The Uncomfortable Questions · An AI FAQ for leaders

Question #14 · Talent

How many of your managers are using AI to look productive rather than to be productive?

The short answer

AI makes volume cheap: more emails, more decks, more summaries.

Activity metrics go up while real output stays flat. Leaders who reward volume will get a flood of polished noise. Measure decisions made and problems solved, not documents produced.

Norm E. Murray

The psychology

Why it happens

When output costs nothing to produce, the signals leaders use to judge effort stop meaning anything.

01

The measure becomes the target

Goodhart’s law holds that when a measure becomes a target, it stops being a good measure. If managers are judged on visible output, AI gives them an effortless way to hit the target without moving the result.

02

Cheap signals stop signaling

Michael Spence’s signaling theory explains that a signal is credible only when it is costly to send. A polished deck used to signal hours of thinking; when AI makes it nearly free, it signals very little.

03

The easy question wins

Daniel Kahneman described attribute substitution: faced with a hard question, we quietly answer an easier one. Whether a manager is effective is hard to judge; whether they produce a lot is easy, and AI makes the easy answer look great.

04

People do what they are rewarded for

Agency theory describes how people optimize for what they are rewarded on, not for what the organization actually wants. If recognition follows visible activity, AI becomes a machine for producing activity.

The bottom line: If your scorecard rewards volume, AI will deliver volume and very little else.

Your move

What to do about it

  1. Stop counting artifacts.Take documents produced, messages sent and decks delivered out of manager scorecards and reviews.
  2. Ask what changed.In reviews, ask which decisions each manager made and which problems they closed this quarter, and what AI contributed.
  3. Cap the noise.Set a ceiling on recurring reports and status updates. When AI produces a new one, an old one retires.
  4. Trace the output.Each month, pick a few AI-assisted deliverables and check whether anyone acted on them.
  5. Model it from the top.Send fewer, shorter, decision-focused updates yourself, so the team sees what you actually value.