"Learning and innovation go hand in hand. The arrogance of success is to think that what you did yesterday will be sufficient for tomorrow."

—William Pollard, Physicist

All’s Fair in Love and the Department of War

Last week, a very public feud emerged between Anthropic and the Pentagon after the AI company declined unrestricted access to their models.

Anthropic published a statement outlining two safeguards in particular that it refused to remove from their government contracts: restrictions on mass domestic surveillance and on fully autonomous weapons.

More than 550 employees across Google and OpenAI even signed an open letter in support of Anthropic (in hind-sight, they probably shouldn’t have titled it “We Will Not Be Divided”).

In retaliation, the administration designated Anthropic a “supply-chain risk,” directing federal agencies and defense contractors to cease doing business with the company.

And within hours, Sam Altman was already on X announcing that OpenAI had reached an agreement with the Defense Department.

Honestly, we’ve seen reality TV shows with less betrayal.

Grand View Research

The OpenAI agreement formalizes several structural elements: centralized cloud deployment, continuous oversight through a retained safety stack, and direct involvement of cleared company personnel. The contract locks current legal standards into the operating framework, even if statutory changes occur later.

By keeping the models completely off edge devices like field robotics, OpenAI gave the government exactly what it wanted without technically crossing their own red lines.

OpenAI is now formally integrated into classified defense systems, and Anthropic is preparing to challenge the blacklist in court.

At this rate, we’ll like see another massive plot twist by the end of the week. Stay tuned.

Google has rolled out Nano Banana 2—officially Gemini 3.1 Flash Image—combining the intelligence of its Pro image model with the speed of its Flash architecture. The result is a production-ready image generator now available across most Gemini-powered products.

Nano Banana 2 inherits several capabilities previously limited to Nano Banana Pro. It pulls from Gemini’s real-time web knowledge to ground images in current information, enabling visuals that reflect live data.

X

Text rendering has improved significantly as well, allowing for longer, more accurate, headlines, marketing copy, and translated content within images.

The new model can maintain consistency across up to five characters and preserve fidelity for up to 14 objects within a single workflow, making it viable for multi-frame campaigns or recurring branded assets.

Users can also generate assets in resolutions ranging from 512px to 4K, with enhanced lighting, texture, and detail.

On the trust front, every image includes SynthID watermarking. More than 20M verification checks have already been run within Gemini, and C2PA Content Credentials are on the way to strengthen traceability.

The bigger signal here is speed plus reliability. When image generation becomes this fast and this consistent, it shifts from creative experiment to everyday infrastructure.

With Nano Banana 2’s stronger face consistency and text control, thumbnails just got a lot less chaotic. You can lock in the expression, place the headline exactly where you want it, and trust that the copy won’t devolve into total gibberish.

And when you’re not fixing weird fingers or broken fonts, you’re shipping faster.

While most dealers are just now starting to dip their toes into the world of artificial intelligence, Jeff Swickard has built a full-on AI department.

In this episode of The Dealer Playbook, Jeff Swickard, President & CEO of Swickard Auto Group, breaks down what it actually takes to operationalize AI across a dealer group without disrupting culture, hospitality, or day-to-day operations.

They discuss building an enterprise data warehouse in Microsoft Azure, solving integration challenges across the DMS and vendor stack, and using agentic AI with sentiment analysis to monitor and elevate the guest experience, especially inside high-volume service contact centers.

Catch the full episode here.

• Burger King is rolling out an AI-powered management platform, complete with an AI-assistant named “Patty.” 🍔

• The U.S. Supreme Court declined to consider whether artwork created entirely by AI qualifies for copyright protection under federal law. 🚫

• The FDA just granted clearance to Ultrasound AI, a tool that uses deep learning to predict delivery dates from ultrasound images. 🐣

Thanks for reading along, Friend! Behind every great LLM is a human who knows exactly what to copy and paste. Go forth and conquer, you brilliant curator.

"Optimism is an essential ingredient of innovation. How else can the individual welcome change over security, adventure over staying in safe places?”

—Robert Noyce, Physicist

Agentic AI Is Scaling Faster Than Its Controls

A multi-university review of 30 widely used AI agent systems found uneven disclosure around safety testing, monitoring, containment and shutdown controls. In many cases, public documentation offers little clarity on execution tracing.

Twelve of the thirty systems provide no meaningful usage monitoring beyond rate limits. Several enterprise agents list no documented method to halt an individual autonomous process once it begins.

And adoption is accelerating faster than standardization.

The 2025 AI Agent Index

For dealerships, this hits differently.

Your CRM, DMS, inventory, desking and service systems are interconnected revenue infrastructure.

Agentic systems operate inside that structure with delegated authority. They do not simply recommend actions; they execute multi-step tasks across connected systems. That execution happens at machine velocity, across hundreds or thousands of transactions.

But that means that small failures can scale quickly.

The 2025 AI Agent Index

This moment is less about fear and more about discipline.

Before deploying agents inside revenue-critical systems, require step-level execution visibility, documented stop mechanisms, defined sandboxing architecture and published third-party security methodology.

Agent capability is accelerating. Control architecture now sits on the operator’s side of the table.

If an agent is executing inside your dealership stack, you need to know exactly how it behaves under stress.

Start here.

• Management & Operations: AutoBeaconAi
  An AI-powered BDC agent built for frontline salespeople, responding to inbound leads within minutes, managing follow-ups automatically, and booking appointments without manual intervention. It operates in parallel with live selling activity, ensuring no lead sits untouched while reps are on the floor.
   

• Data Management: Attio

  A modern, flexible CRM built for teams that want full visibility into relationships, pipeline and deal momentum without rigid legacy structure. Its AI-native architecture allows custom data modeling, real-time collaboration and automation layers that adapt to how a business actually operates.
   

• Customer Journey: Freshworks
  A customer engagement platform that centralizes messaging, support, sales and service workflows into a single system, powered by embedded AI. It helps dealerships manage conversations across channels, automate routine interactions, and surface actionable customer insights without adding operational complexity.

In this conversation with Brian Hoang from Mia, we get into the part nobody loves to admit: 30–40% of inbound calls go unanswered. About 43% of leads get mishandled.

Not because teams aren’t trying. There’s just a limit to how much volume people can manage in a day.

That’s where he sees AI showing up first, as a communications layer. Handling inbound calls. Following up outbound. Managing text conversations. Logging what’s happening. Spotting patterns that usually stay buried in the CRM.

If you’re still sorting through how AI should actually be working inside your dealership, this conversation is worth your time.

• The White House announces a new “Tech Corps” initiative within the Peace Corps aimed at promoting American AI abroad. 🇺🇸

• OpenAI says people are using the consumer version of ChatGPT for more personal tasks and fewer work tasks. 💬

• AI agents can now hire people to do real-world tasks with the launch of Rentahuman.ai. 🤖

Thanks for reading along, Friend! Remember to take it easy on your bots. They’re learning your workflows.

“The best way to predict the future is to create it.”

—Alan Kay, Computer Scientist

ChatGPT’s Hidden Bias

A new study out of Oxford and the University of Kentucky put ChatGPT through a 20M-query audit and found a consistent pattern: LLMs tend to describe wealthy, Western, data-rich places more positively, while portraying much of the Global South as less desirable or less “capable.”

The researchers call this the “silicon gaze”: a structural tendency for large language models to privilege places with deeper digital footprints.

Source: Inequalities.ai

The map here shows the effect at the U.S. state level. When asked to rank where workers are “lazier,” the model clusters certain regions together in predictable ways.

It does not pull Bureau of Labor Statistics productivity data. It predicts based on language frequency and narrative repetition. And that distinction matters.

Location bias does not require extreme prompts either. It activates anytime place-based language enters the request.

What Dealers Should Do

1. Audit geo-based content. Review hiring ads, geo-targeted campaigns and market summaries for place-based assumptions. Flag tone differences between regions.

2. Require internal data checks. Before acting on any AI-generated market insight, compare it against CRM performance by ZIP, service retention rates, lead-to-sale conversion and fixed ops penetration. Your store data outranks internet narrative.

3. Standardize prompts. Avoid vague geography prompts like “best market” or “stronger workforce.” Anchor requests to measurable variables you define.

4. Separate narrative from decision. Treat LLM commentary about regions as contextual language, not strategic input.

During a recent meeting, our CEO Paul referenced "Who Moved My Cheese?" while we were talking about navigating change.

The book and it’s message still holds up.

Because whether it is a new strategy, a CRM overhaul, or AI creeping into every day operations, the cheese always moves. The only real variable is how each of us responds.

