AI at the Start, Then Debrief.

For years I treated software as the final polish. Think it all through myself, then use a tool to clean it up. Now I bring the tool in at the beginning, when the problem is still messy. I ask it to argue with me, to generate options I have not considered and to find the holes in my first instinct.

Then I do what only I can do. I decide. Afterward, I debrief, the same way I debriefed every flight. What did the tool see that I missed? What did I see that it could not? Every debrief makes the next loop better.

Why the start matters

Artificial intelligence is very good at producing answers. It is much less reliable at knowing which question matters, which answer to trust and what to do when two good answers point in opposite directions. Those are judgment calls. Used at the end, the tool polishes a view you have already locked. Used at the start, it widens the picture before you commit.

Two old disciplines, one new partner

The habit borrows from techniques that predate the tools. Psychologist Gary Klein’s premortem, described in Harvard Business Review in 2007, asks a team to assume the project they are planning has just failed and to generate plausible reasons for its demise. Klein built it because people are reluctant to voice their reservations during planning. A machine has no such reluctance, and no ego invested in your first instinct. It makes a tireless premortem partner.

The debrief comes from the cockpit. Wharton professor Ethan Mollick’s first rule in Co-Intelligence is to always invite AI to the table. I agree, with one addition. The decision stays at the table with you, and so does the debrief.

Run one loop this week

1. Bring a real problemSomething you are deciding this week, not a demo.
2. Open with the machineBefore you form your own view, ask for options, objections and blind spots. Ask it to argue the other side.
3. Decide yourselfThe tool widens the picture. The call stays with you.
4. Write a short debriefWhat did the tool show you that you would have missed? What did you see that it could not? What would you do differently next time?