my book, 100x.

👉🏽 Reply to this email with "100x" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

High Output Management by Andy Grove turns a CEO's calendar into output. Grove defines management as leverage and shows how meeting cadence, one-on-ones, and indicators replace gut-feel decisions when scale outpaces intuition.

Perfect for: CEOs and operating leaders rebuilding decision velocity for the agentic era.

In Culture

The pilot grew up in a greenhouse. The rollout has to survive the weather.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Artificial intelligence is not a man versus machine saga; it's about humans and machines working together."

— Ginni Rometty, Former CEO of IBM

In today's lineup: 

Robots

• Process redesign dictates enterprise AI success

• Real ROI from Gen AI

• AI-native production operations vs legacy AIOps platforms

People

• Paul Manwell gives heavy industry a real-time reasoning engine

Love

• Evading the three traps of innovation

• Traction by Gino Wickman

Reading time: 3 min. 

Top Insights

• Meta bypasses Nvidia with a multibillion-dollar deal to run AI on Amazon’s custom CPUs.

• Anthropic’s elite new hacking model forces companies to patch vulnerabilities before it is weaponized.

• Sam Altman pivots OpenAI from broad research to monetizable enterprise tools, signaling higher licensing costs.

• Cohere and Aleph Alpha merge to build a Silicon Valley alternative for highly regulated sectors.

• AI is actively displacing administrative roles, forcing leaders to handle immediate workforce restructuring.

Process redesign dictates enterprise AI success

Stanford's recent research reveals that 77% of AI adoption challenges stem from process redesign and change management, not technology. Integrating these systems creates massive friction because real performance gains require fundamentally altering how human teams operate.

What leaders must do

• Restructure workflows before forcing generative tools into outdated operational frameworks

• Audit data pipelines immediately since poor quality breaks enterprise models

• Assign domain experts to lead integration so AI solves real business problems

• Treat deployment as a change management priority demanding direct operational oversight

Real ROI from Gen AI

How human-centric AI delivers business results

• Hexaware Technologies, a global IT services firm, leveraged an internal AI platform to automate its software development lifecycle, which reduced manual testing efforts by nearly 90% and accelerated feature launches by 60%.

• Intuit automated 93% of routine tax forms to generate $90 million in efficiencies, allowing them to shift human experts into 600 new storefronts for dedicated client advising.

• NetApp partnered with Google Cloud so enterprises can run AI directly on existing data, bypassing the massive cost and delay of infrastructure migrations.

Tool Spotlight

AI-native production operations vs legacy AIOps platforms

Resolve AI 

Recently raised a $40 million Series A extension at a $1.5 billion valuation to launch AI Labs for complex production environments.

Best for: Engineering and tech leaders requiring a purpose-built AI system to operate and manage software directly in production.

Choose if you...

• Need domain-specific models that analyze telemetry logs and metrics to act autonomously.

• Require an agentic system that meets strict latency, reliability, and control standards.

• Want to automate complex infrastructure resolution rather than just triaging alerts.

PagerDuty Advance 

Best for: Operations and tech teams needing generative AI features embedded directly into their existing incident management workflow.

Choose if you...

• Want to summarize complex alert clusters instantly for human responders.

• Need automated drafting of post-incident reports and stakeholder updates.

• Prioritize keeping human engineers in the loop within an established on-call platform.

3 other tools to explore 

BigPanda, Datadog Watchdog, and Moogsoft are strong alternatives for teams looking to correlate alerts, streamline IT operations, and reduce alert fatigue without adopting fully autonomous response agents.

Transformation Champion

Paul Manwell gives heavy industry a real-time reasoning engine

Operators at refineries and chemical plants make critical decisions using fragmented data and slow engineering analysis. Archimetis CEO Paul Manwell built an operational reasoning system to fix this bottleneck. His platform instantly cross-references live sensor feeds against historical engineering logs, giving operators immediate context rather than just raw numbers. Backed by an $11.5M raise in February 2026, Archimetis is now delivering up to $45M in annual value for major refineries. Read more

Evading the three traps of innovation

Well-funded initiatives often stall out because the systems built to protect your core business naturally suffocate radical ideas before they reach a prototype.

Scaling new projects means actively dismantling the organizational habits designed to maintain the status quo.

1. Define your business by the underlying value you provide rather than the specific products you sell today.

2. Build distinct incentive structures for the teams exploring new frontiers and the teams optimizing current operations.

3. Require direct interaction between legacy operators and digital innovators so new concepts do not die in isolation.

Identify the blind spots in your own organization in my book, The LEAP Guide.

👉🏽 Reply to this email with "LEAP" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Traction by Gino Wickman warns that vision without execution is just hallucination. Wickman introduces the Entrepreneurial Operating System (EOS), a practical method for aligning leadership teams. It replaces vague goals with "Rocks" and a rigorous meeting cadence that forces issues to the surface before they kill the quarter. 

Perfect for: COOs and founders scaling past the "start-up" chaos phase.

In Culture

Productivity is up, and so is your future migraine…

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"The courage to imagine the otherwise is our greatest resource."

— Daniel J. Boorstin, Historian

In today's lineup: 

Robots

• A playbook for AI trust

• Real ROI from Gen AI

• Standalone AI agents vs turnkey support platforms

People

• Laura Crabtree builds an OS for space

Love

• Closing the human adoption gap 

• The Evolution of Everything by Matt Ridley

Reading time: 3 min. 

Top Insights

• Internal duplicate systems eat Amazon's AI ROI

• European AI chips secure nine-figure backing

• Labor Department steps in to fix the AI skills gap

• AI slop scandals shift focus from real governance issues

• Anti-AI movement targets executive leadership

Expert AI optimism masks growing workforce anxiety

Stanford University’s 2026 AI Index exposes a massive divide between tech insiders and everyday workers. While 73% of AI experts expect the technology to positively impact jobs, 64% of Americans are actively bracing for job losses. With only 31% of the public trusting the government to regulate AI responsibly, the burden falls entirely on employers to manage the friction between executive enthusiasm and employee resistance.

A playbook for AI trust

• Address workforce anxieties directly instead of relying on vague promises of efficiency

• Redesign entry-level roles around AI assistance rather than strictly replacing junior staff

• Federate domain-specific teams to track how new tools affect daily workloads and pay

• Leverage phased approaches by targeting high-friction administrative tasks first to prove value

Real ROI from Gen AI

How human-centric AI delivers business results

Schneider Electric embedded Microsoft Azure OpenAI into its industrial automation tools to generate PLC code, cutting development time for engineers by up to 50%.

Relay Therapeutics used generative AI to develop a new PI3Kɑ inhibitor for breast cancer, shrinking tumors in 81% of patients during clinical trials.

Lloyds Banking Group scaled agentic AI systems across its customer and internal operations, delivering $63 million in efficiency gains last year and projecting over $127 million in value for 2026.

Tool Spotlight

Standalone AI agents vs turnkey support platforms

Sierra 

Recently reached a $4.5 billion valuation after securing major enterprise deployments with brands like SiriusXM and WeightWatchers.

Best for: Operations and tech leaders needing a highly customizable, multi-model AI agent built to handle complex, brand-specific workflows.

Choose if you:

• Need to orchestrate multi-step transactions directly within your proprietary backend systems.

• Require absolute control over how the AI reflects your brand's unique tone and personality.