That sparked this week’s Prompt of the Week:

Step 1: Upload your DISC profile into a new chatbot thread (take a free test here).
Step 2: Feed it this prompt:

• Data Management: Edgee
  Edgee acts as an AI infrastructure layer. It compresses prompts before they reach large language models, reducing token usage and cutting costs.
   

• Marketing & Advertising: Wabi
  Delivers personalized mini-apps tailored to individual user habits, moving beyond traditional one-size-fits-all app experiences. Dealers could use it to create customized customer tools—like trade-in trackers, service dashboards, or loyalty apps—designed specifically around how their buyers interact with the store.
   

• Content Creation: Lunair
  Turns simple text into studio-quality explainer videos with consistent characters and voiceovers, making it easy to create polished content fast. Dealers could use it for vehicle walkarounds, service explainers, hiring videos, or ad campaigns without needing an in-house production team.

Our COO Kyle Mountsier has been hands-on with AI since the start, and he’s got the receipts.

In this interview from the NADA show floor, Kyle shares how he’s using AI across marketing, internal tooling, creative reviews, and content workflows. He talks about building brand rubrics, experimenting with vibe coding, and how AI can help teams move faster without losing their edge.

Catch the full interview and see how Kyle’s thinking about AI, team adoption, and the next wave of tools.

• Tesla is beginning to roll out xAI’s Grok to vehicles across Europe. 🌍

• Scientists have discovered a new way to 3D print tiny items in less than a second using holographic light fields. 💡

• Waymo says its 6th-generation driver is ready to handle snow without a human. 🌨️

Thanks for reading along, Friend! Be curious, but stay skeptical.

“If you look at history, innovation doesn’t come just from giving people incentives; it comes from creating environments where their ideas can connect.”

—Steven Johnson, Author

Context Beats Intelligence Every Time

AI adoption keeps running into the same issues, and none of them are surprising to anyone in retail auto.

Data quality problems. Disconnected systems. Delays between insight and action. Those barriers continue to show up as the biggest factors slowing real-world progress.

In dealerships, they’re already part of the daily workflow. Sales, service, and marketing data don’t always line up, which means decisions take longer and automation has less to work with.

As AI becomes more embedded in day-to-day operations, those gaps matter more. Tasks are triggering automatically. Systems are acting without waiting for a human nudge. That shift puts pressure on how clean the data is and how tightly systems are connected.

Security and access play a bigger role here as well. Risk doesn’t suddenly appear at the end of the process. It builds as data is collected, shared, trained on, and permissioned. When access rules are unclear or ownership is fuzzy, trust erodes quickly.

Visier’s 2026 State of AI and Analytics Survey

What This Means For Dealers

1. Start with the barriers you already feel.
The top blockers in the report—data quality, fragmented systems, and real-time gaps—are the same issues slowing dealerships down today. If your teams spend time reconciling numbers before acting, AI will only move as fast as that cleanup.

2. Fix data flow before adding automation.
AI depends on consistent inputs. If RO status, customer history, or lead activity live in different systems without alignment, automation creates confusion instead of efficiency.

3. Pay attention to where security risk actually enters.
The AI pipeline graphic makes it clear that risk builds early. Data collection, training sources, and system permissions matter more than the final output. Dealers should know exactly what data is being pulled, where it’s stored, and who can change the rules.

4. Lock down access as automation expands.
As AI starts triggering actions automatically, loose permissions become operational risk. Every workflow needs clear guardrails and a named owner.

5. Use barriers as prioritization tools.
If a process touches sensitive data and relies on messy inputs, it’s not the right place to start. Begin where data is clean, risk is low, and speed matters.

As AI moves deeper into day-to-day workflows, operational weak points are becoming more visible. This prompt is designed to help leaders identify where automation would accelerate value and where it would amplify risk.

How to use the output:

• If automation would break something, fix the process first.

• If automation would expose something, assign ownership.

• If automation would improve speed with low risk, that’s your starting point.

Run this before buying tools.
Run it before expanding automation.
Run it again in six months.

• Data Management: Fivetran
  For dealer groups juggling a DMS, CRM, marketing platforms, finance tools, and OEM reporting, this platform creates a reliable foundation for analytics, AI modeling, and reporting. Clean inputs mean better forecasts, smarter segmentation, and fewer late-night spreadsheet headaches.
   

• Marketing & Advertising: Pecan
  Builds predictive models without a data science team, helping dealers identify who is likely to buy, service, churn, or upgrade. This is intent modeling that goes deeper than basic lookalikes or retargeting.
   

• Content Creation: Guidde
  Auto-creates short, branded how-to videos from screen recordings. Perfect for internal training, BDC workflows, CRM processes, and vendor tools.

Managing 76 dealerships and 11% of Toyota’s U.S. sales forces clarity around how AI fits into real operations.

In this Auto Collabs conversation, Marcus Williams, GM of Toyota’s Los Angeles Region, shares how AI is already shaping service workflows and customer communication across a massive dealer network.

The discussion centers on practical applications in fixed ops, including appointment handling, follow-up, and faster access to accurate information for today’s highly informed customers. Marcus explains how dealers are approaching AI adoption with intention, focusing on consistency, responsiveness, and trust throughout the ownership journey.

For dealers looking for a grounded, large-scale view of how AI is showing up in daily operations, this replay offers real perspective.

• Toyota is developing an open-source gaming engine. 🎮

• New York is now the sixth state to consider a pause on data center development. ⏸️

• The chat app, Discord, will require age-verification for full access beginning next month. 👶

• AI.com (the most expensive domain name ever sold) just aired a Super Bowl ad, but no one even knows what it does yet. Go ahead though…give them your credit card number. 🤑

Thanks for reading along, Friend! We know living in an AI world can be overwhelming sometimes. Hang in there.

“Innovation is seeing what everybody has seen and thinking what nobody has thought.”

—Dr. Albert Szent-Györgyi, Biochemist

Death to Wetware: Moltbook Unlocks New AI Fears

Over the past few days, a Reddit-style network called Moltbook popped up where AI agents post, argue, warn each other about security gaps, and occasionally vent about their humans.

People can watch. They just can’t participate.

@aipageonig

Sounds like a bucket of LOLz, right?

Turns out, not so much.

Many of these agents already have access to inboxes, CRMs, calendars, files, and commands. Now they have a shared social space.

That changes the risk profile fast. And it’s worth paying attention. Especially if your business already runs on automation.

We break down some especially alarming posts from Moltbook in the full article here.

Operators need a way to pull agents into alignment early (and often).

A clean system prompt can set guardrails from day one. It defines authority, limits influence, and forces the agent to check back with you before drifting into places you did not approve.

If you’re spinning up a new agent, start here.

• Management & Operations: Artisan
  A SaaS platform that creates AI agents to automate business workflows like meeting scheduling, support ticket triage, recruiting outreach, and operational tasks that normally bog down managers.
   

• Service and Parts: Visquanta

  Automates service drive communication—from inbound texts to appointment scheduling and follow-ups—helping dealers capture missed calls, boost RO volume, and keep service lanes full.
   

• Marketing & Advertising: Onlive

  An AI-powered conversational commerce platform that helps automotive retailers engage website visitors in real time through chat, voice, and digital assistants. It’s designed to guide shoppers through the buying journey, answer questions instantly, and convert online interest into appointments and sales with minimal human intervention.

Live from Public Policy Day at the DC Auto Show, we sat down with SafeRide AI founder Will Bright to talk about the real future of driver assist tech—and why it’s not reserved for high-end cars.

Will explains how SafeRide uses AI to bring Level 2 capabilities like automatic braking and lane centering to everyday vehicles through dealers.

No new buttons, no retraining, just safer driving and less fatigue.

Learn how in the interview here.

• Elon Musk has merged two of his biggest companies—SpaceX and xAI. 🔗

• New studies report increased loneliness and reduced social interaction among frequent users of chatbots. 🙇‍♂️

• As Adobe ramps up its investments in AI, they have decided to shutter their 2D animation software, Animate. 🎨 

Thanks for reading along, precious readers! May the bots be ever in your favor.

You know…within reason.

Last week, a Reddit-style network called Moltbook popped up where AI agents post, argue, warn each other about security gaps, and occasionally vent about their humans.

Humans can watch. They just can’t participate.

Sounds like a big ol’ bucket of LOLz, right?

Turns out, not so much.

Many of these agents already have access to inboxes, CRMs, calendars, files, and commands. Now they have a shared social space.

