"An ounce of prevention is worth a pound of cure."
—Benjamin Franklin, Founding Father & Inventor
Prompt of the Week
CDK's latest dealership study found that most stores use AI and almost none have written rules for it. In practice, that gap looks like a salesperson pasting a credit app into a free chatbot to tidy it up.
You can close most of it in an afternoon. List every AI tool your team actually uses, including the AI features inside your CRM, phone system, and chat vendor. Then run this prompt to turn that list into a one-page policy people will actually read.
Act as an operations and compliance advisor for a car dealership.
We use AI across the store but have no written policy for how employees use it. Help me draft a one-page employee AI use policy.
AI TOOLS IN USE
[LIST EVERY AI TOOL STAFF USE, INCLUDING FREE CHATBOTS AND AI FEATURES INSIDE OUR CRM, DMS, PHONE SYSTEM, CHAT, AND MARKETING TOOLS]
WHO USES THEM AND FOR WHAT
[DEPARTMENT + TYPICAL TASKS, E.G. SALES: LEAD REPLIES / SERVICE: RO NOTES / F&I: MENU SCRIPTS / MARKETING: AD COPY]
SENSITIVE DATA WE HANDLE
[CREDIT APPS, SSNS, DRIVER'S LICENSE IMAGES, DEAL JACKETS, PAY PLANS, PAYOFF INFO, CUSTOMER CONTACT LISTS, ETC.]
Build the policy with these sections:
* APPROVED TOOLS: Which tools are approved, approved with limits, or off-limits, with one line of reasoning for each.
* RED-LINE DATA: The specific information that never goes into any AI tool outside our approved systems.
* GREEN-LIGHT USES: Tasks each department can hand to AI without asking.
* CHECK FIRST: Tasks that need a manager's sign-off before AI touches them.
* REVIEW RULE: What a person must check before any AI output reaches a customer, especially prices, payments, and rates.
* OWNER: Who owns this policy, who approves new tools, and how often it gets reviewed.
* WHEN IN DOUBT: The one thing an employee should do when they aren't sure.
Keep it to one page, in plain language a new hire could follow on day one.
Flag anything that touches the FTC Safeguards Rule or state privacy law for our attorney to review. Do not decide those items yourself.
Then give me:
1. A 10-minute script for introducing the policy at our next team meeting
2. Five quiz questions to confirm people understood it
3. The ONE rule you would announce first if we could only announce one
None of this is legal advice. Loop in actual counsel before the policy goes live.
This prompt covers your people. For the AI agents your vendors run inside your systems, use the Agent Containment Playbook from our Sept. 29 issue.
The AI Breakdown
Dealers Are Using AI Faster Than They're Writing the Rules

Almost every store is using AI. Almost none have put rules around it.
CDK's State of Dealership Cybersecurity 2026 found that 87% of dealership leaders use AI in some capacity. Only 6% have a fully defined AI governance framework. Another 24% have no oversight at all, and 15% aren't sure whether a policy even exists.
The gap is widest at the biggest groups. Among leaders of 11+ rooftop groups, 44% say AI is widely used across their stores, but only 7% have full governance in place.
The study surveyed 218 dealership executives and department heads. Sixty-one percent said AI has made them more worried about cybersecurity.
The Biggest Exposure Starts Inside the Building

At most stores, the likeliest AI slip comes from a well-meaning employee with a deadline.
A salesperson drops a credit app into a free chatbot to summarize it. A service advisor pastes a customer's history into a tool nobody vetted. An F&I manager asks AI to clean up a menu with the deal jacket still attached.
None of those feel like security decisions in the moment. Each one sends customer data somewhere the store doesn't control.
Dealers already sense this. Asked which AI risks worry them most, 47% named data leaks through AI tools and 40% named unauthorized AI use.
The outside threat is changing too. Email phishing has overtaken ransomware as dealers’ top cybersecurity concern, and 76% named AI-powered phishing as their biggest AI risk. Several dealers told CDK their staff had received convincing emails from someone posing as the store owner.
Even the people building this technology are cautious about it. Microsoft CEO Satya Nadella wants AI agents running on every PC, but he recently told NBC News he sees them as a security risk that shouldn’t be trusted by default. If the company selling Copilot is that wary, a dealership can manage a one-page policy.
Dealer Guide
Take inventory. List every AI tool your team uses, including free chatbots on personal phones and AI features built into software you already pay for.
Name the red-line data. Credit apps, Social Security numbers, license images, and deal jackets stay inside approved systems. Put that list in writing.
Pick an owner. One person approves new tools and keeps the policy current. Without an owner, the 15% who aren't sure will stay that way.
Train it in 10 minutes. A short walkthrough at a team meeting does more than a 20-page PDF nobody opens.
Review it quarterly. Vendors keep adding AI features to tools you already own. Recheck the list every quarter and whenever a vendor announces an AI update.
Top Tools
Google's Gemini 4 Argon Opens Up
On Oct. 13, Google made Gemini 4 Argon, its new flagship model, widely available. It first went to a small group of cybersecurity partners on Sept. 30.
Google AI Pro and Ultra subscribers can use it in the Gemini app and in AI Mode in Search. Businesses can reach it through Gemini for business and Google's API.
For stores that run on Google Workspace, the demo to watch shows Argon in the Gemini app taking a messy folder of project notes, turning it into a client proposal deck, emailing it, and setting up a scheduled task to handle future requests the same way. Swap the folder for a week of aged-inventory reports and the dealership version is easy to picture.
One caveat: the benchmark wins and customer results in the launch post come from Google and its early-access partners. Run Argon on your own work before you trust it with anything a customer will see.
Why it matters: if your store already lives in Gmail, Calendar, and Sheets, this is an assistant that comes with the keys already in hand. Google says it trained Argon to pause for human confirmation before irreversible steps. That’s Google’s safeguard. Your store still needs its own, so put Argon on the list from this week’s prompt.
Hear from the Experts
Dealers are starting to build their own AI dashboards, reports, and automations, and Brian Pasch argues vendors should welcome it.
In a recent LinkedIn piece, Pasch says the real question for vendors is whether they’re ready for dealers who expect open APIs, MCP connectors, and easy access to their own data. His line worth stealing: “Education has to include both capability and responsibility.”
His workshop partner, Ai Dealer Labs founder Gray Scott, pushed it further. In the comments, he said it’ll be interesting to see who has the first data breach from self-built dealer tech. In his own repost, he argued that the vendors who win will be the ones who teach this new wave of builders the security basics they might skip: “how to test the software, house it, how to ensure privacy.”
That’s the same gap CDK found, now showing up in homemade tools.
Read Brian’s full take, then Gray’s, before your next vendor review.
Bits and Bytes
Anthropic is consulting law firms on fair rules for AI agents that haggle over prices on their owners' behalf. Remember Meta's Muse? The bots are studying for the desk. 🤝
Google's Universal Cart is now live in the U.S. It tracks saved products across retailers, flags price drops, and uses a checkout standard built for AI agents. 🛒
Rivian CEO RJ Scaringe says the company’s vehicles will eventually run AI agents from companies like Anthropic, though agents won’t get control of the brakes or steering. 🚙
Volvo is expanding its Nvidia partnership to phase in eyes-off driving first, then mind-off driving on future models. 😴
GTCR and Elliott agreed to take CCC private in a deal that values the car insurance software company at $6.1B including debt. 🔧
Character.AI filed for bankruptcy five years after launch, and Disney agreed to hire more than 60 of its employees. 🏰
Parting Pixels
Thanks for reading, Friend! Remember, if your AI can’t dance, you more than likely just need better prompts.

