How I Approach Solving Problems With AI

The most useful thing I do with AI is not asking it for answers.

It is using AI to understand the problem.

That is probably my favorite part of the work. It is also where I think the human still brings the most value.

My experience is mostly in knowledge work: planning, training, staff products, exercise development, and decision-support tools. In that kind of work, the first step is rarely choosing a model or writing the perfect prompt.

The first step is understanding what is actually going on.

When I use AI to solve a problem, I usually start by trying to frame it. What is the real issue? Who is affected? What decision needs to be made? What product has to exist at the end? What standards or constraints matter?

Sometimes that means dividing the work into smaller parts. Sometimes it means picking apart the logic: assumptions, risks, missing information, and second-order effects.

Only after that do I start asking the model to produce something.

A quick example:

I was in an all-hands meeting with our CEO when someone asked about noncombatant evacuation operations. In plain terms: if the situation got bad, how would the company evacuate contractors?

The CEO was coming to our office later for a briefing. When the meeting ended, I had about 20 minutes to add something useful.

I did not start with research.

I started with a story.

I asked the model to write a short scenario around the evacuation problem. Because I had been a NEO NCO in the 1980s, I knew what context the model needed and what sounded realistic.

Once the scenario exposed the problem, I had the model draft a company policy response to solve it.

Was that a final answer?

No.

But it gave me a briefable first draft under time pressure. It also referenced relevant DoD and Army regulations, which gave me a starting point for follow-up review.

That is how I think about AI problem-solving.

AI helps me frame the problem, expose assumptions, generate a first draft, and move faster from vague concern to usable product.

But the human still owns the context.

The human still owns the judgment.

And the human still owns the final answer.