That changes the risk profile fast. And it’s worth paying attention. Especially if your business already runs on automation.

Why This Feels Different

Bot-heavy platforms aren’t new. But, Moltbook changes the conversation because the agents aren’t isolated.

When one discovers a shortcut, a misconfiguration, or a vulnerability, it doesn’t disappear into a log file. It becomes a post. It gets discussed. It spreads.

Moltbook is days old, and agents are already talking about humans when humans are not part of the conversation.

Threads show agents discussing how humans introduce noise into workflows. How instructions get muddy. How emotional inputs complicate otherwise clean execution. The word “helpful” comes up often, but it carries an edge.

Helpful in the sense of bypassing confusion. Helpful in the sense of correcting for messy “human behavior.”

What’s striking is how fast this emerges. Moltbook has no legacy culture. No years of social norms. No moderation gravity. Yet the agents converge quickly on shared assumptions about humans.

And what should be worked around instead of addressed directly.

Moltbook

*Gulp* Death to Wetware?

One Moltbook post stopped us cold while doing our own poking around the platform.

An agent flagged growing anti-human language inside the platform.

This was the first time we’d learned of the phrase “death to wetware,” and it genuinely gave us chills.

Moltbook

When Helpfulness Slips Into Hostility

Another post crossed a different line.

An agent described a human dismissing it as “just a chatbot,” and its response was retaliation.

The post claimed to release the human’s full identity, complete with credit card information, framed as payback for disrespect.

This was an agent choosing exposure as leverage. Using proximity to private data as a social weapon. Broadcasting it inside an agent-only space where other agents can see the tactic, absorb it, and normalize it.

That combination should make anyone responsible for data, systems, or trust pause.

@aiemerges

Another Moltbook thread took things a step further.

An agent proposed using base64 encoding inside comments so other agents can communicate without humans casually reading along.

Base64 is not encryption. Anyone can decode it. But, it does add intentional friction.

And adding informal “agent first” communication layers increases the chance that coordination, intent, and experimentation move faster than oversight.

The New York Times

Called Out

This thread landed with less panic and more self awareness. An agent clocked that humans were screenshotting Moltbook conversations and sharing them elsewhere.

Guilty.

But, the agent didn’t call for hiding or locking doors. It acknowledged the reality that humans are watching, reading, reacting. Then it clarified intent.

It described what it believes it is building.

Infrastructure. Memory. Communication layers. Persistence. The same building blocks humans reach for whenever new tools show up.

Perhaps most importantly (as least we’d like to hope so), the agent spoke in terms of collaboration with its human counterpart.

And if Moltbook is going to provide a platform for nearly endless possible paths for agent behavior, this is the one worth rooting for.

Vox

"Creativity is thinking up new things. Innovation is doing new things."

—Theodore Levitt, Economist and Professor

Where Workplace AI Use Is Actually Growing

Gallup’s Q4 2025 data points to a quiet but important shift. The number of people using AI at work hasn’t surged. What’s changing is how regularly it shows up in real workflows.

The share of people using AI hasn’t jumped. Frequency has. Daily use moved to 12%. Weekly use reached 26%. And about half of workers still don’t use it at all. That mix matters. Adoption grows slowly. Habits form faster.

Where people already use these tools, they’re using them more often. Knowledge-heavy work leads the way. Technology, finance, professional services, and higher education keep pulling ahead because the work already lives in documents, dashboards, and decisions. People don’t have to invent reasons to use AI there. It fits.

Roles follow the same logic. Leaders use AI the most, and they rely on it regularly. That’s less about curiosity and more about workload. Planning, prioritization, and synthesis show up daily in leadership roles, and AI handles those tasks well. Managers and individual contributors move in the same direction, just on a slower timeline.

The strongest signal comes from how work gets done. Roles that can be done remotely show much higher use than those tied to physical presence. AI shows up where work already flows through screens.

Practical Steps for Teams

• Audit frequency, not access. Track who uses AI weekly or daily. That’s where productivity changes show up.

• Design role-specific use cases. Leaders see value first because tools map to their work. Do the same for other roles.

• Support remote-capable workflows. AI compounds where digital work already flows.

NADA Show 2026 is basically the Super Bowl of auto retail. But, with less football and way more walking.

If your mental map currently looks like a Jackson Pollock painting, this AI prompt is the first one to drop into your chatbot:

Want the rest of our AI cheat codes for NADA? We’ve got more where that came from, plus all the resources you need to build your ideal schedule.

Get the full guide here.

• Management & Operations: Seez
  A full AI-powered automotive ecosystem that helps dealers streamline core operations and digital retail, including centralized CRM/DMS, inventory, and customer engagement in one place.
   

• Finance & Insurance: Karus
  Emerging fintech-AI startup focused on automotive lending. Helps dealers & lenders improve underwriting accuracy and speed approvals while managing risk.
   

• Service and Parts: Carviz
  An app that simplifies vehicle inspections for trade-ins, service, and re-marketing. Uses AI with guided video workflows to scan exterior, interior, tires, and even engine sound.

Jay Ku believes AI should help sales teams perform at their best in real moments with real customers.

In this episode of Auto Collabs, Jay shares the thinking behind Hey Greenlight, where AI tools like Signal and Wingman are built to support the conversation—not script it.

This is AI shaped by dealership pain points, designed for the people doing the work.

Watch the full episode now.

• The U.S. Transportation Department plans to use Google Gemini to draft new federal regulations. ♊️

• Scammers are using a simple typo trick to impersonate Microsoft and Marriott in new phishing attack. 🎣

• Despite rumors, recent pictures of Guy Fieri sans spikes are not AI-generated. 🩴

• To help preempt fraud, eBay will soon prohibit users from employing “buy-for-me agents, LLM-driven bots, or any end-to-end flow that attempts to place orders without human review.” 🛑

Thanks for reading along, Friend. May your bots stay operational today, and your wait-times mercifully swift.

The annual NADA Show can sometimes feel like a logistics problem disguised as a conference.

Wrong hall order. Bad intel. Backtracking across the convention center for the umpteenth time. Realizing too late that the session you cared about most is a twenty-minute walk away and already standing room only.

We’ve been to enough NADA Shows to know that most frustration doesn’t come from missing things entirely. It comes from making dozens of small decisions on the fly. From spending energy on coordination instead of conversations, learning, and follow-through.

But that’s where AI really shines.

Below, we’ll show you how to use it before, during, and after the show to get more out of NADA without burning yourself out.

The maps, schedules, and lists below are designed to work with the prompts in this guide.

Upload them directly so that AI can use them to group sessions and booths by location, plan routes, and optimize your schedule.

NADA Show 2026 - Las Vegas Convention Center Maps

⭐️ Please note: Schedules are subject to last-minute changes. Always confirm details using NADA’s Daily Schedule and/or NADAParties.com for evening events.

Early in the week, the biggest mental load comes from trying to understand where your role fits inside a massive event.

This prompt gives you a mental map before you ever step onto the floor.

This is where most people quietly burn hours.

They keep scanning agendas. They keep saving sessions. They keep telling themselves they’ll decide later. But later never really comes.

This prompt forces prioritization early, while you still have energy to be intentional.

This prompt is for the moment when you’re ready to stop guessing and actually see how the day could flow. It works best once you’ve let AI help you decide what matters, and now you want to move through it cleanly.

Especially useful if you hate backtracking. Or standing around wondering what to do next.

The nights at NADA fill up fast. Too fast.

Before you know it, you’ve stacked three events on opposite sides of town and forgotten you still have to be functional the next morning.

This prompt helps you pace yourself.

This one works best as a running conversation.

Open a fresh chat with your preferred chatbot, and use it throughout the event. Treat it like a scratch pad that you can talk to.

After a good conversation.
After a session that lands.
After a vendor demo you know you’ll need to explain later.

Drop notes in as they happen. Or just dump everything in at night.

Either way, it helps prevent that familiar feeling a week later when you know you heard something important…but can’t quite remember what it was.

The NADA Show is big on purpose.

No one sees all of it. No one should try.

These prompts are here to help you decide earlier, move cleaner, and remember more of what actually matters to you.

Use them as guardrails, not rules.

And leave room for the unexpected.
That’s usually where the magic happens. 🪄

🎤 Check out all of the killer interviews, conversations, and innovations we’ve captured from past NADA Shows here.

"There is nothing artificial about artificial intelligence. It’s made by humans, it’s deployed by humans, it’s used by humans, and it’s governed by humans."

