Somebody sent me a Geoff Woods video this week — How to Use AI to Make Money, Grow Your Business, and Be More Productive — and I want to write about it honestly, because most of what gets published in response to a video like that is either uncritical applause or a competitor pretending the idea is obvious.
It is not obvious. Woods is the founder of AI Leadership and the author of The AI-Driven Leader. Before that he was Chief Growth Officer at Jindal Steel & Power during a stretch when its market cap went from roughly $750 million to $12 billion. He is not a prompt-tips guy who discovered ChatGPT in a webinar. He is an operator who ran a real transformation and then went back to describe the method.
And his framework contains one move that almost nobody makes.
CRIT, and the step everyone drops
Woods teaches CRIT: Context, Role, Interview, Task. You give the model rich context about your actual situation. You assign it a specific expert role. You have it interview you, one question at a time, to surface the context you did not think to give it. Only then do you hand it the task.
Watch what people actually do with that. They do Context — a paragraph about their business. They do Role — “act as a marketing strategist.” They do Task — “write me a plan.”
They skip the I.
The Interview step is the entire framework. The other three are hygiene. Interview is the one that changes what comes out the other side, and it is the one that gets dropped, because it is the only step that costs you something. It costs you fifteen minutes and the mild discomfort of being asked questions you do not have clean answers to.
What that looks like in the trades
Here is the version I would hand a contractor. You want a plan for next season. The instinct is to type “build me a marketing plan for an HVAC company in Dallas” and grade whatever comes back. You will get something competent and generic, because you asked a generic question, and you will conclude the technology is overrated.
Instead, end the prompt with this:
Before you answer, interview me one question at a time until you have enough to give me a plan I could not have written myself. Do not give me the plan until you are done asking.
Then answer honestly. It is going to ask what your close rate is on replacements versus repairs. What your average ticket is in July versus March. Which two zip codes your best techs hate driving to. Whether your phones are answered by a person at 7pm. How many of last year's maintenance agreements actually renewed.
Most owners cannot answer four of those without opening something. That is not a failure of the exercise. That is the exercise. The questions you cannot answer are the report you actually needed, and you got it for free from a machine that was only trying to help you write a marketing plan.
Why it works, in one sentence
A model with no context gives you the average of the internet. A model with your context gives you something about your business. The Interview step exists because you do not know what context you are withholding — nobody does, about their own operation, because the things you know best are the things you have stopped noticing.
Interrogation surfaces them. That is the whole mechanism, and it is why the step is worth more than every prompt template ever posted.
And now the part where I push
Woods is moving leaders from AI as a tool to AI as a thinking partner. That is a real promotion and most of this industry has not made it yet. I would put the video in front of any contractor who is currently using AI to write Facebook captions.
But a thinking partner is rung two.
A thinking partner still waits for you. It still requires you to open the window, bring the problem, and carry the answer back out to the business by hand. Every insight it produces is gated on you having the time and the discipline to go ask. That is enormously better than nothing. It is not the end state, and I do not want the trades to stop there and think they have arrived — the way a lot of companies stopped at a stack of automations and called it a strategy.
Rung three is delegation. Not a partner you consult — agents that hold context permanently, act across your systems, and route the expensive decisions back to a human on purpose. The difference between asking a brilliant advisor what to do and having a team that already did it. That is orchestration, it needs an architecture underneath it, and you cannot prompt your way there — bolt-on versus built is the whole question.
The full progression is on the maturity ladder. CRIT gets you cleanly onto rung two. Most people reading this are on rung one and think they are on three because they have a subscription.
The honest caveat
None of this replaces the judgment. The model can interview you brilliantly and still be confidently wrong about your market, because it has no stake in the outcome and will not be there in March when the decision lands. It does not know which of your people can carry a change and which one will quietly stop trying. Leadership is not the part that automates.
What Woods gets right, and what I would underline for anyone in the trades, is that the constraint was never the technology. Everybody has the same models. The constraint is whether you are willing to sit still for fifteen minutes and answer hard questions about your own business before you ask for an answer.
Do this on Monday
- Take one real decision you are carrying right now — a hire, a service line, a market, a price change.
- Give context and a role, then add: interview me one question at a time until you can give me something I could not have written myself.
- Answer honestly, including “I don't know.”
- Write down every question you could not answer. That list is worth more than the plan. Go get those numbers.
- Then ask for the task. Compare it to what you would have gotten cold.
Do that once and you will understand why the I is the framework. Do it four times and you will start wanting the thing on rung three, which is when we should talk.
Credit where it belongs: the video is here, and Woods' book is The AI-Driven Leader. Watch it, then go get interviewed by a machine about your own company. It is a strange fifteen minutes and I have never seen anyone come out of it unchanged.