Your Vendor Wrote the AI. Your Dealership Owns the Risk.
By Nathan Platter
Every boat dealer I talk to is getting pitched an AI tool right now. It answers web chats at 10 p.m. It follows up on show leads. It texts the guy who asked about the pontoon back in March. The demos are slick, and some tools are genuinely good.
Here’s what rarely comes up in the demo: when that tool tells a customer something, your dealership said it. Not the vendor. You.
I build AI systems that keep a person in the loop, and the most common misunderstanding I run into is that liability travels with the software. It doesn’t. It stays with the name on the message. I’m not here to scare you off AI. I’m here so you buy it with your eyes open.
The Chatbot is Not a Separate Person
In a 2024 case that every dealer should read, Moffatt v. Air Canada, the airline’s website chatbot told a grieving customer he could claim a bereavement discount after his trip. Wrong. The airline then argued the chatbot was responsible for its own actions.
The tribunal didn’t buy it. Per the American Bar Association’s summary, the airline “still bore responsibility for all the information on its website, whether it came from a static page or a chatbot.”
Closer to the showroom, a California Chevrolet dealer learned this in public. A user told the store’s chatbot to agree with everything, then asked for a new Tahoe for $1. The bot agreed and called it “a legally binding offer.” The dealer pulled the bot and wore the headlines.
Swap the Tahoe for a 24-foot center console and the story works exactly the same.
Three Places it Lands on a Marine Dealer
What you advertise. A text from your AI quoting a price, a show special, or “still available” is advertising. When the FTC launched Operation AI Comply, it was blunt: “there is no AI exemption from the laws on the books.” If the bot invents a discount, that’s your discount.
What you say about financing. Boat loans run long, so monthly payment talk is everywhere in this business. Under Regulation Z, stating a payment amount, a down payment, or the number of payments in an ad triggers more disclosures, including the APR. A bot that cheerfully texts “about $389 a month” has just written a credit ad, and probably an incomplete one.
How and when you text. The FCC requires consent before autodialed texts to a cell phone, and written consent for marketing texts. Customers can opt out “in any reasonable manner,” not just by typing STOP. And in 2024 the FCC ruled that AI-generated voices count as “artificial” under the TCPA, so an AI voice agent calling back leads plays by robocall rules.
There’s a fourth one underneath all three: customer data. If your store arranges financing, the FTC’s Safeguards Rule may well apply to you. It requires you to monitor service providers that touch customer information, with contracts that spell out your security expectations. An AI vendor reading your CRM is one of those providers.
The Off-the-Grid Problem
Plenty of reps are pasting chatbot replies into their personal phones or Facebook Messenger to keep up with leads. Same exposure, no record of what was promised. If a customer later says your guy guaranteed a trade number, you have nothing to pull up.
Five Gotcha Questions for Any AI Vendor
Many vendors haven’t thought this through either, so ask. You don’t need to speak AI to do it. A competent vendor knows exactly why you’re asking.
- “Who presses Send?” Professional answer: “A person on your team approves every message. Here’s the approval screen.” Red flag: “The AI handles it. It’s smart.”
- “How do you track consent and opt-outs?” Professional answer: they show where consent is stored, how quiet hours are enforced, and how any version of “stop” ends texting on every channel. Red flag: “Customers can just reply STOP.”
- “What customer data does your AI touch, and where does it go?” Professional answer: they name the data, where it’s stored, and agree to Safeguards Rule security terms in the contract. Red flag: “It’s all in the cloud.”
- “Show me a log, then show me the off switch.” Professional answer: one thread with every draft, edit, sender, and timestamp, exportable to you, plus a one-step shutoff. Red flag: “We can pull that if something happens.”
- “Show me the tech sheet under the hood.” That means the system prompt, the guardrails, the model card, and the eval and red-team results. Professional answer: they pull it up or explain it in plain English. Many keep agent instructions in an AGENTS.md file, an open format a growing list of AI coding agents read. Red flag: a deer-in-the-headlights stare, or “our AI is safe.” That’s likely a vibe coder, not someone who understands the risk you take on when you sign with an unregulated AI vendor.
If the answers are vague, the risk is still yours. Nobody has told you yet.
A Simple Ownership Plan
You don’t need a legal department. You need three habits.
- Hold the money words. Any draft that mentions price, payment, APR, down payment, trade value, discounts, or availability waits for a person to check it against the deal and the inventory.
- Keep your own records. Every draft, edit, sender, and timestamp, stored by the dealership.
- Name one owner. One person accountable for the tool, with a short written list of what AI may and may not say.
None of this slows down a good sales team. It just makes sure the fastest reply is also one you’d sign.
The vendor wrote the AI. Your name is on every message it sends. Run it like it’s yours, because it is.
Nathan Platter builds human-in-the-loop AI systems and runs Hold the Send, which helps dealers keep a person in charge of AI customer messages. This article is general education, not legal advice.