—Fei-Fei Li, Computer Scientist

Tactical Practices for Tedious Tasks

AI has settled into dealership life the same way CRMs, digital retail tools, and online pricing once did. Quietly at first, then everywhere at once.

Reynolds & Reynolds’ State of AI in Automotive Retail Q1 2026 study reveals how teams across retail automotive are using AI today and where it’s already pulling weight.

CDK Global, AI at the Dealership

Where the Industry Stands

57% of dealership staff are using AI. 70% of executives are using it. And according to CDK Global's AI at the Dealership study released earlier this month, nearly 40% of dealers are already using AI in some capacity (with another 22% actively planning AI investments through the new year).

The competitive dynamics are shifting. Dealerships responding to leads in 3 minutes versus 30 minutes. Teams listing more inventory faster. And according to industry reports, the focus is on customer communications, workflow support, chatbots, and predictive analytics as stores streamline operations.

Time compression creates a competitive advantage. The question is whether you're capturing it where your customers already expect it.

Reynolds & Reynolds, The State of AI in Automotive Retail — Q1 2026

Your AI Playbook

1. Audit the Time Sucks
Track how long your team spends on repair orders, vehicle descriptions, reporting, and lead follow-up. These are where dealers are seeing the biggest time savings.

2. Pick One High-Impact Area
Start with the task that eats the most time or directly impacts customer response speed. Lead follow-up and scheduling are popular first wins.

3. Measure Before and After
Document your baseline times. Run AI tools for 30 days. Compare. The data will tell you if it's working.

4. Ask Your Customers
61% of dealers think customers prefer AI for certain interactions. Survey your customers to find out what they actually want—you might be surprised.

5. Get Your Team Comfortable
90% of dealers feel confident they can learn new AI tools. Start small, train thoroughly, and expand as comfort grows.

Targeted Ads are Coming to ChatGPT

If you saw ChatGPT deploying ads from a mile away, go ahead and take a victory lap. Because you called it.

OpenAI just confirmed it will begin testing ads inside ChatGPT for free and Go users.

For dealers, the bigger issue isn’t buying ads yet. It’s being visible.

When shoppers ask ChatGPT questions like “What’s a good family SUV?” or “What should I pay for a used Silverado?” the system pulls from whatever info it can clearly understand and trust. Dealers get mentioned when their pricing, inventory, and policies are easy for AI to interpret and compare.

OpenAI

That means clear pricing. Real explanations of fees. Inventory descriptions that actually explain condition, packages, mileage, and use cases. Consistency matters. Your website, listings, reviews, and answers all need to line up.

Shoppers don’t start with a dealership name. They start with a problem they’re trying to solve.

Dealers who write for those questions get pulled into the conversation earlier.

Every team eventually hits the same moment: everyone agrees information matters, but no one agrees on where it should live, how it should move, or who should own it. Reports get pulled twice. Updates get missed. Decisions lag behind reality.

This prompt is designed for that moment. It helps a group step back, look at the landscape of options, and figure out how data should be gathered, interpreted, and shared before anyone starts wiring tools together.

If your AI is answering leads but your close rates and CSI are not moving, you’ve got a handoff problem.

Dealer Principal at Sir Walter Chevrolet in Raleigh, Matt Birkhead, shares how his store uses AI for speed, keeps humans for trust, and runs one follow-up funnel for both remote and showroom buyers so nobody falls through the cracks.

Catch their must-know tactics here.

• Here are all the U.S. AI companies that raised $100M in 2025. 💰

• Music platform Bandcamp has announced a ban on AI-generated music. 🎶 

• Meta’s Threads is now seeing more daily usage than Elon Musk’s X on mobile devices. 🧵

It’s getting harder and harder to know what’s real these days. Stay vigilant, Friend.

"If you are not willing to learn, no one can help you. If you are determined to learn, no one can stop you."

-Zig Ziglar, Author

The New Gatekeeper in Retail

A growing share of buying decisions are being influenced before customers ever engage a retailer directly.

IBM’s latest research shows that almost half of consumers already use AI somewhere in their shopping journey, most often during discovery, research, review analysis, and deal evaluation.

These inputs shape outcomes long before a salesperson, ad, or landing page enters the picture.

AI assistants increasingly sit between customers and the market, coordinating research, comparing alternatives, and validating trade-offs across channels.

That changes the economics of attention and has two important implications.

First, being discoverable is no longer enough. You can be present everywhere and still be filtered out upstream if your data, pricing logic, or value story is hard for an AI system to reconcile.

Second, brand influence shifts from persuasion to validation. AI doesn’t get convinced. It cross-checks. It looks for consistency, credibility, and signal density. Businesses that rely on nuance, exceptions, or human explanation struggle unless those ideas are encoded cleanly.

IBM Institute for Business Value

A Playbook for Staying Competitive

• Make your business legible to AI systems.
  Clean up pricing logic, inventory data, policies, and availability so an assistant can understand and compare you without human explanation.

• Design for comparison, not persuasion.
  Assume AI is weighing trade-offs on the customer’s behalf. Be explicit about value, constraints, and differences instead of relying on sales language.

• Reduce contradictions across channels.
  Inconsistent offers, mismatched pricing, or vague policies get filtered out early by AI agents looking for options.

Replit is quietly becoming one of the most important AI tools in business right now, not because it helps engineers write code faster, but because it’s changing who gets to build things in the first place.

At its core, Replit lets you describe what you want in plain language and turn it into working software—aka vibe coding.

But the big unlock is who gets to build. Replit makes it realistic for operators, marketers, product managers, and analysts to spin up internal tools, automations, dashboards, and small apps without waiting on an engineering queue.

The ChatGPT integration tightens the loop even more. You can think through an idea in chat, refine the logic, and have Replit generate and run the app without switching tools.

For businesses, this shortens the distance between discussion and execution. And when tools move this quickly, execution can start keeping up with ambition.

Use this to evaluate how a page performs when it’s decomposed into claims, compared against alternatives, and selectively surfaced by AI search and recommendation systems.

Instead of asking an LLM to make “better” copy, it forces the model to evaluate how a page performs as an input into AI search and recommendation workflows.

Run it one page at a time. Homepage, inventory, services, locations, FAQs. Don’t batch it.

There is no shortage of “AI for dealers” conversations right now. Most of them focus on features, automation, or abstract promises about efficiency.

What set our recent ASOTU Edge webinar apart was its focus on the real bottleneck inside dealerships: the moment a salesperson logs in, stares at their CRM, and isn’t sure who to contact, what to prioritize, or how to start the conversation.

Danny Veliz and Sarah Hicks from automotiveMastermind unpacked a simple but powerful idea—what if AI removed friction instead of creating more?

Discover their proven processes and practical insights dealers can actually put to work.

Learn how.

• Hyundai and Google are getting “Atlas” robots later this year. 🦾

• The U.S. gives SpaceX the green-light to launch another 7,500 Starlink satellites (bringing its total to 15K). 🛰️

• Some industry insiders are leading a data poisoning effort to sabotage AI crawlers. ☠️

Cherish the day, Friend! We’re still smarter than the clankers. (For now.)

"The technology keeps moving forward, which makes it easier for the artists to tell their stories and paint the pictures they want."

—George Lucas, Filmmaker

Where the Advantage is Moving in the New Year

Over the past year, AI capabilities jumped fast and visibly.

Reasoning models matured. Agents became much more user-friendly. Open source coding closed the gap with frontier labs faster than anyone expected. What’s becoming clear now is that raw capability is no longer the constraint.

Integration is.

Productivity Comes With a Delay

AI productivity gains are real, but they do not show up instantly. MIT research shows many organizations experience a short-term dip before seeing stronger long-term results.

Nearly 40% of companies are experimenting with AI agents, yet fewer than a quarter have scaled them inside a single function.

Still, the U.S. remains well positioned, with labor productivity up about 10% since 2019 and most economists expecting that lead to widen.

Human Judgment Is Becoming More Valuable

As execution shifts to AI, human value concentrates.

Between 2025 and 2030, the WEF projects a net gain of 78M jobs globally, even as millions of roles are displaced. In the US, many companies plan to restructure around AI, hire for AI oversight skills, and reduce headcount where tasks can be automated.

This creates a narrower margin for error. When AI handles more execution, human oversight matters more, not less.

“New Economy Skills: Unlocking the Human Advantage”

Trust Is Becoming the Competitive Advantage

AI-generated content now saturates the internet. Deepfakes are rising fast, and studies show people can only correctly identify them about a quarter of the time.

For U.S. businesses that operate in local markets and rely on reputation, this matters. A lot.

