HALDEMAN Notes · 02 ·

The cost of waiting belongs in every AI decision.

Doing nothing has a price. Put it on the page.


AI decisions arrive with one side priced: the cost of acting. The cost of waiting shows up blank. So committees pick the blank.

Waiting costs two ways. Value you don’t scale where something works. And damage you don’t manage where it doesn’t: BetterUp Labs and Stanford found 40 percent of full-time US employees received AI “workslop” in a month, at about two hours to untangle each time.

A waiting estimate is honest when you can argue with it: how many people, how much time, what an hour costs, and how much saved time turns into anything real.

Waiting is the right call in cases like these. The evidence is still too thin to tell a working pilot from a lucky one. Something it depends on isn’t ready, like data access or a quality check. Or a small test you can undo will settle it this quarter. Then waiting is a decision with a date on it.

One page, both prices

Act now
Your vendor’s quote, plus your teams’ time to set it up and change the work.
Wait a quarter
The hours the change would have saved, at your own hourly cost.
Biggest uncertainty
Acting: how much of your teams’ time the change takes. Waiting: how much of the saved time becomes value.
What would change the choice
A second team reproducing half the change by its next check.

So: make every AI decision paper price both sides, each with its biggest uncertainty. A committee comparing two prices decides, and sometimes the answer is to wait, with a date.

Sources checked September 2026

Source

BetterUp Labs and Stanford Social Media Lab, “AI-Generated ‘Workslop’ Is Destroying Productivity,” Harvard Business Review, September 2025. Read the original

See how a 90-day plan sets its gates Ask Albert: what does waiting cost us? Next note
Ask Albert