Write down the Unwritten rules…
On our latest Auto Collabs, Conversica CEO Mike Gilson described what top salespeople do every day. They keep five screens open across the CRM, the DMS, email logs, and call logs, and piece together a customer’s story by hand. The data already exists. What doesn’t exist anywhere is the reasoning that connects it.
“[T]here’s no AI system that can reason across data without that underlying context that powers the industry.” Mike Gilson, Conversica CEO
Kyle Mountsier has spent most of this year on that missing layer: the decisions managers and top performers make offline that nobody has ever written down. Software teams solve the same problem with a context file, often called AGENTS.md, that tells AI tools how their codebase works. Dealerships need an equivalent, and the people who carry it are already on staff.
The takeaway: Any AI tool you bring in can only use the context someone has documented. Start with the judgment calls your best people make.
Run this prompt with one top performer in a single sitting
Pick one person and one type of opportunity: a payment swap, a lease maturing in three months, a declined service, an open recall, or a household with several vehicles. Let them answer out loud or by typing. Paste this into ChatGPT, Claude or Gemini:
You're helping a car dealership document the unwritten decision rules its best people use. Interview me, one question at a time, about how I handle [TYPE OF OPPORTUNITY, e.g. "a customer whose lease matures in the next 90 days"]. I'm a [ROLE] at a [BRAND] dealership.
Rules for the interview:
- Ask about specific past situations, then ask what rule I was following.
- When I give a rule, ask about the exceptions and the signals that tell me an exception applies.
- Ask what I check, in what order, and in which system.
- Ask what I would never do in this situation, and why.
- Stop after 10 to 12 questions, or sooner if I say "done."
- Remind me not to share customer names, contact details or account numbers.
When the interview ends, write a one-page playbook with these headings:
Situation | Signals I look for | Where I find them | What I do next | What I never do | Exceptions
Write it in plain language a new hire could follow.What to do with the playbook it produces
Have a manager review it. Where they disagree is where your store has never had a standard.
Run the same scenario with a second top performer. Comparing the two playbooks shows which rules belong to the store and which belong to one person.
Save each playbook in one shared place. Over time, that collection becomes the context any future AI tool will need from you.
If your best salesperson left tomorrow, which of their decisions could anyone else in the building explain?