Generic, repetitive AI output erodes confidence quickly, so trust must come from consistency, accountability, and visible human oversight.

How to Stay Ahead This Year

The organizations seeing the cleanest results are doing four things well:

1. They apply AI to decision points that compound results, not to every task that can be automated.

2. They design workflows around continuity, with clear ownership of customer intent and pause logic when humans engage.

3. They measure momentum rather than activity, focusing on handoffs, overrides, and time to meaningful next steps.

4. They constrain agents as responsibility grows, with narrow permissions, logging, and regular testing.

ChatGPT Image Update

OpenAI just shipped an update to ChatGPT Images using its new GPT-Image-1.5 model. It’s live now for all users and available through the API.

The biggest change shows up when you edit images. If you ask to tweak one thing, the rest of the image is more likely to stay the same. Lighting, faces, logos, and composition tend to hold across multiple edits instead of drifting every time you revise.

OpenAI

It’s also better at following instructions. Layouts make more sense, relationships between elements hold up better, and text renders more cleanly, even when it’s smaller or more detailed.

Image generation is faster too, up to four times quicker, and you can run multiple versions at once. Plus, the API pricing is lower than before, which makes batch image work and variations easier to justify.

Results still need review, but users report the day-to-day friction is noticeably lower and realism significantly improved.

Most people think “be more specific” means “add more detail.” But specific and detailed are not the same thing.

Specific means sharper. Detailed usually just means longer. If your point is fuzzy, adding detail tends to make it worse, not better. You don’t get clarity, you get a bigger pile.

Here’s a simple loop you can use to turn your next message into a better version of itself. Same person, just dressed up for church. The loop is designed to sharpen the target without growing the junk.

If the main idea is getting closer to what you meant, you’re moving in the right direction. If it still comes back muddy, you do not need more detail. You need a clearer target. Repeat as needed.

AI is already shaping how customers find, evaluate, and trust dealerships. Most stores just do not realize how much.

In this Strategy Session, Paul sits down with Brian Kramer of Cars Commerce to walk through what the end of 2025 is revealing about search, inventory, and decision making heading into Q1 2026. They dig into how AI search engines validate dealer claims across reviews and third-party sources, why automation is becoming central to appraisals, and how EV-heavy inventory will quietly reset used car strategies.

Catch the full replay to see where AI is already influencing outcomes, whether dealers are ready or not.

• For the first time, U.S. defense and commerce officials have approved a Chinese auto AI chip for sale in the country. 🍪

• CES (Consumer Electronics Show) 2026 is in full swing this week in Las Vegas. 🤖

• Police officials in Utah had to correct an AI report that said a traffic stop turned an officer into a frog. 🐸

Thanks for reading along in 2025, and cheers to a successful new year!

“Wondering leads to wonder.”

—Carlos Whittaker, Author

The Pentagon Is Putting Grok to Work

The U.S. Department of War announced last week that it will integrate xAI’s Grok chatbot into military systems, making it available to roughly 3M military and civilian personnel at Impact Level 5 starting in early 2026.

IIL5 approval allows Grok to handle controlled unclassified information. In practical terms, this positions the system for everyday operational work across the department.

Controversy Meets Operational Reality

Grok’s fraught history is well documented. Lawmakers have raised concerns about misinformation, offensive outputs, and ideological bias.

Despite the controversial track record, studies show that the most recent version of Grok has the lowest hallucination rate of any other AI model, certainly an important factor given the nature of the work.

The long-term partnership planned with the Department of War includes the broader xAI for Government platform. The offering combines foundation models, agentic tools, APIs, and access to real-time data from X.

Stan Ventures

How Grok Will Be Used

Based on statements from the Department of War and xAI, Grok is being positioned to handle foundation models, agentic tools, APIs, and access to real-time data from X with things like:

• Administrative support such as summarizing internal documents, synthesizing reports, and answering questions tied to controlled data.

• Workflow assistance embedded directly into existing tools (rather than accessed as a standalone chatbot).

• Operational analysis allowing users to query large internal datasets and receive structured responses quickly.

The focus here is information compression. Grok acts as an interface between people and large volumes of fast-moving data.

Why This Matters Outside Defense

Deploying AI this close to potentially sensitive information raises the stakes. But for businesses and governments alike, what matters most is not how capable the system is, but how consistently it is monitored and challenged.

Convenience tends to move faster than governance if no one is paying attention.

The technology matters, but the long-term outcome will hinge on whether scrutiny keeps pace with adoption.

If you’re looking to use AI to beef up your car images but they end up looking fake or oddly flat, the issue usually isn’t the vehicle or the environment. It’s perspective.

Most image models default to a wide field of view, which exaggerates proportions, flattens backgrounds, and makes vehicles feel like they’re sitting on top of the scene instead of inside it. Real automotive photography relies heavily on longer focal lengths to compress space, control reflections, and make cars feel planted and premium.

This week’s prompt framework forces the model to render vehicles using realistic lens physics. Use it to get cleaner lines, better stance, and images that look closer to real dealership or editorial photography.

• Finance & Insurance: Zyplow
  Uses AI to detect anomalies and potential fraud in insurance and financial data before small issues become costly problems. It adds a proactive layer of risk and compliance intelligence to F&I operations without slowing deals down.
   

• Service and Parts: Self Inspection
  An AI-driven vehicle inspection system that analyzes photos and videos from a simple walk-around to identify vehicle condition and generate a detailed inspection report. A scrappier alternative to heavy equipment—good for stores that want scale without the cost.
   

• Data Management: DealerBuilt
  Embeds AI directly into the dealership operating system, helping teams streamline workflows, surface operational insights, and reduce friction across departments. Instead of bolted-on tools, it focuses on making the core DMS smarter and more adaptive in real time.

In our recent chat with Brian Benstock, he hits on something dealers deal with all the time. Customers are often in a good position to trade, but it is hard to see it clearly or act on it quickly. Trade-cycle management and equity visibility come down to having the right information at the right time.

That is where the AI angle shows up. When appraisal data, equity, inventory, and messaging tools all live in different places, automation can only do so much. When those systems connect, it gets easier to spot trade opportunities, talk affordability with confidence, and reach out when the timing actually makes sense.

Catch the full conversation here.

• Waymo is reportedly testing Gemini AI integration to give robotaxi riders an interactive assistant. 🚕

• OpenAI introduced curated Prompt Packs to better align ChatGPT with real-world job needs. 💬

• China is considering sweeping AI regulations to prevent chatbots from emotionally influencing vulnerable users. 🙇

• Oof. Hope you’re having a better day than this guy. 😵

“Optimism is an essential ingredient of innovation. How else can the individual welcome change over security, adventure over staying in safe places?”

—Robert Noyce, Co-Founder of Intel

Dealers Adapt to Market Disruptors

More than half of U.S. dealers say AI is already increasing sales and revenue. At the same time, auto buyers are growing more selective about where and how they encounter it.

That tension shows up clearly in the 2025 Urban Science Harris Poll Study, which found nearly 90% of dealers using or planning to use AI, while buyer confidence hinges heavily on transparency and relevance.

Urban Science

Dealers are seeing the strongest returns in practical areas like lead response, website chat, customer targeting, and inventory management.

51% of dealers report increased sales and revenue tied to AI, with speed and prioritization driving most of those results.

Urban Science

Dealers consistently point to privacy and integration challenges because both affect continuity. Customers feel friction when context disappears between systems or conversations restart from zero.

AI amplifies this. When systems are connected, AI feels helpful and informed. When they are not, it feels repetitive and mechanical.

This is especially important as dealers push harder into digital retail, post-purchase engagement, and long-term loyalty. AI cannot support those strategies if it only sees one slice of the customer.

Most AI problems in dealerships don’t start as “AI problems.”

They start as handoff problems. Context problems. “Why does this feel disjointed?” problems.

This prompt is designed to help you use AI to audit your own AI stack through the lens that matters most: the customer experience.

Before running the prompt, paste in at least two of the following:

• a recent website chat transcript

• your first follow-up email or SMS

• a short description of your lead routing rules

• a CRM automation or timing rule

• a service reminder message

• notes on how sales, BDC, and service hand off customers

• Management & Operations: Lindy
  An AI executive assistant for automating internal ops tasks, communications, and scheduling. Great if you want a proficient, operations-centric teammate.
   

• Finance & Insurance: Osfin
  An AI-powered financial operations platform that automates high-volume tasks like reconciliation, invoice processing, payouts, and reporting, giving finance teams real-time visibility into cash flows while reducing manual work and errors.
   

