What AI actually changes about product management

Placeholder content, shipped with the site to demonstrate the article layout. Not a finished piece. Replace or delete it.

An outline of the argument, so the real version has somewhere to start.

#The question that is not interesting

Which model is best. It changes every few weeks, everyone has an opinion, and almost none of it affects what a product team should do on Monday.

#The question that is

What happens to a product team's process when producing a plausible first version of almost anything becomes cheap.

Some of the answers are genuinely surprising. Most of them are boring. Being straight about which is which is roughly the entire value of writing this down.

#Research

Synthesis got cheap. Deciding what is worth researching did not.

#Design

The first version costs an afternoon, so the cost of being wrong dropped, so the argument for deciding early got weaker. Teams that have not noticed this are still holding meetings to avoid work that now takes less time than the meeting.

#Shipping

Stage	Used to be	Increasingly is
First draft	Slow, expensive	Cheap, fast
Deciding what to build	Cheap, fast	The whole job
Review	A gate	A bottleneck

#Measuring

Unchanged, and now the most valuable thing in the building. Placeholder note: the real section should be blunt about why.

#The uncomfortable part

If building stops being the constraint, then a team's output is limited by the quality of its decisions. That is a much harder thing to hire for.

#What I would actually tell a team

Placeholder list, to be replaced with the version I would defend:

• Stop asking whether to use it and start asking which step it removes
• Instrument the thing before you accelerate the thing
• Be suspicious of any improvement you cannot see in a number