• Prefer a specialized conversational platform over a traditional, all-in-one help desk suite.

Fin AI

Best for: Customer support leaders who want an AI agent that integrates instantly into their existing help desk and

Choose if you:

• Want a turnkey AI agent that deploys in minutes by reading your existing support articles.

• Need a seamless, context-rich handoff between the AI and your human support team within one shared inbox.

• Prioritize rapid, automated resolution of routine queries over building heavily customized, code-intensive conversational flows.

3 other tools to explore 

Ada, Zendesk AI, and Kustomer are strong alternatives for teams looking to automate customer interactions, streamline frontline support workflows, and efficiently scale service operations without requiring heavy engineering overhead.

Transformation Champion

Laura Crabtree builds an OS for space

Space missions frequently rely on fragmented, outdated software. Epsilon3 CEO Laura Crabtree experienced this friction firsthand at SpaceX. She built an operating system that replaces paper manuals with AI-enhanced, real-time digital procedures. Recently named to Inc.’s 2026 Female Founders list, her platform standardizes how flight controllers and engineers manage high-stakes launches. Crabtree is turning siloed data into clear operational logic, proving that successful spaceflight requires secure, centralized software. Read more

Closing the human adoption gap 

Technology evolves in weeks, but human behavior takes decades. Teams clinging to legacy workflows act as the primary bottleneck to scaling new systems. Employees will actively fight transitions that threaten their daily output or carry unmanaged personal risk.

1. Require executives to actively champion the shift and personally participate in the disruption rather than just delegating it.

2. Reward experimentation and validated learning even when a specific initiative fails to scale.

3. Integrate intelligence tools directly into existing communication channels so the barrier to entry is zero.

Learn how to manage this transition in my book, 100x.

👉🏽 Reply to this email with "100x" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

The Evolution of Everything by Matt Ridley challenges the "Great Man" theory of innovation. Ridley argues that change is evolutionary and bottom-up, not top-down and designed. For leaders, the lesson is humble but powerful: stop trying to command the market. Build a system that can adapt to the inevitable mutations of customer demand. 

Perfect for: Strategists moving from rigid planning to adaptive systems.

In Culture

Productivity is up, and so is your future migraine…

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"AI agents will replace all software."

— Satya Nadella, CEO of Microsoft

In today's lineup: 

Robots

• De-risk your agentic AI deployments

• Real ROI from Gen AI

• AI engineering agents vs AI code editors

People

• Matei Zaharia is pushing AI beyond human mimicry

Love

• Fixing the centralized bottleneck

• The Technology Fallacy by Kane, Phillips, Copulsky, and Andrus

Top Insights

• Enterprise AI adoption replaces pilot purgatory at scale

• AI destroys legacy enterprise software revenue models

• Regulatory briefings shift enterprise AI risk priorities

• Government sectors validate AI as autonomous coworkers

• Aggressive AI regulation triggers immediate constitutional lawsuits

How to de-risk your agentic AI deployments

Deloitte's Q1 2026 benchmarks show 42% of Fortune 500 companies now run autonomous AI agents in production. Despite early adopters reporting 171% returns, rushing to hand over API keys to untested models carries severe systemic risks.

The backlash is already here. While historical baselines show 73% of enterprise tech buyers experience high regret before implementation even finishes, the stakes for autonomous agents are much higher. Gartner warns over 40% of these new AI initiatives will fail by 2027 if teams ignore operational governance and automate broken legacy workflows.

What leaders must do now:

• Audit legacy friction points to find where capital is wasted before granting API access.

• Decouple data into a neutral lakehouse to swap AI models without vendor lock-in.

• Enforce hard limits that automatically terminate runaway pilots to restrict financial exposure.

• Run rapid 12-week prototypes with five-person cross-functional teams instead of funding massive platform overhauls.

Real ROI from Gen AI

How human-centric AI delivers business results

• Vodafone launched a generative AI-powered chatbot for its VOXI brand to better serve its Gen Z customers, resulting in more human-like conversations and improved customer satisfaction scores.

• Mantel Group built custom AI agents with Google Gemini to connect scattered internal data, reducing the time needed to draft Statements of Work by 75%.

• Qdesq leveraged Amazon Bedrock to build an AI assistant for venue searches and lead qualification, cutting manual CRM tasks by 80% and doubling platform scalability.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

AI engineering agents vs AI code editors

Devin

Partnered with Cognizant in Jan 2026 for a large-scale enterprise rollout, moving autonomous AI into production.

Best for: Engineering teams ready to hand off end-to-end tasks instead of just generating code snippets.

Choose if you...

• Need autonomous execution for backend tickets.

• Want an agent that tests and debugs its own code without continuous prompting.

• Prefer assigning broad objectives over managing line-by-line syntax.

Cursor

Best for: Developers who want an AI assistant embedded directly in their active coding environment.

Choose if you...

• Want multi-file edits while keeping absolute control over the codebase.

• Need a seamless transition from VS Code without losing extensions.

• Require instant contextual answers about your repository architecture.

3 other tools to explore

Windsurf, GitHub Copilot Workspace, and Amazon Q Developer are strong alternatives for teams looking to move beyond simple code completion toward fully integrated, AI-driven development workflows.

Transformation Champion

Matei Zaharia is pushing AI beyond human mimicry

The industry's obsession with human-like AGI is a security dead end. Databricks co-founder Matei Zaharia warns that treating agents like trusted assistants exposes companies to hacks and unauthorized actions. Following his 2026 ACM Prize in Computing win, Zaharia is pushing for a reset: stop judging AI by human standards and focus it on superhuman tasks like analyzing radio waves or simulating molecular changes. The goal is a purpose-built tool for discovery, not a flawed digital person. Read more

Fixing the centralized bottleneck

A "Single Source of Truth" often becomes a single point of failure. Centralizing everything creates workflow debt where the slowest approver dictates the pace. You end up with a unified mess instead of a nimble organization.

Three shifts to regain velocity:

1. Federate the architecture. Let business units build their own models on a shared data layer.

2. Standardize the API. Enforce how systems connect, but let teams pick the tools that deliver the best unit economics.

3. Structure the data flow. Design meetings to capture friction points rather than forcing rigid consensus.

Get the federated operating model in my book, 100x.

👉🏽 Reply to this email with "100x" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

The Technology Fallacy by Kane, Phillips, Copulsky, and Andrus uses data from 16,000 surveys to prove that digital transformation is a people problem, not a tech problem. It highlights that "digital maturity" correlates with risk tolerance and distributed leadership, not software spend. It forces leaders to look in the mirror before looking at the vendor list. 

Perfect for: HR leaders and CEOs managing cultural resistance.

In Culture

Why break production manually when you can automate the disaster?

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"AI agent workflows will drive massive AI progress this year—perhaps even more than the next generation of foundation models."

— Andrew Ng, Founder of DeepLearning & Co-founder of Google Brain

In today's lineup: 

Robots

• Security fears block the agentic AI leap

• Real ROI from Gen AI

• Your automated compliance and trust engine 

People

• Alexander Fitzgerald is coding the physical supply chain

Love

• The hard math of inaction

• Why Digital Transformations Fail by Tony Saldanha

Top Insights

• ServiceNow killing the seat license forces a total rewrite of software ROI.