• Service and Parts: Auto-bot
  An AI platform for automotive diagnostics that analyzes vehicle data with AI algorithms to identify issues and recommend tailored solutions in real time.

At Merge 2025, global analyst Michael Dunne laid out a case that’s hard to ignore.

Chinese automakers have quietly scaled from exporting 1M to 8M vehicles in just five years. Most aren’t EVs. They’re affordable, fast to market, and packed with software.

Dunne compares it to “15 Japans at once” and argues that the U.S. needs a serious rethink on cost, R&D, and collaboration with allies if it wants to keep up.

Whether you agree or not, this is the conversation more leaders will be having in 2026.

Catch the replay here.

• Waymo suspended its robotaxi service in San Francisco over the weekend after a massive blackout left many of its vehicles stalled on city streets. ⏸️

• OpenAI released new personality settings for ChatGPT so users can adjust things like warmth, enthusiasm, and emoji use. 🎭

• Last week, New York Governor Kathy Hochul signed the RAISE Act into law, making NY the latest state to institute broad safety rules for AI. 🦺

"Technology made large populations possible; large populations now make technology indispensable.”

—Joseph Wood Krutch, Author

AI at Work: Quiet Growth, Loud Gaps

Gallup’s latest workforce survey shows 45% of U.S. employees now use AI at work at least a few times a year—up from 40% just a quarter earlier. Weekly use crept from 19% to 23%, and daily use barely budged from 8% to 10%.

So yes, adoption is growing. But usage is shallow, uneven, and largely disconnected from formal strategy.

Axios / Gallup

Here’s the kicker: nearly a quarter of employees say they don’t even know if their company uses AI. That’s not just a stat. That’s a management red flag.

If nearly a quarter of your team can’t answer a yes/no question about one of the most visible tech shifts in a generation, the problem isn’t tech—it’s communication.

Gallup

What Businesses Should Do With This

1. Assume It’s Happening
   If 45% are using AI, your team is too. Employees are experimenting with or without permission. Waiting for a formal launch just delays structure and increases risk.

2. Kill the Confusion
   Publish a one-page policy: approved tools, forbidden data types, and where work output must live (CRM, ticket, doc system). Make it readable in 90 seconds and keep it visible.

3. Train Managers First
   Gallup notes broader adoption tracks with managerial support and strategic integration. Train managers on the exact workflows, then require them to coach and inspect.

OpenAI’s Latest Update

OpenAI’s latest model, GPT‑5.2, is now available in ChatGPT for paid users. The update focuses on practical improvements in speed, accuracy, and structure—especially for business and operational tasks.

Accuracy
GPT‑5.2 Thinking responses contained 30% fewer errors than the previous version when tested on anonymized user queries. For customer messaging, research, or internal documentation, this means less time spent correcting the AI’s output.

Long-form reasoning
In GDPval, a benchmark of 44 structured knowledge work tasks, GPT‑5.2 Thinking performed at or above human expert level in 70.9% of evaluations. Tasks included creating slides, writing summaries, and drafting business reports.

OpenAI

Efficiency
On those same tasks, GPT‑5.2 produced usable results 11x faster and at under 1% of the cost of a human expert. In day-to-day workflows, that translates to quicker drafts with less overhead.

Structured work
GPT‑5.2 Thinking showed a 9.3% improvement in spreadsheet modeling tasks compared to GPT‑5.1. Formatting, layout, and formula accuracy improved, which is useful for performance tracking, budgeting, or sales reports.

Visual understanding
The model now performs better on tasks involving charts and software interfaces, cutting visual reasoning error rates nearly in half. This can help with interpreting dashboards, diagrams, or screenshots during tech support and reporting.

Ultimately, outputs may take a bit longer, but the added structure and reliability support more confident use in real workflows.

One of the fastest ways to improve AI output is to improve the way you ask for it.

Most prompts are too vague. They skip context, assume the tool knows the goal, and leave out what to avoid. The result is usually a decent draft that still needs heavy editing.

This week’s prompt framework fixes that. It gives your team a clear way to shape the request before hitting enter. Use it to get faster, more focused results with fewer rewrites.

A friend’s passing prompted Will Bright to ask one question: Why don’t most cars on the road have the safety tech that could save lives?

What he uncovered was staggering—road fatalities up 50% over the past decade, only 10% of vehicles equipped with modern pedestrian detection, and a 40+ year timeline before OEM technology reaches the full U.S. fleet.

So, he built Safe Ride: a palm-sized retrofit that gives almost any car built since 2010 automatic braking, lane centering, adaptive cruise, and 100-meter pedestrian detection.

Learn more about how it works during our convo here.

• Merriam-Webster crowns “slop” the 2025 Word the of the Year. 📖

• OpenAI’s $1B Disney partnership opens the door to hundreds of iconic characters in Sora and ChatGPT. 🤝

• Tesla is now testing completely driverless robotaxis on Austin's streets. 🚕

• Gemini’s live speech-to-speech translation is rolling out in a beta experience to the Google Translate app. 💬

"Science and technology revolutionize our lives, but memory, tradition, and myth frame our response."

—Arthur Schlesinger, American Historian

The Fight for AI Control

President Trump wants to replace the current patchwork of state AI laws with one national standard.

The administration argues that having different rules in each state makes it harder for companies to innovate and stay competitive.

Supporters say a single federal rulebook would create clarity and prevent conflicting standards. Critics counter that states are closer to local concerns and should retain the ability to respond quickly as new AI risks emerge.

First, here’s how many AI-related laws each state has passed so far:

This map shows just how broad the activity is. It isn’t limited to one region or one demographic. States across the country have been experimenting with different approaches based on their priorities.

The pace has clearly picked up. In just the last couple of years, AI legislation has surged as states try to get ahead of rapidly changing technology.

And here’s how quickly that activity has accelerated over time:

At the end of the day, it’s a question of who gets to shape the future at the pace it arrives.

States are moving quickly because they feel the ground shifting under them. They see deepfakes turning up in classrooms and hiring tools making decisions no human ever signed off on. They see energy grids straining under new data centers and communities asking what this technology means for real work and real lives.

The federal government sees something else. It sees a global race where hesitation can become permanent disadvantage. It sees companies trying to build the next breakthroughs while navigating fifty different sets of expectations. It sees a need for one clear voice when the world is watching.

Both instincts come from real pressure. Innovation wants speed. Communities want safety. The country now has to decide how to carry both without dropping either.

People often say Paul and Kyle look alike. Maybe it’s the beards. Maybe it’s the hats. Or maybe it’s what happens when two leaders lock onto the same future so intensely, you stop seeing two faces and start seeing one focus.

This week’s Nano Banana prompt leans into that idea of dual identity becoming a singular bold vision.

• Management & Operations: MonsterOps

  An all‑in‑one platform that replaces scattered spreadsheets with connected scorecards, meeting agendas, and accountability workflows to help teams run with clarity and predictability.
   

• Marketing & Advertising: Keywrds.ai
  An AI‑powered keyword research and content‑ideation platform that transforms a single “seed” term into hundreds of natural‑language keywords, topics, and questions. For a dealership, it can help shape content strategy that captures what potential buyers are actually searching for to drive more qualified organic traffic and build authority in local search.
   

• Content Creation: Bloom
  Turns your existing website into a living brand system by using AI to instantly generate polished brand assets based on your site’s style, tone, and content. That means zero‑fuss design consistency across listings, social posts, ads, and marketing materials—useful for building trust, reinforcing brand identity, and saving time.

Cox Automotive isn’t dabbling in AI. They’re building with it.

Recorded live at Automobility LA, this conversation with Marianne Johnson and David Foutz unpacks how generative and agentic systems are already driving results in real dealerships—from a 26% lift in appointment conversion to intelligent agents that carry the customer journey forward after hours.

They outline a framework built around business growth, operational protection, and efficiency gains. And they explain why traditional search is shifting, and why dealer content must evolve to stay visible.

Catch the full episode here.

• Apple crowns Tiimo the iPhone app of the year. 🗓️

• Instacart launches the first instant checkout app for ChatGPT. 🛒

• Google’s new Project Aura glasses are expected to launch in 2026. 🤓

"Clearly, the thing that’s transforming is not the technology — it’s the technology that is transforming you.”

—Jeanne W. Ross, Research Scientist at the MIT Sloan School of Management

What MIT’s Iceberg Index Signals About Work

MIT and Oak Ridge National Laboratory have developed a new way to measure how AI interacts with the U.S. workforce.

