I tried to build a pricing strategy in ChatGPT last year. Not as an experiment. I actually needed one.
The first response was good. Genuinely good. Clear framing, smart structure, the kind of thing that makes you think “okay, maybe I don't need to hire anyone for this.”
By the fifth prompt I was managing the AI more than it was helping me.
By the tenth I was copy-pasting between three conversations, reformatting in Google Docs, and trying to remember which version of the competitive analysis was the one I'd actually validated. I caught myself thinking I should've just done the whole thing from scratch.
The thing nobody talks about
These models are smart enough. That's not the issue. The issue is that chat tools are built for one thread. One question, one answer, follow up, refine. That's great for drafting an email or brainstorming names for your dog.
Strategy doesn't work like that. A pricing strategy needs a competitive analysis that feeds into financial projections that shapes a go-to-market plan. Change one assumption and three deliverables shift. In a chat window, you're the one holding all of that together. You become the project manager, the quality checker, and the integration layer, all while trying to also be the person who thinks about the actual problem.
That orchestration work? That's the hard part. And the tool doesn't do any of it.
The test I keep coming back to
After using an AI tool for a project, can you share the output directly with your VP? Your investors? Your client? Not after reformatting. Not after fact-checking. Not after stitching five conversations into a coherent doc. As-is.
If the answer is no, what you got was raw material. Raw material has value. But the hardest work is still on you.
Projects, not prompts
So I built a system that works the way real projects work. The analysis is examined from multiple angles, not just one. Quality is scored before you see it. Sources are verified before delivery. And the conclusions get challenged before they reach you, because unchecked work, from any source, is a liability when real decisions are on the line.
The output arrives ready to share. Not because it's perfect; no analysis is. But because it's been structured, checked, and challenged before it reaches you. You apply your judgment. You steer it. You make the calls. But you're not also doing the plumbing.
Chat tools are built for questions. The work I care about requires projects. That's a different problem, and it needs a different kind of system.