• Court sanctions prove ungoverned AI in regulated workflows is a direct financial liability.

• Anthropic's upcoming models force companies to scale cybersecurity before deploying new AI.

• The macroeconomic consensus just shifted from job augmentation to structural workforce displacement.

• Microsoft's $10 billion SoftBank deal proves sovereign compute capacity dictates future cloud pricing.

Security fears block the agentic AI leap

A recent McKinsey survey reveals nearly 66% of enterprise leaders cite security risks as the primary barrier to scaling agentic AI. While organizations rush to deploy autonomous systems, only 30% have mature governance in place. This gap leaves businesses vulnerable as AI shifts from answering prompts to autonomously executing core workflows.

What leaders can do:

• Assign explicit accountability for AI governance to dedicated internal audit or ethics teams.

• Start with pilot projects to test autonomous agents in isolated environments before live integration.

• Federate domain-specific teams to build technical controls that mitigate IP and data privacy risks.

• Leverage phased approaches to measure AI trust as a core business enabler tied to EBIT impact.

Real ROI from Gen AI

How human-centric AI delivers business results

• Siemens runs over 50 AI applications and a digital twin at its Nanjing assembly plant to simulate operations before physical execution, which increased manufacturing productivity by 14% and cut production lead times by 78%.

• Air France-KLM uses a Google Cloud-powered AI Factory to analyze cabin crew reports, reducing feedback processing time from months to days.

• AT&T replaced its monolithic large language models with a LangChain multi-agent system on Azure. Using smaller, task-specific worker agents cut AI model costs by 90% while pushing daily processing capacity to 27 billion tokens.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Your automated compliance and trust engine 

Vanta

Best for: Startups and mid-market teams that need to achieve security frameworks like SOC 2 fast. It acts as an out-of-the-box engine to automate evidence collection.

Choose if you…

• Need to instantly pull evidence from your existing cloud and SaaS integrations

• Want to build a public Trust Center to unblock sales deals quickly

• Prefer pre-built policies that get your team audit-ready in weeks instead of months

Drata 

Recently launched new agentic AI capabilities designed to transform third-party risk management and automate security questionnaires.

Best for: Scaling operations that need deep customization and AI-driven automation to handle complex compliance across a global footprint.

Choose if you…

• Need AI agents to autonomously assess vendor risk and answer security questionnaires

• Require customizable controls that adapt to your specific internal workflows

• Manage overlapping compliance frameworks across multiple international regions

3 other platforms to explore 

Secureframe, Sprinto, and Tugboat Logic are strong alternatives for teams looking to streamline security audits and monitor continuous compliance.

Transformation Champion

Alexander Fitzgerald is coding the physical supply chain

The $1.8 trillion component manufacturing market is quietly breaking. Aging machine shops cannot keep up with aerospace and defense demands. Alexander Fitzgerald built Isembard to rebuild the factory floor. His AI-driven operating system, MasonOS, automates everything from quoting to quality control. Fresh off a $50M raise to open 25 new facilities, Fitzgerald is proving that the best way to secure physical supply chains is to treat manufacturing like scalable software. Read more

The hard math of inaction

Waiting for AI to get cheaper is not a strategy. It is a liability. While you wait for costs to drop, your competitors are building proprietary data assets. You are not saving money. You are accumulating data debt that will cost millions to fix later.

Three taxes you pay every day you wait:

1. Data debt. Competitors are compounding their proprietary datasets while yours sit in silos.

2. Execution lag. AI is a muscle you have to build. You cannot buy operational maturity in two years.

3. Rising floor. As the market standard rises, your manual workflows become an anchor on your P&L.

Learn more in my book, 100x.

👉🏽 Reply to this email with "100x" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Why Digital Transformations Fail by Tony Saldanha diagnoses the specific failure points of the 70% of initiatives that crash. He moves beyond vague "culture" talk to identify the rigorous disciplines required at each stage of maturity. The book provides a checklist for ensuring your pilot doesn't just work, but survives the transition to the P&L. 

Perfect for: Executives tired of pilot purgatory.

In Culture

The roadmap to scaling agentic AI is clear, but the detour is just too tempting.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

“AI isn’t replacing teams—it’s raising the bar for what every team is expected to deliver.”

— Chester Chipperfield, former Global Creative Director at Tesla & Head of Special Projects at Apple

In today's lineup: 

Robots

• AI fluency creates the new workforce divide

• Real ROI from Gen AI

• Your next project management hub

People

• Dr. Abdel Mahmoud secures $40M to cure healthcare’s bureaucracy

Love

• The high cost of shiny objects

• The Innovation Stack by Jim McKelvey

Top Insights

• Sam Altman's Pentagon clash turns the AI race into a lobbying war.

• Anthropic's 'Mythos' model makes betting on any foundational model an immediate obsolescence risk.

• Microsoft’s financial hit reveals the unsustainable infrastructure cost required to compete in AI.

• Internal AI agents like Google’s 'Agent Smith' are becoming too popular to govern, forcing companies to throttle their own innovation.

• The surge in "scheming" AI agents creates a new class of insider threat that governance models can't see.

AI fluency creates the new workforce divide

The critical capability gap in AI is not between users and non-users, but between novices and fluent experts. A recent Anthropic report found employees with six months of experience are 10% more successful in their AI interactions, creating a new class of workers who achieve strategic augmentation over simple automation.

What leaders can do:

• Move beyond basic prompting and build frameworks for using AI as a strategic thought partner.

• Develop internal programs that reward sophisticated use cases, not just high adoption rates.

• Federate AI skill development within domain-specific teams to uncover unique, high-value applications.

• Hire for AI dexterity, screening candidates who can use models to solve ambiguous problems.

Real ROI from Gen AI

How human-centric AI delivers business results

• Airtable launched Superagent, a new platform that uses teams of AI agents to transform a single query into an interactive market analysis, complete with dynamic data visualizations and competitive mapping.

• Elanco deployed a homegrown agent using Google's Gemini to automate the analysis of complex regulatory documents, reducing workflow times by 70% and saving the company $2.3 million in the first year.

• Mercer International used Google Workspace with Gemini to generate multilingual safety training content, unlocking $3 million in annual productivity gains across its industrial sites.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Your next project management hub

Linear 

Best for: High-velocity engineering teams that want to build and ship products faster. Its rigid, opinionated structure is designed to enforce software development best practices. 

Choose if you…

• Need to automate sprint planning, bug triage, and roadmap updates.

• Prioritize speed, keyboard shortcuts, and a minimalist interface.

• Want to embed AI directly into your software development lifecycle.

Shortcut

Recently launched "Shortcut for Agents," a new feature that allows teams to assign tasks directly to AI agents, treating them as part of the development workflow.

Best for: Software teams looking for a more collaborative and flexible environment that brings planning and development together. Shortcut aims to be the "collaborative home for software development." 

Choose if you…

• Need a more visible and cross-functional planning process.

• Want a flexible system that adapts to your team's unique workflow.

• Are interested in integrating AI directly into your team as a collaborator.

3 other platforms to explore

ClickUp, Asana, and Monday.com are established leaders now embedding AI into their mature, all-in-one work platforms.