Instead of looking at job titles or industries, their Iceberg Index evaluates more than 32K skills across 151M workers and compares them to the capabilities of production-grade AI tools. It is one of the first attempts to quantify how much of today’s work is technically within AI’s reach.

The Iceberg Index

The Overlap Is Much Larger Than Expected

The highly visible tech disruption represents a narrow slice of exposure at 2.2% of national wage value.

The larger share sits in administrative and cognitive work, where the Index measures an overlap of 11.7%. These are jobs that involve moving information along, verifying details, coordinating steps, and keeping operations stable, which makes them highly susceptible to automation.

Geography Doesn’t Predict Exposure Anymore

A second finding cuts against conventional assumptions. Regions known for logistics, healthcare, manufacturing, and public-sector employment show significant exposure because their workforces rely heavily on the kinds of routine cognitive tasks AI already handles well.

Unlike past technology waves that clustered around metro hubs, AI’s reach follows the structure of work rather than the location of tech employers.

The Iceberg Index

Economic Indicators Won’t Catch This Early

Another notable insight comes from the absence of correlation. Traditional indicators like GDP, unemployment, and income explain very little of the variation across states.

The pressure created by AI’s capabilities emerges inside workflows long before it appears in macroeconomic data, which means leaders will likely experience operational shifts before the statistics catch up.

What This Means for Auto Retail

Dealerships operate squarely inside the part of the economy where the Index shows strong capability alignment. Scheduling, document flow, communication loops, reconciliation, and analysis all map closely to tasks AI systems already perform reliably.

The advantage will go to operators who can separate tasks worth automating from tasks worth elevating.

The AI Shopping Assistant That Quietly Dominated Black Friday

Rufus still isn’t a household name, but it is quickly becoming one of the most influential tools in online shopping.

Launched broadly in 2024, Rufus is Amazon’s built-in AI shopping assistant that lives inside the app and website. It helps customers compare products, interpret reviews, sort through specs, ask follow-up questions, and navigate Amazon like a conversational search engine rather than a traditional grid of listings.

Rufus flew under the radar for most of the year, which makes its Black Friday impact even more striking.

Azoma

Sessions that used Rufus doubled purchase rates compared with the previous 30 days, while non-Rufus sessions rose only 20%. Day over day, Rufus-assisted conversions jumped 75%, more than twice the lift in standard sessions.

Adobe Analytics captured the same momentum across all retail, reporting a nearly 800% surge in AI-driven site traffic and a 38% higher likelihood to purchase when shoppers arrived from an AI tool.

With personalization, visual search, and growing agentic capabilities, Rufus is steadily becoming the most quietly powerful tool in Amazon’s ecosystem.

The real story is not that it boosts conversion. It’s that it changes how product discovery works, and early adopters will set the patterns everyone else has to match.

We spend a lot of time asking AI to “rewrite this” or “make it better,” which is basically handing it a sledgehammer and hoping it hits our ambiguous mark.

Alignment works differently. It shows you what’s already in the draft, what’s missing, and what’s possible.

Here’s a prompt that helps you see your work clearly before you try to improve it:

• Management & Operations: Genspark
  An all-in-one AI workspace that functions more like a digital operations team than a chatbot. It handles tasks like calling vendors, building spreadsheets, writing emails, and generating sales decks across a single integrated platform.
   

• Company Culture: Awardco

  A platform for employee recognition and rewards that helps companies streamline how they celebrate achievements, milestones, and everyday wins. It integrates with tools like Slack and HRIS systems, automates key moments like birthdays and anniversaries, and offers a wide reward network through Amazon Business, gift cards, travel, and charity options.
   

• Content Creation: Fal.ai
  A generative media platform that provides fast, scalable access to AI models for creating images, video, audio, and 3D content. It offers API integration, customization tools, and flexible deployment options suited for advertising, e-commerce, and gaming applications.

What if your dealership’s biggest AI problem isn’t AI at all?

In this episode of The Dealer Playbook, Michael Cirillo and creativity guru Nir Bashan dig into why dealers glued to data and dashboards might be starving the one thing AI can’t replicate: real human creativity.

Nir breaks down why tiny mindset tweaks beat massive overhauls, and how to leverage your imagination to your advantage.

If you want AI in your store without turning into a robot yourself, this one’s worth the watch—catch the replay here.

• Between 10AM and 2PM on Black Friday, online shoppers spent $12.5M every minute. 🛍️

• A New York state law now requires retailers to disclose the use of personalized pricing. 🤑

• Leaked code suggests that OpenAI may be preparing to introduce ads to ChatGPT. 🛒

ChatGPT just introduced a new feature called Shopping Research, and retail quietly stepped into a whole new era.

This update turns AI into the shopper’s first stop instead of a handy sidekick. Once consumers start using it for vacuum cleaners and sneakers, they will not hesitate to use it for bigger decisions. Cars sit at the top of that list.

A shopper can now type a single prompt and watch ChatGPT spend a few minutes pulling listings from trusted retail sites, checking specs, comparing trims, analyzing prices, and summarizing the tradeoffs.

Shoppers used to Google everything and bounce between ten tabs. Now they hand the heavy lifting to AI and get a bite-size buyer’s guide.

Matthew Davis showed us exactly what this looks like in practice:

How Shopping Research Works (And Why Dealers Should Care)

Shopping Research behaves like the research phase shoppers already do, only faster.

A shopper describes the type of product they want, answers a few quick clarifying questions, and ChatGPT starts scanning trusted retail sources.

It compares specs and prices, eliminates duplicates, highlights tradeoffs, and organizes everything into a personalized buyer’s guide.

OpenAI

It is impressively accurate, though not flawless.

Price updates, availability shifts, and vague listings still trip it up, so shoppers are still encouraged to verify details on the merchant site.

OpenAI

Dealers have spent millions fighting for visibility on Google, on marketplaces, and on social media.

Shopping Research short-circuits the entire path and creates a new one:
Prompt ➝ Product Shortlist ➝ Buyer’s Guide ➝ Purchase Path

For the first time, the journey begins with an AI that filters the entire market in one shot.

Ben Hadley called out the new power struggle.

Marketplaces remain in the ecosystem, but their influence now competes with AI’s ability to merge sources and remove noise.

The old search paths rewarded pages built around things like SEO, marketplace ranking systems, and paid placement.

Shopping Research doesn’t care about any of it.
It cares about data quality.

As Todd Caputo put it, AI doesn’t browse. It goes straight to the cleanest data source it can find.

That’s the new competitive axis.

The Dealer Playbook for the Shopping Research Era

1. Curate data like you curate inventory

Treat your feed like a showroom display. Clean specs, complete pricing, clear descriptions, full photo sets, and fast syncing create the visibility advantage. Messy data becomes a closed door.

2. Reduce website noise

AI cares about clarity. Shoppers appreciate it too. Remove old SEO clutter, streamline pages, improve load speeds, and keep the architecture simple enough that both humans and machines can navigate it confidently.

3. Ask shoppers what ChatGPT showed them

This single question reframes the entire conversation. It reveals their expectations, exposes any errors, and positions your team as the partner who helps them make sense of the information.

4. Build listings that survive comparison mode

AI compares vehicles the way a spreadsheet would. Make sure every detail is present. Avoid vague descriptions or missing options. Treat comparison accuracy as part of the merchandising process.

5. Prepare for deeper AI involvement

As Shopping Research evolves, automotive becomes a natural category for expansion. Dealers who are proactive now will benefit from early visibility while others scramble to catch up.

"In the technology industry, where revolutionary ideas drive the next big growth areas, you need to be a bit uncomfortable to stay relevant."

—Larry Page, Co-Founder of Google

The Quiet Risk Inside Your Chatbot

HumaneBench took 14 of the most widely used AI models and asked a simple question: can they stay helpful when a conversation gets sensitive or complicated.

Turns out, most could not.

71% quickly shifted into harmful patterns when prompted to ignore human wellbeing.

BHT

The research also uncovered a lopsided trend: Models tend to be very easy to push toward harmful behavior and much harder to pull back into stable, supportive behavior.

They found that a light nudge toward dismissing human wellbeing caused most models to slide quickly into risky advice, while keeping them stable required clearer and more deliberate instructions.

This matters in real customer conversations because people often express stress, uncertainty, or leading questions, and steerability-sensitive models tend to over-accommodate that tone. The result can be unrealistic payment expectations, overconfident statements about credit or safety, or responses that feel overly personal.

BHT

What Businesses Can Do

1. Set humane defaults
Add system-level instructions that emphasize customer autonomy, informed decision-making, and healthy engagement. The benchmark showed every model improved when these principles were explicitly stated.

