I attended an online panel about AI and behavioral science and walked away with a simple thought: like every wave of tech before it, AI amplifies the best in us and the worst.
"WORKSLOP'
Harvard Business Review: Workslop refers to "AI-generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task."
AI-generated work that looks right but doesn’t move the task forward.
Recent research (Stanford + BetterUp, published in HBR) estimates workslop is costing large companies millions in rework and lost momentum. So not a productivity gain we hoped for I guess.
I was talking to a group of friends about how a lot of AI-generated content is like a speech of a politician - a lot of things said, not a lot of meaning delivered.
The cure?Using AI to make good work, good thinking better, not replace it (which it can’t)
The curse, in the words of a Nobel Prize-winning behavioral economist Daniel Kahneman,
"Thinking is to humans as swimming is to cats; we can do it, but we'd prefer not to"
Another friend, a consultant, recently complained to me, confirming the observation,
“Now with AI I have to do MORE work - people are bringing me all kind of work that “meets all the specs” and is total bullshit, it’s like I can’t trust any of that any more”
Quantity AND Quality (or reps that actually matter)
Behavioral science already gave us a fix for this "perk" of human nature to look for the easiest thing. Don’t just count activity; design the process so the reps stack into desired outcomes.
A mix of quantative and qualitative measures.
Now just what's being done but also how it's being done to produce the outcome we want.
The solution:
-
Lead + Lag metrics. Lead = the reps you can do today that might predict success. Lag = the outcome you want. In AI rollout, leads might be “times AI used for decision-making,” while lags are “quality of services delivered or lowered, measurable risks” and “customer NPS after the change.” In content, leads are “amount of content pieces produced,” lags are “sales cycle shortened” or “more demo requests from potential customers matching predefined profile”
-
Goodhart’s Law. “When a measure becomes a target, it ceases to be a good measure.” The moment you reward only the count, people will game the count. So you must pair quantity with a counter-metric for quality.
Think fitness: drop 10kg fast on a crash diet and you lose water and muscle. The scale applauds; your future performance and health don’t. Same at work: 20 glossy AI decks this week could increase confusion and delay decisions.
If the reps don’t deliver better outcomes, more reps aren't gainz.
The Cobra Effect
The classic cautionary tale from real life.
British officials in colonial India offered bounties for dead cobras to reduce amount of cobras.
People began breeding cobras for the reward.
When the bounty ended, breeders released the snakes.
Outcome: More cobras than before.
The “Cobra Effect” entered the lexicon to describe incentives that backfire.
A documented version exists in 1902 Hanoi: the French paid for rat tails, hunters cut tails and released rats to breed more - rat farms even appeared. Measures hit target; reality got worse.
Back to the Goodhart's Law: the moment you reward only the count, people will game the count.
Over to you dear reader,
Whether you are trying to build a high-performing team, improve your health, or increase revenue or innovation in your business you always need to make sure that what you measure reflects a) some quantifiable results you measure real time (like amount of reps or customers) b) AND you need to make sure that over time that leads to the result you want - vs just doing the same reps for years or getting customers that aren’t profitable and will bankrupt your business if continued.
What’s the goal of the goal? And do you have metrics in place to measure what matters, not just doing more?