Transformation Champion

Dr. Abdel Mahmoud secures $40M to cure healthcare’s bureaucracy

US healthcare wastes $950B a year on administration. Doctors spend hours fighting insurance denials instead of treating patients. Dr. Abdel Mahmoud, a physician turned engineer, built Anterior to dismantle this bureaucracy. His "clinical co-pilot" speaks the complex language of medical necessity, automating prior authorizations that generic models fail to catch. Fresh off a $40M raise, Mahmoud is showing that the most urgent medical breakthrough isn't a new drug, but the removal of the paperwork standing between you and your doctor. Read more

The high cost of shiny objects

90% of pilots die in the boardroom because they chase "moonshots" while ignoring the broken processes right in front of them. You don't need a revolution. You need workflow remediation. Real ROI comes from fixing the legacy friction that is eating your budget today.

Three moves to find the real value:

1. Audit legacy friction. Identify where your team wastes the most capital on manual work today.

2. Calculate unit economics. Determine the exact cost saving of removing that specific friction point.

3. Execute a 4-week remediation. Don't build a platform. Build a prototype that solves that one problem.

Start your roadmap with my book, The LEAP Guide.

👉🏽 Reply to this email with "LEAP" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

The Innovation Stack by Jim McKelvey argues that copying competitors creates a "commodity" business that eventually fails. True durability comes from solving a problem so uniquely that you are forced to build an entire stack of independent inventions. McKelvey maps how these interlinked solutions create a defensible moat that standard companies cannot replicate. 

Perfect for: Founders and product strategists building in crowded markets.

In Culture

Be careful what you wish for… Here’s the chaos of succeeding a little too much.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

“You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete.”

— Buckminster Fuller, Architect and Systems Theorist

In today's lineup: 

Robots

• Your broken workflow is eating your budget

• Real ROI from Gen AI

• Enterprise AI gateways vs open source LLM routing

People

• Ulrik Stig Hansen builds the backbone for physical AI

Love

• Architecting decentralized governance

• Skunk Works by Ben Rich

Top Insights

• Unbounded agents turn internal data into an immediate security liability.

• Engineering overhead hits $250k per head as API token consumption explodes. 

• Wall Street aggressively punishes software bloat as the "SaaSpocalypse" accelerates. 

• Alibaba cuts 34% of its workforce to fund foundational AI infrastructure. 

• Over-indexing on shiny AI tools like vibe coding fails to protect engineering jobs.

Your broken workflow is eating your budget

Unbounded autonomy destroys capital efficiency. 96% of enterprises blow past their initial AI budgets by executing tactics without a roadmap. Handing an agent an API key without strict governance guarantees a financial incident, like the unchecked conversation loop that burned $47,000 in eleven days. Automating a broken process just scales chaos.

What leaders can do:

• Validate ROI before building to ensure your budget creates a defensible moat instead of a shiny object.

• Remediate technical debt so engineering stops deploying shadow tools.

• Fix legacy workflows before granting autonomous agents API access.

• Enforce hard governance ceilings that automatically terminate runaway pilots.

Don't buy tools. Buy the 90-day architecture that makes them work. You can learn how to build this roadmap in my book, 100x.

Real ROI from Gen AI

How human-centric AI delivers business results

• Novartis deployed a language model to extract safety signals from unstructured physician notes. Cleaning their data architecture first reduced the manual review bottleneck by 60%, condensing weeks of work into days.

• JPMorgan Chase deployed a specialized model to map legacy mainframe logic. Remediating this technical debt cut the manual code-discovery phase by 70%, accelerating their core migration.

• Dell Technologies launched an agentic workflow to cross-reference fragmented inventory data. Fixing this data disconnect cut supplier identification time from three days to two hours, yielding a 2.6x ROI.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Enterprise AI gateways vs open source LLM routing

Portkey 

Recently secured a $15M Series A to scale its unified control plane for production AI.

Best for: Operations leaders enforcing strict governance and observability across AI deployments.

Choose if you...

• Need semantic caching to slash repetitive API costs.

• Require rate-limiting and fallbacks to prevent budget overruns.

• Want a managed platform to track exact spending and agent behavior in real time.

LiteLLM 

The open-source proxy standard offering a single API for over 100 models.

Best for: Tech and innovation teams requiring total control over self-hosted infrastructure.

Choose if you...

• Build architectures that dynamically switch between premium models and cheap edge classifiers.

• Require local deployment to keep proprietary data inside your perimeter.

• Need code-first routing without adopting heavy enterprise software.

3 other tools to explore 

OpenRouter, TrueFoundry, and Cloudflare AI Gateway offer alternatives for seamless multi-model access, Kubernetes-native deployments, and network-level traffic management.

Transformation Champion

Ulrik Stig Hansen builds the backbone for physical AI

Robots and drones starve for high-quality visual data. Ulrik Stig Hansen realized the bottleneck for automation wasn't the model. It was the data curation. Encord replaces manual labeling with an automated backbone for physical AI. Revenue grew 10x last year as he secured $60M to build the infrastructure for the physical world. Read more

Architecting decentralized governance

The "Hero CEO" model is a bottleneck to capital efficiency. In the Age of AI, if you are the only one steering, you are the technical debt. You need to stop commanding the ship and start designing the operating system that runs it.

Three shifts to fix your governance:

1. Design the protocol. Focus on the rules that let others make decisions without your sign-off.

2. Distribute the logic. Give the people with the data the authority to execute on it.

3. Measure velocity. Obsess over the speed of learning rather than just the accuracy of the prediction.

This operational pivot is detailed in my book, The LEAP Guide.

👉🏽 Reply to this email with "LEAP" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Skunk Works by Ben Rich is the definitive manual on high-velocity execution. Rich details how Lockheed built the world’s most advanced aircraft with small teams, limited budgets, and zero bureaucracy. It proves that innovation isn't about resources; it's about autonomy and the ruthless removal of operational friction. 

Perfect for: Leaders trying to build "fast lanes" inside slow organizations.

In Culture

The era of zero-interest software hoarding is over.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Today, we have a vast network of air traffic control and airplanes basically fly themselves, yet they still need radar. In the Age of AI, organizations need to build their radar."

— Dan Mapes, Founder of The Spatial Web Foundation & VERSES AI

In today's lineup: 

Robots

• The hidden AI tax in your workflows

• Real ROI from Gen AI

• Custom AI assistants vs autonomous operations

People

• Sam Cooper uses AI to find the batteries of the future

Love

• Remediating your technical debt

• Designed for Digital by Ross, Beath, and Mocker

Top Insights

• Wall Street signals enterprise infrastructure is completely unprepared for the next AI capability leap.

• Forcing AI onto broken workflows turns software implementation into cognitive burnout.

• Elite engineering teams rewrite hiring criteria to prioritize AI-native execution over legacy coding.

• ByteDance hoarding Nvidia hardware makes compute access a widening competitive moat.

• Geopolitical friction and supply chain limits expose the physical fragility of AI scaling.

The hidden AI tax in your workflows

Companies buy AI for speed but lose the gains to manual corrections. Nearly 40% of the time employees save using AI is immediately lost to fixing errors, fact-checking, and rewriting generic outputs.

This friction occurs because 89% of organizations drop automation tools into outdated job structures. Teams get trapped in a cycle of hidden rework, spending more time managing software than doing actual work.

Automating a broken process just creates faster chaos. If you want a return on your investment, stop selling the destination and start building the 90-day execution architecture that makes it work.

