Every board asks about the AI strategy. Every team has ChatGPT, or Claude, or Copilot. And the work that actually runs the company — the quote, the onboarding, the close, the incident at 2am — looks almost exactly like it did three years ago.
Chat logs are not a process. Completions are not a workflow. People have AI on their desktops; the business they work inside does not.
The models are good enough, and have been for a while. What they don’t have is any standing in your business — no authority to act, no ownership of how it turns out, no sense of what needs an approval, and no way to reach across the four systems a single request touches. So they offer advice, and a person does the work anyway.
A better model won’t change that. One that scores higher still doesn’t know your spend limits, your escalation path, or the vendor you stopped using last quarter. Those aren’t things a model can be smart enough to guess. It demos well. It doesn’t run anything.
There is a reason AI can’t take the work. How your business actually runs — how you quote, how you onboard a client, how you close the books, how you handle that incident — was never written down in a form anything could act on. It lives in people’s heads, in half-updated documents, and in chat threads nobody can find.
None of this is a failure of effort. It is what happens to any business that grows faster than it can write itself down. The companies that escape it are the ones large enough to afford the ceremony — six-figure implementations, multi-year customizations, and a standing room of consultants to keep it all current.
Follow a single request all the way through — an incident, a new hire, a renewal. It will take somewhere between six and ten human touches before it is finished, and almost none of them are judgment. They are re-typing, chasing, re-explaining, and waiting on someone to notice. Every handoff adds a delay, and the delays compound.
Most of what your team does in a day isn’t the job. It is the coordination the job requires, carried by people because nothing else can carry it.
What is missing is a system — one that knows the shape of your business, is trusted to carry the routine work end to end, and changes when the business changes. Until recently, a system like that could not be built at a price most businesses could pay.
A wave of AI-native enterprise platforms has arrived, and most of them are being built by people who have never had to run enterprise software. It shows in the same place every time: there is nothing underneath. No system of record. No durable thing the work attaches to, no lifecycle, no permissions worth the name, no history you could audit a year from now.
You can see it in the interface. A great many of these products live inside Teams or Slack — the assistant is a participant in a channel, and the state of your business is whatever has been scrolled past in that thread. That is a fine way to demo. An interface is not a platform, and the record is the one part you cannot add afterward.
Which is the problem at the top of this page, seen from the other side. AI cannot take authority over your work when there is nothing durable to take authority over.
We built the platform first — a real enterprise system of record, modular, shaped by thirty years of learning what enterprise software has to get right — and then, as we went, worked in every place where an LLM can genuinely carry the load. p6k will run without a single LLM interaction. Doing that would waste the point of it. But the platform is what makes the intelligence useful, not the other way round.
You already have playbooks. They just aren’t running anything yet.
See the Platform