In 1999, I watched a talk by Douglas Engelbart, the man who invented the mouse. He said technology moves on two dimensions. Access is whether you can reach the information, and connectivity is whether it can move once you have it. Everything advances when one of those expands, and at any given moment, one of them is what’s holding you back.
I was twenty-something at the time, running a web development company out of Eugene, and his words landed because I was living the second half of it. We could build almost anything, and almost nothing could talk to anything else.
Ten years later I was running a marketing agency for car dealers, and the same bottleneck showed up wearing different clothes. We would drive more leads into a store and the store would not sell more cars. For longer than I want to admit, I treated that as a marketing problem. The information that would have closed those deals already existed inside the building. It could not move to the person who needed it in time to matter.
Twenty-seven years later, it is still the most useful idea I know about the car business.
What makes it useful is pairing it with the thing operations people call the theory of constraints. The plain version is something every store already lives. Your business runs at the speed of its worst bottleneck. One lift down in service and it does not matter how fast anybody else is working. So the only work that changes the number is work on the bottleneck. Everything else keeps you busy.
Which means when you see a business running with remarkable efficiency, they are not smarter than you. They are running closer to the ceiling of what their access and connectivity allow.
Which is where our industry has been sitting for a while. We all spent fifteen years raising one of those dimensions (access) and ignoring the other (connectivity), so we are massively over-softwared and dramatically under-connected. The average store runs more tools than it did a decade ago and can answer fewer questions about itself.
Two things changed, and they only matter together
The first one everybody noticed, which is writing software got close to free. Work that took a developer three weeks only takes a conversation in Claude or ChatGPT now. On its own, that produced a lot of demos and not much else.
The second one nobody noticed primarily because “data plumbing” is boring. That is, AI is useless without connected, accurate data to connect to. Point it at a dealership with no connected data and you have a yappy intern. It will write you a beautiful application about nothing whatsoever.
Put those two together and a kind of problem opens that has been closed since the day this industry started.
I’m talking about every problem in your store that was too small and boutique to build a software company around. It might be the strange one that is specific to your market, your brand, your layout and your people, or the one that eats four hours of somebody’s Tuesday every week. Nobody was ever coming to build that for you, because the economics of packaged software never justified it and never will.
That category just opened.
The people doing it are not who you’d guess
In September, about sixty operators spent two days in Atlanta at a hands-on AI build event. They were not engineers. They were GMs, BDC directors and variable ops leaders, the people who know what’s wrong with a DMS because they fight it every morning.
What came out of that room included appointment prep, call sheets for return web visitors nobody had contacted, and flags for customers sitting in positive equity. These were things that had been sitting on a vendor roadmap for two years, or more often, had never made it onto one because nobody would have paid for them.
What I keep thinking about is who turned out to be good at it. The people who got the most out of it were the ones who could describe their own problem most precisely, which had very little to do with how technical they were. That moves power toward operators who understand their own business in detail, which is most of the people reading this.
What happens next is not chaos
The reflex response is that dealers building their own tools ends badly. I’ve noticed the people most confident about that are usually the ones selling packaged goods.
What I saw looked like collaboration. Operators showed each other what they made, then took somebody else’s idea and bent it to fit their own store. It felt closer to the early web than to a support nightmare.
Markets don’t die when this happens, they get smarter. And software doesn’t go away either. It just has to be worth paying for in a world where the alternative is a Tuesday afternoon and a clear description of the problem, which is a healthier market than the one we’ve had.
If you’re waiting for permission to start, you won’t get it. Nobody sends a memo.
And when you start, don’t build the impressive thing. Build the ugly, specific, unglamorous thing that annoys you every Tuesday afternoon. That’s the one nobody was ever going to build for you, and it’s the one that pays.