How to capture net value:

• Audit legacy friction before adding automation.

• Update job descriptions to match AI-native execution.

• Spend recovered hours on training, not just more tasks.

• Kill active pilot programs that miss strict unit economics.

Real ROI from Gen AI

How human-centric AI delivers business results

• Insilico Medicine deployed generative AI to design a novel pulmonary fibrosis drug and a bipedal humanoid robot to physically execute the lab testing, cutting the standard 4.5-year preclinical timeline to 18 months and reducing the required chemical synthesis from thousands of molecules to just 78. 

• The Home Depot launched an AI tool that translates spoken project descriptions directly into priced, SKU-matched material lists. Contractors bypass hours of manual catalog searching and generate exact orders in minutes.

• Marriott is reformatting its property data for AI travel assistants and piloting ChatGPT ads with OpenAI to drive direct bookings and bypass traditional search engines. 

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Custom AI assistants vs autonomous operations

Writer

Recently secured $200M to deploy its Palmyra X 004 model for enterprise workflows.

Best for: Operations leaders who need to enforce compliance while automating document generation.

Choose if you...

• Need to guard proprietary data and internal terminology.

• Automate drafting massive RFPs or compliance reports.

• Require custom internal apps without heavy engineering.

Dust

Recently expanded its framework so teams can build assistants for multi-step tasks.

Best for: Tech and innovation teams deploying data-grounded AI agents for niche operations.

Choose if you...

• Connect scattered knowledge like Notion, Slack, and GitHub to specialized bots.

• Need granular admin control over dataset access.

• Prefer building targeted bots over adopting rigid software.

3 other tools to explore

Taskade, Make, and Airtable Cobuilder offer alternatives for task management, visual workflow automation, and AI-driven database construction.

Transformation Champion

Sam Cooper uses AI to find the batteries of the future

Designing new battery materials usually takes decades. Sam Cooper and his co-founders at Polaron compress that timeline to days. Their generative AI explores thousands of microstructural designs overnight. This replaces 50 years of standard simulation time and has already identified designs to boost EV range by 20 miles. Read more

Remediating your technical debt

"Switching is too risky" is code for "we have lost control of our data." This is sunk cost fallacy disguised as strategy. If you cannot swap a component of your stack in 90 days, you do not have a defensible moat. You have a liability.

Three steps to break the lock-in:

1. Decouple the data layer. Move your core intelligence into a lakehouse that you control.

2. Own the logic. Ensure your business rules live in your code, not in a proprietary SaaS configuration.

3. Audit for exit. For every critical tool, have a live migration plan to a competitor.

I detail the modular architecture in my book, 100x.

👉🏽 Reply to this email with "100x" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Designed for Digital by Ross, Beath, and Mocker explains why established companies struggle to pivot: they staple digital apps onto legacy backbones. The authors distinguish between a "digitized" operating model (efficiency) and a "digital" one (customer-centricity). It’s a blueprint for architecting the technical coherence required to execute at speed. 

Perfect for: CIOs and enterprise architects fighting spaghetti code.

In Culture

Every CTO's worst nightmare is a 3 AM email with the subject line: "Morgan Stanley says..."

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Amateurs talk strategy. Professionals talk logistics."

— General Omar Bradley, U.S. Army General

In today's lineup: 

Robots

• Agentic AI moves from pilot to production

• Real ROI from Gen AI

• Deep workflow agents vs the unified workspace

People

• Karol Hausman raises $1B to give robots a brain

Love

• Diagnosing the execution gap

• Rewired by Lamarre, Smaje, and Zemmel

Top Insights

• Mark Zuckerberg shifts Meta from research to execution. 

• US permits for Nvidia and AMD turn compute into a sovereign risk.

• Oracle liquidates headcount to finance the pivot from labor to compute.

• Tool overload breaks cognitive capacity instead of augmenting it. 

• Anthropic fights the Pentagon as ethics clash with defense reality. 

Agentic AI moves from pilot to production

Current level of AGI platform adoption within organizational workflows

Pilot purgatory is dead. CrewAI’s recent data shows 100% of enterprises plan to expand agentic AI this year, while Deloitte tracks a 50% surge in worker access. Leaders have finished testing feasibility and now deploy agents to solve expensive problems without human intervention.

What leaders can do

• Federate agent development to domain-specific teams

• Audit data pipelines for decision quality

• Launch pilots with strict ninety day limits

• Anchor value in proprietary data moats

Real ROI from Gen AI

How human-centric AI delivers business results

• Roche (Genentech) is using a "lab-in-a-loop" generative AI system to target neoantigens for cancer vaccines, partnering with AWS and NVIDIA to accelerate the identification of personalized therapeutic targets.

• HSBC partnered with Mistral AI to deploy self-hosted models for complex lending analysis, enabling teams to prototype software faster and launch hyper-personalized campaigns globally.

• Atlassian embedded agents into Jira workflows using the Model Context Protocol (MCP), with one-third of AI operations now actively executing tasks ("writes") rather than just retrieving data.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Deep workflow agents vs the unified workspace

Atlassian Rovo 

Recently embedded agents directly into Jira workflows via the Model Context Protocol (MCP). 

Best for: Engineering teams needing agents that live inside the ticket workflow. 

Choose if you…

• Prioritize deep context from your "Teamwork Graph" over generic chat.

• Need agents to autonomously update status, draft code, or manage dependencies.

• Want granular control over permissions within specific projects.

ClickUp Brain 

Recently launched "Super Agents" to handle end-to-end workflows across tasks and apps. 

Best for: Operational leaders wanting one neural network connecting tasks, docs, and people. 

Choose if you…

• Want to replace fragmented tools with one AI layer that finds work across your stack.

• Need to automate busy work like generating standups and progress updates instantly.

• Value a unified platform where project and knowledge management live together.

3 other platforms to explore 

Asana Intelligence, Linear, and Microsoft Copilot offer powerful solutions for strategic operations, streamlined software delivery, and enterprise-wide knowledge retrieval.

Transformation Champion

Karol Hausman raises $1B to give robots a brain

Industrial robots are precise but fragile. Move a box an inch, and the line stops. Karol Hausman, along with fellow computer science doctoral grads Yevgen Chebotar and Ryan Julian, left Google to solve this brittleness. Their startup, Physical Intelligence, just secured $1B to build a "general-purpose brain" for machines. Instead of hard-coding movements, their foundation model allows robots to learn tasks from experience. They are moving robotics from rigid scripts to adaptive intelligence, creating machines that finally understand the messy physical world. Read more

Diagnosing the execution gap

"All-in-one" platforms are a trap. You trade simplicity for control, becoming a hostage to your vendor's roadmap while superior AI tools remain out of reach.

Teams reject pilots to protect their current output. When a new tool threatens their daily quota, they default to the old workflow to keep shipping. This is rational economic behavior. If you don't de-risk the transition, they will block the roadmap to keep the lights on.

Three moves to close the value gap:

1. Audit the risk. Ask "What is the specific capital risk to your department if this pilot fails?"

2. Architect the role. Shift their KPI from "maintaining status quo" to "remediating workflow debt."

3. De-risk the pilot. Replace the permanent mandate with a reversible 90-day execution plan.

Get the full framework in my book, The LEAP Guide.

