Own the System, Keep the Lift

The system that compounds is the one you own. I have watched dealers build entire BDC operations on a vendor platform, train their team on it for two years, and then get a price-hike letter with sixty days notice. The lift was real. The dependency was real. The leverage was entirely on the other side of the table. There is a better way to structure this, and it starts before you sign anything.

The lock-in spectrum: where most vendor deals actually land

Vendor lock-in is not a yes or no question. It is a spectrum, and most dealership deals land further toward the locked end than anyone admits at signing.

At the loose end sit the tools you could replace by Friday. A photo service. A scheduling link. If the price doubles, you switch and nobody on the floor notices.

The middle is where deals get expensive. These are the tools that technically let you leave but make leaving miserable. The export exists, but it arrives as a flat file with cryptic column names and no documentation. Your CRM history comes out, but the call recordings do not. You can go, but you go naked.

At the locked end, the vendor owns the workflow itself. Your BDC was trained on their screens. Your pay plans reference their reports. Your Monday meeting runs off their dashboard. The vendor knows exactly where you sit on this spectrum, and the renewal pricing reflects it.

Here is the test: if this vendor sent a 30 percent price increase tomorrow, could you credibly walk? If the answer is no, you are not a customer. You are a hostage with a login.

How to read a contract for portability before you care about price

Most people read a contract for price and term. Read it for exits first. Price only matters if you can leave.

Four clauses tell you almost everything:

Data export. What format, how often, at what cost. "Available upon written request" is a red flag. You want self-serve export, in a documented format, at no charge, while you are still a customer.
Derived data. The leads are yours. But who owns the call transcripts, the lead scores, the enriched records the platform built from your raw data? If the contract is silent, assume the vendor will claim them.
Renewal mechanics. Auto-renewal with a 90-day cancellation window is a trap that catches stores every year. Put the notice date on a calendar the day you sign.
Termination behavior. What happens to integrations, phone numbers, and historical reporting on the day the contract ends? Some platforms shut the door at midnight and keep your history behind it.

None of this shows up in the demo. All of it decides whether the lift you build over two years stays with the store or walks out with the vendor.

What "owned AI" means in practice: data, models, and workflows

Owned AI does not mean you wrote every line of code. It means three layers stay under your control.

Data. Your leads, your sales records, your transcripts, your service history, stored somewhere you control. Vendors get copies. You keep originals. I learned this early: at Lazare Auto Group, from 2009 to 2011, we built vendor-independent lead-source attribution with early Google Analytics and cookies, precisely so no vendor could grade its own homework. The principle has not aged a day.

Models. You do not need to own model weights. Frontier models are rentable by the API call, and they leapfrog each other every quarter. What you own is everything around the model: the prompts, the evaluations, the routing that decides which model handles which job. I run my own governance layer for this, Tool-Bag, which orchestrates Claude, Codex, and Gemini across 14 plugins, 108 skills, and 12 CI workflows. The point of that layer is that any single model can be swapped without touching the system around it. The models are interchangeable. The governance is the asset.

Workflows. The process lives in your documentation, not in a vendor portal. When the tool changes, the store does not have to relearn its job.

The build vs. buy vs. own framework for dealership tech decisions

Here is the framework I use with operators. Three options, strict order.

Buy when the capability is a commodity. Payroll, scheduling, basic chat. Nobody wins a market on payroll software. Buy the cheapest competent option and move on.

Build when the capability is your differentiator, and only if you can maintain what you build. That bar is high, and most single rooftops should not try to clear it. At OOMDO we built proprietary lead-capture and conversion tech because conversion was the entire business, and that decision helped drive $60M+ in client sales over a decade. An agency had that case. A store rarely does.

Own, always, the connective tissue: the data, the definitions, the reporting, the documented process. It is the cheapest layer to control and the most expensive to lose, and it is the layer dealers hand to vendors without noticing.

Most decisions are buy. A few are build. Every single one should leave the connective tissue in your hands.

Three questions to ask any AI vendor on the first call

You learn more in five minutes of pointed questions than in an hour of demo. Ask these on the first call.

If we cancel after a year, what exactly do we walk away with, and in what format? A confident vendor answers in a sentence. A lock-in business model answers in a paragraph.
Which AI models are you using, and what happens for us when a better one ships? You want to hear that models are swappable parts. If the answer is the name of a proprietary engine and nothing else, the engine is the lock.
Can our data flow out to systems you do not control, while we are still a customer? Webhooks, an API, scheduled exports. Anything less means your data only moves when it serves their renewal.

Write down the hesitations. The pauses are the real contract.

How to transition off a locked system without burning the store down

Leaving a locked platform is a sequencing problem. In the wrong order it torches a quarter. In the right order, the floor barely feels it.

Start exporting now, before you have decided anything. Monthly pulls of every record you can reach. You want a year of history in hand before any cancellation notice goes out.

Move reporting first. Rebuild your core numbers from your own exports and run the Monday meeting off them for a month, in parallel with the vendor dashboard. When the team trusts your numbers more than the portal, the psychological lock is already broken.

Then move the workflow one lane at a time. One lead source, one team, one process on the new system while everything else stays put. Expand as it proves out.

Cancel last. The contract should end after the lift has already moved, not before.

The lift was never the tool. It was the data, the process, and the people who run both. Structure the exit so all three stay home.

Keep what you build

Twenty years in automotive taught me one rule that survives every technology cycle: the store keeps the lift. From attribution at Lazare to the platforms I ship today, every system I build is structured so the leverage sits on the operator's side of the table. If you are staring at a renewal letter, or at a contract you have not signed yet, look at the work to see how owned systems get built, or check pricing if you want help reading the deal in front of you.