The Uncomfortable Questions · An AI FAQ for leaders

Question #27 · Economics

What’s the most expensive AI mistake your company could make that would never show up on a balance sheet?

The short answer

None of it is booked as a loss. All of it shows up eventually.

The costly failures are rarely the failed pilots. They’re the invisible ones: talent that leaves because the work stopped being meaningful, customers who drift away from a colder experience, judgment that erodes across the leadership bench.

Norm E. Murray

The economics

Why it happens

Organizations manage what they can count, and the most valuable things AI can damage are the hardest to count.

01

What gets measured is all that gets managed

The McNamara fallacy is the habit of treating what is easy to measure as what matters. AI business cases count licenses, hours and headcount, so the losses that sit outside those columns never enter the decision.

02

Meaning keeps people; pay only keeps them quiet

Frederick Herzberg found that pay and conditions prevent dissatisfaction, while meaningful work and achievement are what motivate. Strip the meaningful parts out of a role with automation and your best people leave long before anyone fixes the pay.

03

Customers leave without complaining

Albert Hirschman described two responses to decline: voice and exit. Most unhappy customers skip voice and simply exit, so a colder AI-driven experience shows up as quiet attrition, not as a complaint anyone can act on.

04

Automation erodes the skill it relies on

Lisanne Bainbridge called it the irony of automation: the more a system does, the less practice people get at the judgment needed when it fails. The leadership bench thins out one delegated decision at a time.

The bottom line: The balance sheet records what AI saves. Nobody is recording what it quietly spends.

Your move

What to do about it

  1. Add an invisible-cost line to every AI business case.Next to the savings, require a short note on what the change could do to talent, customers and judgment. If nobody can fill it in, the case is not finished.
  2. Watch who leaves after a rollout.Track regretted departures in teams where AI changed the work, and ask in exit conversations whether the role stopped feeling worth doing.
  3. Measure what customers stop doing.Complaints will not tell you. Watch repeat purchases, renewals and contact rates in the journeys you automated, before and after.
  4. Keep judgment in practice.Decide which calls your rising leaders still make themselves, without the tool, so the bench stays deep enough for the day the system is wrong.
  5. Review the soft costs at the board.Once a year, put talent, customer and judgment effects of AI on the board agenda next to the financial return, so they are owned by someone.