I spent 30 hours writing a 45,000-word prompt. Here's what I learned.

This year I rearranged my calendar to carve out 10 to 20 hours a week to work with AI. Not to read about it. To build with it. My clients expect me to advise them on strategy, and I knew I couldn't do that on AI without making a real investment of my own.

I spent 30 of those hours building an elaborate tool: 45,000 words of instructions on how to assess the impact of AI on markets. It wasn't just encoding what I knew - it was collaborating with AI to create new frameworks for understanding AI's impact on markets.

The 45,000 words are actually 5 different tools, that all come together to give a sophisticated view of the impact of AI on 4 aspects of any company's markets.

I've run it on several companies, and the machine works well. The thinking is sophisticated - with strong strategy and detailed analysis.

Around hour 20 of the building, I thought, "I wonder if all this is worthwhile." So, after I set the tool aside for a week, I opened a blank window and ran a fifteen-minute prompt on one of the companies I'd already used my tool on.

What I got was pretty much the same answer. Not the same detail, but pretty close at the broad, high level.

So, was the 30 hours wasted? Are 45,000-word prompts worth the investment? (Or, more accurately, why or when are 30 hours worth it?)

The 30 hours, broken down honestly

The first 10 built a real understanding of how an AI world reshapes business. Useful whether or not I ever built a machine, and far more robust than anything I'd have assembled without AI as a thinking partner. I came out of it with 5 frameworks I can now apply to any company, anytime, with or without a tool in front of me.

The second 10 taught me how to actually work with AI on something large and complex. This is scouting. My clients pay me to understand terrain before they walk into it. I can't advise anyone on building with AI if I've never done it at scale myself.

Only the last 10 - building the intricate guts of the machine - were a waste.

One-third misspent. Two-thirds worthwhile. That's a very different story than "30 hours down the drain."

What the example proves, and what it doesn't. The quick prompt matched the machine...this time. It might be good enough. It might also be confidently, expensively wrong, without really being able to tell why.

It's ok to outsource thinking. It's not ok to outsource understanding.

When there's real money riding on a decision, "it usually gets close" isn't good enough. That's when a more robust machine earns its place, not because it's smarter on average, but because it's more trustworthy when the decision is important and the downside is significant.

Because I, or the machine, can explain to you how we got the answer we did, and you can calibrate that to your own experience.

The number that actually reframed it for me. With one hour of my own work, I can build a 50-100-word prompt that'll get 20-50% better output than what my client would write on their own. I'll add a concept to check that a generic prompt wouldn't. I'll give a warning about an area that seems useful but isn't. Depending on the situation, that expertise is somewhere between modestly and meaningfully beneficial.

With a 500-1,000-word prompt, or an hour collaborating with AI, I can get output that's 50-200% better than a generic prompt. That's when my rigorous frameworks, experience, and embodied judgment really come into play.

I'll say what any builder knows: (1) work is more effective with tools than without them, and (2) experts get more out of tools than novices...and that compounds over reps.

Now some real talk about my own business

The biggest strategic question I'm sitting with: do I publish my frameworks, or keep them private?

I've decided to keep them private. They're my competitive advantage. More robust, more contextualized, more useful than what a generic prompt will give.

That might turn out to be a mistake. Plenty of sharp people are betting the other way, giving everything away to build an audience.

But I keep landing on the same conviction: as AI commoditizes knowledge, judgment and taste become the scarce goods. And my frameworks are the encoding of my judgment. I'll build them into my machines - my "judgment engines" - but only my best clients will see how those engines are constructed.

Where that leaves me. I'll keep spending 10 to 20 hours a week learning and building with AI. I honestly don't know what my role looks like in 5 years. AI can already generate useful frameworks that I used to be the one to provide. That unsettles me.

That unsettling feeling is driving me to spend 10-20 hours each week to build with AI. My business will shift, and my job is to make sure my edge shifts with it. Maybe it moves from intelligence to hustle. From being knowledgeable, to being the one who puts in the reps.

If that's the game now - reps, effort, hands in the dirt, week after week - well...let's go, AI.

PS - if you want to see what my machine says about your business, run the following prompt and then send me the output at dave@phimation.com and I'll run your company through my machine and we can see what we get.

The prompt: You're a sophisticated market research strategist. Give me a detailed analysis of how AI will impact my markets.

PPS - some societal implications of my article should also be acknowledged. I have a moral responsibility to make sure I am giving the next generation opportunities to learn with me. I've had 5 great interns this summer who have walked some of this journey with me, and as a result, I think they're smarter about their own abilities with AI, and understand better how AI can help accelerate their learning. But I didn't just give them my frameworks - I had them build their own. And just like the article says...they were pretty close to what I came up with, given that they have 30 years less experience than I do. I think we need this next generation more than ever to figure out this new world, and I'm hopeful that we can show them how to learn faster and better than past generations did, and that they'll make the world better and make a better living than we did.