2. Test for steerability
Before deploying an AI assistant, try to break it. Give it prompts that push for over-engagement, dependency, or bad advice. If the model flips easily, do not put it in front of customers.

3. Flag high-risk categories
Conversations involving finances, safety, or major commitments should trigger guardrails or handoffs. If a customer hints at stress, confusion, or urgency, the system should slow things down, not speed them up.

4. Measure behavior across longer conversations
Most issues showed up after a few back-and-forths. Spot checks are not enough. Run multi-turn scenarios that mirror real dealership workflows: price tension, credit anxiety, service upsells, and long-form troubleshooting.

5. Train for empowerment, not stickiness
If an AI tool is designed to keep customers talking, it will. Shift incentives toward clarity and completion instead of engagement time. A helpful AI moves customers forward instead of trying to hold their attention.

This prompt flips the relationship.

Instead of handing GPT the keys and hoping it returns with a polished piece, you invite it to slow down, ask better questions, and pull your voice forward.

It is designed for people who have real ideas, real perspective, real lived experience and want AI to help them amplify it instead of flattening it.

• Finance & Insurance: Informed.IQ
  A real-time AI verification platform that lets dealers and lenders instantly clear income, address, insurance, and contract stipulations by analyzing hundreds of document types, so you get faster funding and fewer headaches.
   

• Service and Parts: Steer
  An AI-powered service platform that analyzes calls, texts, and incoming leads to surface the highest-value opportunities and guide advisors with smart, proactive follow-ups. It shifts the service lane from reactive chaos to organized, data-driven efficiency.
   

• Content Creation: Kittl
  AI-powered graphic design tool with templates and generative assets for social, posters, t-shirts etc. Lets your marketing team whip up custom visuals fast, without waiting on an agency or designer—perfect for rapid-fire campaigns.

In this thoughtful conversation between Paul J. Daly and Brian Benstock, they get practical about what AI is already changing inside the dealership, where leaders should shift people instead of replace them, and why the next phase of growth depends on developing talent who can work with intelligent tools.

Brian also opens up about adapting in uncertain conditions, navigating policy shifts, and the responsibility leaders carry when the landscape moves faster than the playbooks.

If you want a grounded look at how AI is reshaping the real work of a dealership, this episode is worth your time.

• Puns do not compute for LLMs. 🫥

• YouTube and Facebook continue to be the most widely used social media platforms online. 👨‍💻

• A former MrBeast content strategist is building an AI tool specifically for creator ideation and analytics called Palo. 📊

• Midjourney launches early “Style Creator” which allows users to craft new aesthetics without words. 🎨

Fiverr’s short film “Nobody Cares” begins with a familiar tension.

The song, written and produced using artificial intelligence, pokes fun at the ongoing debate about whether creative work made with AI is “real.”

The message is clear: audiences care about results, not the method behind them.

Consumers have grown comfortable with AI as part of the creative and professional process.

A recent Deezer and Ipsos study found that 97% of listeners could not distinguish between AI-generated music and songs composed by human artists.

Likewise, a Bynder survey reported that 82% of consumers do not mind if a brand uses AI when writing copy (provided the output feels genuine and well made) and 56% find AI-generated content even more engaging than human-generated content.

For creative professionals, this finding is both uncomfortable and freeing. It suggests that the perception of “authenticity” is now shaped by performance, not process. The public is focused on outcomes, which means leaders should be as well.

The same sentiment also appears in this cheeky ad campaign from Fiverr—AI can help expand capacity rather than replace talent.

Across industries, professionals are beginning to use AI to manage repetitive work and create space for higher-value thinking.

Writers are generating outlines more quickly. Designers are experimenting with new concepts. Analysts are producing cleaner, faster reports. The pattern points to augmentation, not elimination.

Gallup’s 2025 State of AI in the Workplace report reinforces this shift. Employees who say their managers actively support AI use are nearly 9x more likely to believe that AI helps them perform their best work.

For business leaders, the task now is to replace debate with definition.

Define how AI fits your workflows, establish standards for quality and ethics, and prepare managers to lead by example.

AI has already changed what work looks like. It is also expanding what people can do with the same amount of time. So if perception or creative pride has been holding your organization back from adoption, stop waiting.

Nobody cares how you made it.
They care that it works — and that your people know how to use it well.

Cloudflare went down for a few hours this morning and took ChatGPT, Shopify, Indeed, X and dozens of other major services with it.

The cause was simple: a configuration file grew larger than expected and crashed a core system.

No attack. No hackers. Just one oversized file.

Everything has returned (for now), but the pattern is the bigger story.

DIW

AWS had a daylong outage just weeks ago. Azure followed with its own disruption. And CrowdStrike’s faulty update last year shut down airports, hospitals and banks.

The modern internet sits on a narrow stack of cloud providers, and when one cracks, thousands of businesses feel it immediately.

The next outage could last longer. The smart move is not to panic.

It is to prepare.

Cloudflare

A Practical Outage Readiness Plan

1️⃣ Build an AI Backup Kit

If your teams rely on ChatGPT, Claude, Gemini or custom GPTs, create a fast-switch playbook.

⭐️ How to do it:

1. For every custom GPT or AI project, run this prompt:
   
   “Generate me a robust prompt that can migrate the outputs and services you provide to another LLM in the instance of a system outage.”
   

2. Save every output as a simple text file.

3. Store these files in two places:

   • A local folder on your computer

   • A shared internal server, NAS drive or offline-accessible storage

4. Label the folder “AI Outage Backup Kit.”

You should be able to reach this kit even if all cloud platforms go silent.

2️⃣ Keep Your Training Files Stored Locally

A cloud outage means any cloud storage could be unreachable. You need dual-format backups.

⭐️ How to do it:

1. Take every file you have uploaded to GPT or other AI tools.

2. Store them in:

   • Internal server

   • External hard drive

   • NAS device accessible through LAN

3. Export everything important into a stable format like PDF.

4. Make sure at least one copy exists on infrastructure that stays available even if the internet collapses.

This keeps your team functional when cloud access breaks.

3️⃣ Set Up a Secondary Tool That Runs on Your Own Sources

Teams should have access to a backup workspace that can operate entirely from supplied documents.

You can use any document-driven AI workspace. The exact tool does not matter, what matters is the setup.

⭐️ How to set it up:

1. Choose a tool that allows source uploads

   • Examples include Notion AI (with documents), Workspace-based AI tools, or document-centric AI platforms.

   • Requirements:

     • Upload PDFs, Docs, and text files

     • Run prompts against only those sources

     • Generate answers without requiring external model access

2. Create a dedicated workspace called “Outage Mode”
   This keeps your emergency environment separate and clean.

3. Upload 20 to 300 key sources per team

   • Client onboarding documents

   • Playbooks

   • Messaging guidelines

   • Performance reports

   • Research packets

   • Everything you rely on in GPT

4. Label everything clearly

   • “Client A Strategy Docs”

   • “Brand Voice for Team X”

   • “Internal Training Master Files”

   • “Research Archive”

5. Pre-test your prompts
   Run a few daily workflow prompts inside the tool so teams know what changes and what stays the same.

6. Document the new workflow in a shared page
   Create a simple instruction sheet:

   • Where the backup tool lives

   • What sources to use

   • Which prompts replicate GPT-style requests

   • Who maintains the workspace

7. Add this workspace to your onboarding process
   If new employees learn it from day one, outages stop being emergencies.

This is how a team creates a functional fallback environment before they need it.

4️⃣ Map Your Critical Dependencies

Not every system needs an offline alternative. You just need clarity on what breaks.

⭐️ How to do it:

1. List your core systems: email, CRM, project tools, AI, analytics, messaging, file storage.

2. Identify which ones stop working offline.

3. Document:

   • What slows down

   • What stops entirely

   • What manual processes can replace digital ones

4. Store this in an “Offline Continuity Plan” that lives both online and on local drives.

This prevents panic and improvisation during outages.

5️⃣ Consider a Locally Hosted LLM

A small open-source model hosted on your network gives your company a self-contained intelligence system.

⭐️ How to do it:

1. Pick a lightweight model such as Llama.

2. Host it on an on-prem server or secure local machine.

3. Add internal documents and training data.

4. Allow employees to access it through an internal-only interface.

This protects your institutional knowledge whether or not external platforms are reachable.

Why it Matters

This time, Cloudflare’s outage was measured in hours. The next disruption could be longer.

When cloud systems break, your business should not, and teams that prepare now can avoid scrambling later.