There is a version of my own story that is not true, and it is the flattering one.
In that version somebody builds a platform for seventy people alone, in ten months, and the interesting part is the person. In the actual version the interesting part is what had to be true around him.
I wrote about the risk side of this directly in why the first version needs one person and the second needs a team. This is the other half: what made it responsible rather than reckless.
What does a company have to do to let one person build a platform?
Two things, and both are decisions that get made once and then have to hold.
Ümit Tenekeci, our CEO, gave the go-ahead before there was anything to see. Not a budget, not a project plan. Permission to find out. Nobody asked for a specification, a vendor comparison or a return calculation for a thing that started as a folder for logo files. Had they asked, the honest answer would have been that I did not know yet, and the project would have ended there.
Ahmet Özbek, our CFO, did not treat it as a cost line. A finance function has every reason to want a number attached to an initiative before it consumes a person's attention for a year. That reason is legitimate and it kills most internal builds. Here it was set aside deliberately, not accidentally.
Why this is not a thank-you note
Because it is the load-bearing part of the argument I make everywhere else on this site.
I keep saying that the people who should build your AI already work for you. That is only half of it. The other half is that internal capability needs the same thing an external vendor gets automatically: a mandate, and someone senior who absorbs the uncertainty while it is still uncertain.
An outside supplier arrives with a contract, a scope and an invoice, and nobody questions whether they are allowed to spend the time. An internal person has to earn that permission from a standing start, usually against colleagues who are measured on something else.
That asymmetry is why most companies buy instead of build. Not because buying is better, but because buying is easier to authorise.
The transferable part
If you want the outcome, the decision to copy is not the technology. It is the willingness to let someone competent work on something for a while without knowing what it will become.
That is uncomfortable. It is also cheaper than the alternative by an order of magnitude, and I have the invoices we stopped paying to show it.