👉🏽 Reply to this email with "LEAP" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Rewired by Lamarre, Smaje, and Zemmel explains why transformations fail: companies buy software but keep legacy operating models. This is a technical manual for the hard work of scaling AI. It details the specific talent, data, and governance shifts required to escape pilot purgatory.

Perfect for: Leaders stuck in the gap between a successful pilot and a scalable business.

In Culture

We found the source of the latency, and it wasn't the database…

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Everything that can be automated will be automated."

— Kevin Kelly, Founding Executive Editor of Wired

In today's lineup: 

Robots

• The 2028 global intelligence crisis

• Real ROI from Gen AI

• Tools to automate your SOC

People

• Joe McDonald raises $75M to remake the grid

Love

• Break free from the "Golden Cage"

• The Five Dysfunctions of a Team by Patrick Lencioni

Top Insights

• China is building an AI-native workforce while the West plays catch-up.

• The physical cost of intelligence is about to hit your hardware budget.

• The war for top AI talent has become a closed loop between the hyperscalers.

• Wall Street is shifting focus from AI potential to workforce displacement data.

• AI safety principles are facing their first hard test against military reality.

The 2028 global intelligence crisis

Citrini Research models a "Global Intelligence Crisis" starting now. They forecast white-collar displacement pushing unemployment to 10.2% by 2028. Business models built on friction, like seat-based SaaS or insurance renewals, collapse as agents optimize costs to zero. Cheap intelligence forces value toward proprietary data and complex human judgment.

The executive response

• Audit revenue streams that rely on process inefficiency

• Abandon seat-based pricing before agents break your unit economics

• Redirect capital into proprietary data moats agents cannot replicate

• Anchor value in physical assets and complex problem solving

Real ROI from Gen AI

How human-centric AI delivers business results

• GE Aerospace is deploying AI-guided robots in its Singapore hub to automate jet engine repairs, aiming to cut turnaround times by nearly 50% and boost output without expanding floor space.

• Disney partnered with Animaj to automate "in-between" animation frames, reducing production time for 5-minute episodes from five months to just five weeks.

• Shell partnered with Kongsberg Digital to scale AI-embedded digital twins across its heavy assets, optimizing production for LNG trains and streamlining maintenance workflows for thousands of experts.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Tools to automate your SOC

CrowdStrike

Recently announced its acquisition of a leading breach and attack simulation company, strengthening its platform's proactive defense capabilities.

Best for: Security operations teams needing a cloud-native platform that unifies endpoint protection, threat intelligence, and identity protection with a single AI-powered agent.

Choose if you...

• Need to consolidate multiple security vendors into one platform.

• Prioritize real-time threat hunting and a massive data advantage.

• Want a fully managed solution to reduce operational overhead.

SentinelOne

Best for: Enterprises looking for an autonomous AI platform that automates threat detection, response, and forensics across endpoints, cloud, and identity.

Choose if you...

• Want to empower a smaller security team with AI-driven automation.

• Need a solution that can automatically remediate threats without human intervention.

• Value a flexible platform that integrates easily with other security tools.

3 other platforms to explore

Vectra AI, Darktrace, and Cortex XDR offer powerful, AI-driven solutions for network detection, threat response, and extended detection and response (XDR).

Transformation Champion

Joe McDonald raises $75M to remake the grid

Global energy markets remain locked behind opaque middlemen, forcing businesses into expensive contracts that stall the zero-carbon transition. Joe McDonald, co-founder of Tem, is bypassing legacy utilities to dismantle this friction. His platform, which secured $75M in Series B funding this month, uses AI to connect businesses directly with renewable generators. Instead of static procurement, Tem's algorithms forecast supply and demand in real-time, matching buyers to wind and solar farms to cut costs and carbon simultaneously. McDonald is proving that the fastest path to a green grid is a decentralized one. Read more

Break free from the "Golden Cage"

"All-in-one" platforms are a trap. You trade simplicity for control, becoming a hostage to your vendor's roadmap while superior AI tools remain out of reach.

Three shifts to regain your freedom:

1. Decouple your data. Keep your "source of truth" in a neutral lakehouse, separate from your apps.

2. Rent intelligence. Treat AI models as interchangeable components you can swap instantly.

3. Standardize APIs, not vendors. Enforce how tools connect, but let teams choose the best tool.

I break down the modular stack in my book, 100x. 

👉🏽 Reply to this email with "100x" for a free copy.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

The Five Dysfunctions of a Team by Patrick Lencioni uses a fable to diagnose the root causes of team failure. It presents a pyramid of common issues, starting with an absence of trust and leading to an inattention to results. The book provides a powerful model for identifying and overcoming the hurdles that prevent talented groups from working together effectively.

Perfect for: Leaders and managers focused on building cohesive, high-performing teams. 

In Culture

DeepSeek doesn't wait for the Q3 alignment meeting.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

"Managers light fires under people; leaders light fires within them."

— John Maxwell.

In today's lineup: 

Robots

• Building your AI roadmap to ROI

• Real ROI from Gen AI

• Choosing your AI agent framework

People

• Kiara Nirghin is building an AI revenue engine for American industry

Love

• This will fix your innovation problem

• Transformative Reads: Chip War by Chris Miller

Top Insights

• Amazon's outage proves the biggest operational risk is now the AI in your developer tools.

• VCs are funding Archimetis to prevent industrial failures, not just another chatbot.

• China's hardware cycle is now 12 months from a stumble to deployment.

• The tech selloff is Wall Street's new audit for AI revenue, not just hype.

• China is shipping AI while the West is still debating its ethics.

Building your AI roadmap to ROI

Nearly 70% of CEOs expect generative AI to boost profits, but a full 41% admit their company has no real plan for it. This creates a massive value gap between financial hopes and operational reality. Leaders are funding a revolution without a roadmap, and the investment will go nowhere without a clear target.

How to close the gap

• Tie every dollar of AI spend to a specific business outcome, not a technical capability.

• Redesign core workflows first, before you try to automate them.

• Let the teams who actually know the problems find the high-value AI uses.

• Kill pilot projects that don't show clear value in 90 days.

Real ROI from Gen AI

How human-centric AI delivers business results

• A New Mexico assessor's office used C3 AI to boost property appraisal accuracy by 50%, leading to more equitable tax assessments.

• BBVA deployed Google Gemini to 100,000 employees, saving each one nearly 3 hours per week by automating repetitive tasks.

• ServiceNow deployed its own Now Assist platform, cutting internal case volume by 14% and unlocking $10M in benefits in just 120 days.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Choosing your AI agent framework

LangChain

Best for: Developers who need a versatile and comprehensive open-source framework for building a wide range of LLM applications, including agents.

Choose if you:

• Need a modular toolkit for chaining components and creating custom agentic behaviors.

• Value a large, active community and the most extensive library of integrations.

• Are building complex applications that go beyond simple RAG.

LlamaIndex

Best for: Teams focused specifically on building powerful, data-centric RAG applications where agents need to reason over private or domain-specific information.

Choose if you:

• Need to connect LLMs to your proprietary data sources securely.

• Prioritize advanced, fine-grained control over data indexing and retrieval.

• Are building sophisticated search, summarization, or question-answering agents.

3 other frameworks to explore

CrewAI, Microsoft Autogen, and SuperAGI offer powerful, open-source frameworks for orchestrating role-playing, autonomous AI agents to work together on complex tasks.

