In this shadow AI case study, a franchised dealership’s staff were pasting customer data into personal ChatGPT and Claude accounts. Helion replaced those accounts with one managed AI platform tied to the dealership’s sign-in, so anyone who left lost access immediately.
Last updated: November 16, 2026
Employees at a franchised dealership had signed up for ChatGPT and Claude on their own, with personal accounts. They were pasting customer data into those accounts and running prompts that included dealership information, and nobody at the store had any visibility into who was doing it or what data was going where. Nobody.
If one of those employees left, the data in their account left with them. So did everything else they’d built in it, customer data or not. That’s a real problem. It doesn’t announce itself, either, because nothing stops working on the day somebody walks out with it.
I should say where I stand on AI, because this isn’t an anti-AI story. Years ago, while I was teaching an MCSE course part-time, I earned the certification myself in 7 weeks, one test a week, studying from books. AI got the same treatment. Learn it early, use it, get there first. What I push on is the part nobody wants to hear about, the guardrails, because they’re what actually protect the business.
This post is one dealership’s story of putting those guardrails in place with managed AI for dealerships. If you’re still finding out what’s in use at your own store, start with our guide to shadow AI risks at the dealership and come back to this one after.

What Was Happening With Staff AI Use at This Dealership?
Shadow AI is shadow IT with a new name. Somebody finds a tool, signs up with an email address, and starts using it for work. The business finds out later, if it finds out at all. That’s how the mess starts, one sign-up at a time. At this dealership the tool was a chat window, and what went into it was the work itself: customer data pasted in, and prompts built around dealership information.
None of it was planned. There was no record of who had an account, no view of what had gone into one, and no way for the store to shut down an account it never opened. Customer data was sitting in accounts the dealership didn’t own. Who owns that? Right then, nobody the dealership could point to.
The accounts were theirs. The data wasn’t.
What Did Employee AI Accounts Put at Risk?
Employee AI accounts put two things at risk at once: customer data the dealership couldn’t track, and work it couldn’t keep. Whatever went into a personal account stayed with the person who owned it, and when that person left, the data and everything they’d built in the account went with them.
Dealers aren’t alone in this. Staff in the FTC’s Office of Technology made a related point in a January 2024 post aimed at AI companies, noting that customers “may reveal sensitive or confidential information when using a company’s models, such as internal documents and even their own users’ data.” Read “users” as your customers and that’s this store.
The second risk is easier to miss, and I think it’s the one dealers end up feeling later. Even when nothing sensitive was in an account, the prompts an employee had written and the work they’d built up in it were dealership work, and the dealership had no copy of any of it. When they left, all of it walked out the door with them.
Both risks share a root. The account wasn’t the dealership’s.

How Did Helion Bring Staff AI Use Under Control?
We replaced the personal accounts with one managed AI platform and connected it to the authentication system the dealership already used. From that point on, AI access at work was the dealership’s call, and staff who had a seat kept the major models they’d been using all along.
- It started with the sign-in. The platform hooked into the system staff already used to sign in at work, so AI access got handled alongside the rest of their access instead of in accounts they opened on their own.
- Seats were the dealership’s call. Before, that decision got made one personal sign-up at a time.
- What about people who left? They were locked out of the AI environment immediately.
- Each prompt was routed to a suitable model, so nobody had to decide whether a task was a Claude job or a ChatGPT job.
- The bill was monthly, per seat, and seats could be added or removed.
Then there’s cost. One seat on the managed platform covers all the major models: GPT, Claude, Gemini, and others. It also costs less than paying for separate subscriptions to each AI tool. We deploy the platform and manage it. If you want the piece-by-piece view of the workspace your staff would actually use, it’s on our page about secure AI chat for dealerships.
From then on, the store decided.

What Changed Once Staff Moved to One Managed Platform?
The short version is that the personal ChatGPT and Claude accounts were replaced by one managed platform the dealership controlled. The longer version fits in a table.
| Question | Before | After |
|---|---|---|
| Where staff used AI | Personal ChatGPT and Claude accounts they opened themselves | One managed platform tied to the dealership’s sign-in |
| Who decided who had access | Each employee, by signing up | The dealership, seat by seat |
| What happened when someone left | Their data and their work left with them | Their AI access ended immediately |
| Which models staff could reach | Whatever their own account offered | GPT, Claude, Gemini, and others, with each prompt routed to a suitable model |
What changed was whose account the work happened in. The dealership knew who had a seat, because it was the one handing them out.
Same models. New owner.

What Should Your Dealership Take From This Shadow AI Case Study?
Start with ownership. Before anybody argues about which AI tool is best, settle whose accounts the work happens in, and build everything else around that answer. Here’s the order I’d take it in at your store.
- Ask your people which AI tools they use, and on whose account. Ask it as a question, not an accusation.
- Write down what can and can’t go into AI. Our dealership AI use policy template gives you starting language, including where personal accounts stand.
- Who decides who gets a seat? Put one name next to that decision and write it down. That’s dealership AI governance at its simplest.
- Give staff an approved place to use AI, on a seat the dealership owns, before you ask them to stop using their own.
- Add AI to the exit process. Our employee offboarding checklist for dealerships covers AI accounts alongside Microsoft 365 and the rest of a departing employee’s access.
The longer nobody owns this, the longer that first list gets.
Ownership first. Tools second.

Questions About This Shadow AI Case
Did staff lose ChatGPT and Claude when the dealership moved off personal accounts?
What happens to an employee’s AI access on their last day?
Was the point to stop staff from using AI?
Was customer data the only thing at risk?
Couldn’t each employee just change the training setting on their own account?
Own the Accounts
At this dealership the employees chose their AI tools before anyone at the store had a say. By the time the dealership stepped in, the choice that was left was who would own the accounts the work happened in. It chose to own them.
If you want to know whether the same thing is happening at your store, start with our complimentary IT and cybersecurity assessment. It’s a no-cost, no-obligation review of your technology and security, and it also covers AI tool usage, which the assessment page describes as “whether staff are using AI tools in personal accounts outside your security controls, and what a governed program would look like alongside your existing stack.” The findings come back in a written report that goes to the people at your store who need to see it, instead of getting lost in a hallway conversation.
Keep the AI. Own the accounts.

