Welcome to Your Token Improvement Plan
Every knowledge worker knows the three most ominous letters in corporate life: the PIP. The Performance Improvement Plan.
I’ve written three and been on zero. But a strange thing is happening in this particular moment in business:
A growing number of my friends in the tech space are joking that they’re on a “Token Improvement Plan.”
The new yardstick
For most of working history, you were measured against the people around you. Your number against the team's number. Your campaign against the last campaign. Your year against your previous year. The comparisons were human, and unless you’re in sales, mercifully slow.
That's no longer the only contest.
Increasingly, the standard isn't your coworker, the industry average, or even your own track record. It's a new benchmark:
What a reasonably competent person armed with AI could have produced in the same amount of time.
How effectively you're using AI compared to your peers.
The performance review used to ask, "How did you do compared to expectations?" Now, sitting underneath the polite phrasing, there's a second question the manager may not even say out loud: "How did you do compared to what was available to you?"
Three years ago, a manager looked at a finished report and asked whether it was any good. Today a certain kind of manager looks at the same report and asks a different set of questions:
Why did this take three days?
Why did this require three people?
Why wasn't AI anywhere near it?
The question is no longer "Can AI do this?" That argument is over. The question is now far more uncomfortable: "Why didn’t AI get this done in 80% less time than you did?"
What the dashboards know
If that sounds paranoid, I have unwelcome news. In a fair number of technology companies, this isn't a vibe. It's a metric.
Somewhere in those companies is a dashboard most employees never see. It tracks prompt usage. Token consumption. AI-assisted output. Experimentation with new tools. Who's leaning in, who's dabbling, and who hasn't logged a single prompt since the all-hands where leadership swore this was "the future of how we work."
I'll let you guess which group gets the benefit of the doubt at review time.
Which brings us to the joke now making the rounds in tech circles:
"Tokens are becoming more expensive than salaries."
It's funny because it's almost true, and it's almost true in a way that says something about us.
For a hundred years, payroll was the most scrutinized line on the ledger. Companies counted heads the way misers count coins. Every hire was a debate. Every raise, a negotiation. Headcount was sacred, watched, defended, agonized over.
If they’re not already, those same companies should watch their AI usage with the identical intensity they once reserved for salaries. The finance team that spent a decade hunting down a rogue $14 software subscription is now staring at a six-figure model bill that didn’t go through any type of formal Finance or Legal approval.
The joke lands because companies have started treating AI usage as a proxy for effort
Fair or unfair, a team consuming tokens looks busy, curious, and engaged.
A team consuming none looks like it missed the memo.
We spent decades measuring labor.
Now we're starting to measure leverage.
The irony nobody ordered
Here's the part that should make you smile and squirm at the same time.
For a generation, the great workplace fear was being replaced by someone cheaper. Cheaper labor across the ocean. Cheaper labor down the hall. The anxiety was always about the low bidder.
The new anxiety is more intimate. It isn't that someone cheaper will take your job. It's that the person in the next chair — same salary, same title, same coffee order — has figured out how to wield these tools and you haven't. They're not cheaper than you. They're faster than you. And to a manager, faster reads as better, whether or not that's entirely fair.
Nobody warned us the threat would have the same job description.
Two employees, same job
Let me make this concrete.
Two employees get the same assignment. One starts with a blank page. The other starts with AI.
The first spends two days researching, outlining, drafting, revising, and second-guessing. Honest work. Real effort. The second spends twenty minutes getting a rough first pass from AI, then another ninety applying judgment, experience, and taste. The work comes back surprisingly similar. The timeline doesn't.
Or take two managers preparing for an important meeting. One spends half a day digging through emails, notes, and Slack threads trying to piece together what happened over the last six months. The other asks AI to summarize the history, identify open issues, surface risks, and suggest discussion points before lunch.
Both show up prepared.
One simply got there faster.
Here's the trap.
The manager reviewing the work doesn't see two different philosophies of work. They don't see craftsmanship versus convenience. They don't see old school versus new school.
They see a productivity gap.
One person consistently produces more output in less time. One turnaround is shorter. One number is bigger.
And managers, bless them, have always been very good at noticing which number is bigger.
You may find that unfair.
You may be right.
The yardstick does not care.
Now Turn It Around
So far, this reads like a warning to employees.
It's actually more interesting for business owners.
The obvious conclusion is that everyone needs more AI. I don't think that's right. Giving a team new tools without helping them understand where those tools fit usually just creates more noise.
The real lesson is much older than software.
Every generation gets a new lever.
The contractor who bought the first pneumatic nail gun didn't become a better carpenter overnight. The office that put a computer on the front desk didn't suddenly employ smarter people. In both cases, the same people became more productive because they found a better way to apply their time.
That's what makes this moment so interesting.
AI isn't changing the pattern. It's just the newest lever.
The people who learn where it helps become more productive. The people who ignore it aren't suddenly replaced. They simply start to look a little slower, a little more expensive, and a little less competitive than the people who figured out how to use it.
That's why I find this story oddly optimistic.
Nobody is being asked to become a programmer. Nobody is being asked to reinvent themselves. They're being asked to do what good operators have always done: notice when a new lever shows up and decide whether to pick it up before their competitor does.
The line worth remembering
If you take one thing from this, take this.
The fear was always that we'd be measured against the machines. That turned out to be the wrong fear.