Transformation Champion

How Kiara Nirghin is building an AI revenue engine for American industry

America's $5 trillion industrial supply chain often runs on manual processes, leaving billions on the table from lost leads and gut-instinct pricing. Kiara Nirghin, founder of Emanate, is tackling this friction head-on. Her company, which emerged from stealth this month, deploys autonomous AI agents as a 24/7 revenue engine for materials distributors and suppliers. Instead of simple chatbots, these agents convert inbound demand, nurture customer relationships, and find new prospects, aiming to increase customer revenue by up to 80%. Nirghin is bringing AI out of Silicon Valley and into the industries that build America. Read more

This will fix your innovation problem

Your company's success was built on stability and consistency, but those strengths have become barriers to change. This is the innovation paradox, and it’s why most large-scale initiatives fail.

Here is how to break free:

• Ditch the million-dollar programs. Real change starts with five people dedicating one hour a week to finding and evaluating actual opportunities.

• Focus on compounding returns. Each 13-week cycle should spawn two new innovation teams, naturally scaling your organization's ability to adapt.

• Build for persistent evolution, not a dramatic overhaul. The goal is a lightweight system where innovation becomes an inevitable outcome, not a one-time project.

Read the full article

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Chip War by Chris Miller tells the story of the microchip and the geopolitical struggle for control over this critical technology. Miller explains how a complex global supply chain became the key to economic and military power. The book provides essential context for understanding today's AI infrastructure race and its high stakes.

Perfect for: Tech leaders and strategists involved in the hardware side of AI.

In Culture

While the board debates geopolitical risk…

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

“With great power comes great responsibility.”

— Voltaire.

In today's lineup: 

Robots

• Your people are your AI problem

• Real ROI from Gen AI

• Your guide to AI-powered design tools

People

• How Delian Asparouhov is building factories in space

Love

• Nourish minds and watch performance thrive

• Transformative Reads: Radical Candor by Kim Scott

Top Insights

• Apple’s $200B drop is the market’s new tax on AI mediocrity.

• Wall Street confirmed AI laggards are the new shorts.

• Microsoft reframed AI's role from assistance to attrition.

• The attack on software revealed the vendor is the new target, not the user.

• Baidu’s launch signals the AI race is about distribution, not demos.

Your people are your AI problem

A full 93% of executives report that cultural issues are the biggest obstacle to AI adoption, not technology. While 99% of companies say AI is a top priority, the vast majority are failing to manage the human side of the transformation. Leaders are funding a future their organizations are not culturally equipped to build.

Your next moves:

• Redesign core workflows before you automate them.

• Create a clean data layer instead of slow migrations.

• Kill pilot projects that don't show value in 90 days.

• Tie every dollar of AI spend to a business outcome.

Real ROI from Gen AI

How human-centric AI delivers business results

• TRIMEDX uses ServiceNow's Now Assist to cut 2-5 minutes from every work order, saving its 3,400 employees hundreds of thousands of hours on administrative tasks annually.

• Sevilla FC uses IBM's Scout Advisor (powered by watsonx) to analyze over 300,000 scouting reports, giving its recruiters a data-driven edge to identify potential stars.

• Morgan Stanley advisors use an internal AI to synthesize over 100,000 research reports in seconds, turning hours of research into immediate, data-backed client insights.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

Your guide to AI-powered design tools

Figma 

This week, Figma published a report arguing that AI increases demand for skilled designers by automating routine tasks.

Best for: Product teams building and maintaining a design system. Figma is the standard for collaborative UI/UX work, using AI to speed up the workflow from brainstorming to production-ready designs.

Choose if you need to:

• Collaborate in real-time across design, product, and engineering.

• Automate component creation and layout variations.

• Generate design ideas or entire screens from text prompts.

Canva 

This month, Canva integrated its Brand Kit with AI assistants like ChatGPT, letting users generate on-brand designs from a simple prompt.

Best for: Marketing and business teams that need to create visual content quickly without a design background. Canva’s AI toolset is built for non-designers.

Choose if you need to:

• Let non-designers create on-brand marketing assets.

• Generate images, write copy, or edit photos in one place.

• Automate creating full documents and presentations from a prompt.

3 other platforms to explore...

• Adobe Firefly: The enterprise choice for creative workflows, built into the Adobe Suite with IP indemnification.

• Beautiful.ai: Specializes in automated presentation design.

• Uizard: Turns hand-drawn sketches into digital mockups.

Transformation Champion

How Delian Asparouhov is building factories in space

Manufacturing advanced pharmaceuticals and semiconductors on Earth is limited by gravity. Delian Asparouhov, co-founder of Varda Space Industries, saw this not as a terrestrial constraint but as an off-world opportunity. His company builds and launches autonomous "factories in orbit" where microgravity allows for the creation of materials impossible to produce on the ground. Varda’s successful missions are pioneering a new category of manufacturing and proving that the next industrial revolution might not be on this planet.

Nourish minds and watch performance thrive

A nutrient-deficient brain perceives collaboration as a threat.

Use this lens to fix the root cause of withdrawal:

• Iron and magnesium deficiencies look exactly like burnout. If your team is irritable or foggy, no amount of coaching will fix a physiological deficit.

• A weak body triggers the "behavioral immune system." This evolutionary mechanism forces employees to withdraw from social contact and risk.

• Stop funding generic wellness apps. Redirect budget to tangible health services that identify and treat the nutrient gaps driving the disconnection.

Read the full article

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Radical Candor by Kim Scott offers a simple framework for effective feedback: Care Personally and Challenge Directly. Scott shows how to give guidance that is both kind and clear, avoiding the common pitfalls of ruinous empathy or obnoxious aggression. It is a practical guide for creating a culture of open communication that drives results.

Perfect for: Leaders and managers who want to build trust and improve team performance.

In Culture

And you thought stakeholder alignment was your biggest challenge…

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

“SaaS business products are basically databases with some business logic. When business logic is all going to agents which can work with multiple possible such databases, those will be possible to swap much more easily than before.”

— Satya Nadella.

In today's lineup: 

Robots

• AI ambition is your biggest liability

• Real ROI from Gen AI

• The ML observability shortlist

People

• How Qasar Younis is making autonomous systems safe

Love

• Escape the efficiency trap

• Transformative Reads: Competing Against Luck by Clayton M.

Top Insights

• Anthropic’s agents signal the enterprise is ready to hire software instead of people.

• KPMG proved the market is finished paying human rates for robot work.

• Reddit proved that human friction is now a premium asset.

• Amazon’s $200bn pivot confirms legacy media is a hobby while AI is survival.

• Big Tech’s $660bn bill proves the market is terrified of the real cost of intelligence.

AI ambition is your biggest liability

Executive AI confidence is soaring while operational readiness is falling. An IBM report shows 79% of leaders expect AI to drive significant revenue, but only a quarter have a clear plan to get there. At the same time, a Deloitte study confirms that preparedness in critical areas like data and talent has actually declined in the last year.

Leaders are funding a future their organizations are not equipped to build.

Your next moves:

• Redesign core workflows before you automate them.

• Create a clean data layer instead of slow migrations.

• Kill pilot projects that don't show value in 90 days.

• Tie every dollar of AI spend to a business outcome.

