A perfect ai agent does work that people were never good at.
TL;DR
Three things people never did well.
- Measuring enormous numbers of changes in real time.
- Collecting data and turning it into information.
- Reacting to a complex event, the moment it happens.
People are bad at all three, and not through any fault — we sleep, we get bored, and we cannot watch and run calculations on ten thousand boring details at once.
Point an agent at work that suits a person and you get a mediocre agent. Point it at the never-human work and you get something nobody was doing well.
Wise Crack, Defined
Agentic Employee is our wise crack terminology for a
- non-human ai agent
- doing the work that a human employee used to do
- day in and day out
- as if it were an employee.
Your boss might be salivating about his future profits, but we know first hand that this isn't the work that we miss doing.
Yes, the economy might or might not crash. But no one asked our opinions about crashing an economy, or whether it is going to happen with, or without, our participation.
Measuring
Not "producing a report." Watching every change across a system, continuously, and noticing which ones matter. A person can audit a sample once a quarter. That is not a worse version of the same job — it is a different job, and the real one was never on offer.
Turning data into information
Most organisations are not short of data. They are short of anyone with the hours to turn it into something a decision can be made from. That conversion is tedious, endless, and exactly the sort of thing that gets deferred until it is worthless.
Reacting in real time
Something happens at 3am on a Sunday. A person finds out on Monday, if the logging was good. The gap between those two facts is where most of the damage lives, and closing it has never been a question of people trying harder.
And the inverse, which matters just as much
Agents are poor at the things people are good at. Judgement in an unfamiliar situation. Knowing when a rule should not apply. Reading a room. Deciding what is worth doing at all.
A perfect "agentic employee" is not a person-shaped worker who costs less. It is a colleague with a genuinely different shape — brilliant at three things, unreliable at the things you would trust a junior with.
What this means for the "replacing people" question
Listen to this excerpt for what Mathew Prince, Cloudflare CEO explains about his own company's layoffs due to Cloudflare OS rollout.
The short version is that it is not all bad news. There's a real upside here, too.
No one - least of all, some inconsequential bootcamp - could predict what is about to happen. But we're going to find out. We are pioneers.