A sequenced 90-day plan for bringing AI into a dealership: clean the data first, automate one workflow end to end, then measure and decide to scale or kill.

Every store needs one number the whole team trusts before it needs any AI, and you can build it in thirty days with the tools you already pay for.

What to automate first in the BDC, what must stay human, and the three-question test that tells you which is which.

Every report in your store disagrees for the same quiet reason, and a one-page data contract is how you make CRM, DMS, GA4, and call tracking tell one story.

What actually works, what is vendor theater, and the five plays I would run in your store this quarter.

What an agent actually is, what it is not, and when a loop with tools beats a workflow or loses to a cron job.

Why one model is never enough in production, and what a distribution and governance layer actually has to do about it.

The five-layer architecture that gives a dealership one set of numbers everyone trusts, and why AI belongs at the top of the stack, not the bottom.

Most dealers have a visibility problem. Ten dashboards that never talk to each other is not a strategy.

The design principles behind market-intelligence pipelines: freshness, lineage, and locality, and why most market reports fail all three.

What self-taught actually means when the classroom was a showroom floor, the curriculum was an agency you owned, and every claim has an artifact behind it.

Hourly billing punishes fast builders and rewards slow ones, which is why every engagement I sell is a defined outcome with a fixed price and a clean exit.

Why owned AI systems beat vendor lock-in, and how to tell the difference before you sign.

Per-seat pricing, switching costs, and the five-year math that decides whether your dealer tech stack is an asset or a permanent tax.

Decisions live in meetings, and the record evaporates the moment everyone hangs up, which makes meeting intelligence a governance problem, not a gadget.

Permissions, audit, distribution, and human gates: the four controls that decide whether AI agents become an asset or an incident report.