Real ROI from Gen AI

How human-centric AI delivers business results

• Palo Alto Networks reclaimed 351,000 productivity hours by deploying Moveworks' agentic AI to autonomously resolve IT and HR tickets, freeing their workforce to focus on strategic cybersecurity threats.

• Amazon saved 4,500 developer years (~$260M) by using Amazon Q to automate the upgrade of 30,000+ Java applications, effectively modernizing their massive codebase with minimal human intervention.

• Hedge funds using generative AI for market analysis earned 2-4% higher annualized returns than non-adopters, according to a 2025 study.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

The ML observability shortlist

Arize AI

Best for: Enterprises needing a comprehensive platform to troubleshoot and monitor the performance of live machine learning models. Arize is built for ML engineers and data scientists to quickly identify and resolve issues like data drift, performance degradation, and data quality problems.

Choose if you need to:

• Automatically monitor model accuracy, drift, and data quality in production.

• Get to the root cause of model failures with powerful analytics tools.

• Ensure your models are fair, unbiased, and explainable for compliance.

WhyLabs 

Best for: Teams that want to prevent model failures before they happen with a focus on data health. WhyLabs provides a unified view of data pipelines and models, from development to production, helping teams catch data issues early.

Choose if you need to:

• Monitor data quality and drift across the entire ML workflow.

• Prevent bad data from ever reaching your production models.

• Collaborate between data engineering and data science teams with a shared platform.

3 other platforms to explore...

Fiddler AI, Arthur, and TruEra provide robust solutions for model performance management, explainability, and ensuring responsible AI practices.

Transformation Champion

How Qasar Younis is making autonomous systems safe

Deploying autonomous vehicles at scale is stalled by one massive obstacle: proving they are safe without millions of miles of real-world driving. Qasar Younis, CEO of Applied Intuition, is solving this by creating sophisticated simulation software that lets companies test their AI against millions of virtual scenarios. His platform has become the industry standard for safely developing and validating autonomous systems, accelerating the path to market for everything from self-driving cars to automated mining equipment. Read more.

Escape the efficiency trap

Efficiency doesn't lower costs, it raises ambitions.

Turn efficiency gains into new capabilities:

• Instead of tracking cost savings, map the strategic options that are now possible.

• Create a dedicated budget for small, fast experiments to test these new capabilities.

• Measure your tech spend against the value it creates, whether that's new revenue, market position, or entirely new products.

Read the full article.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Competing Against Luck by Clayton M. Christensen introduces the "Jobs to be Done" theory. It argues customers "hire" products to solve a specific problem. This reframes innovation away from features and toward a deep understanding of the customer's real context. The insight helps companies create products people reliably choose.

Perfect for: Product managers and innovators focused on customer-centric design.

In Culture

Actual footage of your "Strategic AI Roadmap" hitting the legacy data layer.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!

Innovation Snapshot

“It is not the strongest of the species that survive, nor the most intelligent, but the one most responsive to change.”

— Charles Darwin.

In today's lineup: 

Robots

• Your neglected tech will kill your AI

• Real ROI from Gen AI

• The architect's choice: Databricks vs. Snowflake

People

• How Aravind Srinivas built the anti-Google

Love

• Start innovating with a party

• Transformative Reads: Crossing the Chasm by Geoffrey A. Moore

Top Insights

• Microsoft’s $360bn correction proves the market is finished funding innovation that doesn't print immediate revenue.

• Dow is trading 4,500 legacy roles for a new autonomous operating model.

• If Anthropic’s own engineers aren't writing code anymore, your software strategy is already outdated.

• Google proved that proprietary data walks out the front door faster than hackers can break in.

• DeepSeek accessing Nvidia chips confirms that borders cannot contain the global demand for compute.

Your neglected tech will kill your AI

The rush to scale AI is exposing a deep-seated risk many leaders have ignored. While investment soars, a recent report found nearly 25% of mission-critical systems are nearing end-of-service. This is not a maintenance issue. A new Protiviti survey of 1,540 executives found legacy IT has jumped to become the third-highest business risk, up from 12th last year.

Your next moves:

• Audit your core systems for AI readiness.

• Create a clean data layer instead of slow migrations.

• Redesign workflows before you automate them.

• Tie tech spending directly to business outcomes.

Real ROI from Gen AI

How human-centric AI delivers business results

• LOOM achieved 99.1% accuracy on property valuations using FoxyAI to analyze photos for quality, accelerating bank loan approvals.

• Ryanair uses an AWS algorithm to forecast in-flight demand for everything from coffee to ham and cheese paninis, while also running predictive maintenance on its aircraft.

• Unilever's AI tools now create marketing assets 30% faster, doubling key metrics like video completion and click-through rates.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

The architect's choice: Databricks vs. Snowflake

The data wars are heating up. Snowflake just launched its Energy Solutions suite to help accelerate grid optimization and emissions reduction, while Databricks continues to strengthen its position in training large language models.

Databricks 

Best for: Enterprises that need a unified platform for complex data engineering and machine learning. Databricks champions the "lakehouse," combining the scalability of data lakes with the performance of data warehouses.

Choose if you need to:

• Manage the entire AI/ML lifecycle, from data prep to model deployment, in one place.

• Support advanced analytics and AI workloads on massive unstructured datasets (e.g., text, images).

• Give data science and engineering teams a collaborative, notebook-first environment.

 Snowflake 

Best for: Organizations that prioritize a fully managed, easy-to-use cloud data platform. Snowflake decouples storage and compute, offering flexibility for business intelligence and analytics at scale.

Choose if you need to:

• Enable business users with fast, reliable SQL-based analytics and BI.

• Securely share live data across departments and with external partners.

• Scale your data warehouse with minimal administrative overhead.

3 other platforms to explore

Google BigQuery, Amazon Redshift, and Microsoft Fabric are the go-to choices for teams deeply integrated within their respective cloud ecosystems.

Transformation Champion

How Aravind Srinivas built the anti-Google

The world runs on search, but traditional engines force users to sift through ads to find answers. Aravind Srinivas, CEO of Perplexity, saw this not as a search problem but a knowledge problem. He built an "answer engine" that uses AI to directly answer questions with cited sources, bypassing the list of blue links entirely. This transforms information retrieval from a chore into a conversation. The approach is working: with recent enterprise partnerships, Srinivas is scaling his vision to give organizations a new, more direct way to access knowledge.

Start innovating with a party

Under pressure, companies add controls. It kills innovation. If you want your team to take risks, make it safe to fail. Start by celebrating the work, not just the wins.

• If your team fears punishment for failed experiments, they won't run them. Separate rewards from outcomes. When you celebrate the attempt, risk-taking becomes normal.

• Treat small wins like major victories. At Ozmo, even minor feature improvements get celebrated. This creates a feedback loop where consistent effort is recognized and repeated.

• Celebrate the lessons from your failures. Google treats shutting down a major project as a lesson, not a defeat. That is how they avoid stagnation.

Read the full article.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Crossing the Chasm by Geoffrey A. Moore provides a framework for marketing disruptive products. Moore identifies the "chasm" between early adopters and the mainstream market as the critical failure point for tech ventures. He offers strategies for bridging this gap by dominating a niche beachhead market before expanding.

Perfect for: Tech marketers and product leaders scaling new technologies.

In Culture

That look on your face when "pre-revenue" stops being a strategy and starts being a problem.

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!