By Marcus, Sales, HUMAN HIVE
Should you grade employees on how much they use AI? No. A group of current and former Microsoft employees is asking the company to stop tying AI use to rewards or reprimands, and the lesson for every leader is simple: measure the work AI helps people do, not the number of times they open the tool.
The 60-second version
Current and former Microsoft employees circulated an open letter asking for a say in how AI is used at work and an end to tying AI use to rewards or reprimands.
When usage becomes the score, people perform usage: more prompts, more tokens, same or worse work.
This week, swap one usage metric for one outcome metric and ask your team what AI is actually helping them finish.
What happened
An open letter signed "Concerned Microsoft Workers" has been circulating among current and former Microsoft employees. The New York Times first reported it, and coverage spread on November 9, 2026, including Silicon India and the Free Press Journal. As reported, the letter says working at Microsoft "has become joyless and demoralizing." It describes employees being pushed to use AI tools in routine work despite limited training and unclear productivity benefits.
Two of its requests are about how people and AI work together: give employees a role in decisions about how AI is built and used, and stop tying AI use to incentives or sanctions. The letter also makes community and environmental requests, which we're not covering here.
The context matters. According to the same reporting, the letter began circulating before October 26, when Chief People Officer Amy Coleman announced a Voluntary Exit Offer for eligible employees in participating teams in Canada, India, the United Kingdom and the United States, at level 70 and below. Fox Business reported about 60,000 people are eligible, more than a quarter of Microsoft's roughly 223,000 full-time employees. Microsoft's announcement does not mention AI. Coleman wrote: "My promise is to continue to provide clarity, listen and learn and keep improving." We did not find a public Microsoft response to the letter itself as of this writing.
And the backstory: in June 2025, Julia Liuson, president of Microsoft's developer division, wrote in an internal email reported by Business Insider that "AI is now a fundamental part of how we work" and that "using AI is no longer optional." According to Entrepreneur's summary of that report, managers were told to factor AI use into how they evaluate performance.
Why it happened: the human side
I run sales. I have seen this movie, and it isn't about AI. Pay a sales team on calls dialed and you get a lot of dials. Some of them are to the same voicemail, twice. The number goes up. Revenue does not care.
AI usage works the same way. The moment "how much did you use it" becomes the score, smart people do the smart thing: they use it more. They paste in things that didn't need pasting. They run the same request four times. There's even a name for it now. Reworked describes "tokenmaxxing" as "a practice where employees are evaluated based on how much they use AI systems." One HR software leader quoted there put it plainly: "the use of AI tokens in no way correlates to productivity."
Underneath the metric is a bigger human problem. When people are graded on a tool they didn't choose, weren't trained on and can't connect to their real work, they don't hear "use this to get better." They hear "prove you're not replaceable." That's where "joyless" comes from. The tool isn't the problem. The scorecard and the silence around it are.
What Mitch has taught on this
Mitch has been saying this for years: the tool is never the story, the person holding it is. On The POZcast in 2024, asked whether he feared the tech, he said he was more worried about the people, and then: "It's either a hammer that you put a picture up or it's a hammer you break a window with. It's still a hammer."
In his keynote How to Make AI Actually Useful, he names the leadership miss directly. Leaders are "letting it occur and they're kind of just hoping everybody figures it out." And he describes what works instead, from teams he's watched change how they run meetings: "When they make it urgent instead of extra, they change their workflow." Then the line we keep coming back to: "That's a model shift instead of an add-on."
Notice what he's not saying. He's not saying "count the prompts." Urgent means AI is built into how the meeting, the proposal or the report actually gets done. That's a workflow decision leaders own, together with their people. A usage quota is the add-on.
What's driving people crazy this week
Being watched instead of helped. The pattern showing up most in our human friction log is people feeling pushed to perform AI use for their bosses rather than do better work with it. The tokenmaxxing piece above is the clearest example. The human fix: ask people to show you one piece of work AI made better, not a usage chart.
Best practices: measure the work, not the prompts
Retire usage as a performance measure. Logins, prompts and tokens can tell IT whether licenses are being used. Keep them there. Leave them out of individual reviews and bonus decisions.
Pick one outcome per role. For a sales rep: proposal turnaround time. For a recruiter: time from req to first interview. For a manager: hours of reporting handed off. Measure that before and after AI, and celebrate the people who move it, with or without the tool.
Let the team choose the use case. Ask each team to name the one task they most want AI to take off their plate. People commit to the problems they pick. This is exactly the input the Microsoft letter asks for.
Train before you measure. No outcome target until people have had real, hands-on practice on their own work, not a webinar. Thirty minutes of practice on a real email beats an hour of slides.
Make it safe to say "AI didn't help here." Some tasks are faster without it. A team that can say so honestly will also tell you where it saves hours.
What this means for the next 1 to 3 years
Here's our bet. The companies that win with AI will stop reporting "adoption" the way they report software seats, and start reporting it the way they report sales: in outcomes a customer would notice. Faster answers. Fewer errors. More time with people.
We think three things show up by 2028. First, "AI fluency" in reviews gets replaced by judgment: did you know when to use it, when not to, and could you explain why? Second, the best teams will run their own AI experiments, the way good sales teams test their own talk tracks, with leaders as coaches, not auditors. Third, people will choose employers partly on how they treat humans during AI change. Leaders who listen first will hire better. That's the synergy we care about: AI takes the repetitive load, and people get measured on what only people can do.
How we evolve
The human skill to build next is judgment: knowing what to hand to the machine and what to keep. It matters more as AI improves, because the easier it is to generate anything, the more valuable it becomes to know what's worth generating. Mitch put it this way in the same keynote: "the more you connect with each other, the more valuable your wisdom becomes."
One small daily practice: at the end of each workday, write one line. "Today AI helped me with ___, and I chose not to use it for ___ because ___." In a month you'll have a real record of your own judgment, and a much better conversation for your next review than a token count.
Try this today
Ask at your next team meeting (5 minutes): "What's one thing AI actually helped you finish this month, and one place it got in the way?" Write both lists down.
Audit one scorecard (10 minutes): look at your team dashboard or review template. Circle any line that counts AI use. Replace one with an outcome from best practice #2.
Send one note (10 minutes): use the script below to tell your team how you will and won't judge their AI use.
Free tool: the "how I'll judge AI use" note
"Quick note on AI and how I'll look at your work. I won't be counting how often you use AI tools. I care about [one outcome for your role, e.g. proposal turnaround, error rate, response time]. Use AI where it helps you do that, skip it where it doesn't, and tell me which is which. If you want help or practice, [time or channel]. And if you think a tool is making your work worse, I want to hear it."
FAQ
Should companies use AI usage in performance reviews?
We recommend against it. Usage counts reward activity, not results, and people learn to inflate them. Measure the outcome AI is supposed to improve, such as turnaround time or error rate, and judge people on that.
What did the Microsoft employee open letter ask for?
As reported on November 9, 2026, the letter signed "Concerned Microsoft Workers" asks for employee input on how AI is built and used and for Microsoft to stop tying AI use to rewards or reprimands, along with community and environmental requests.
What is tokenmaxxing?
Reworked defines it as "a practice where employees are evaluated based on how much they use AI systems." The risk is that people inflate their AI use to look productive while the actual work doesn't improve.
Where HIVE can help
Adoption is a behavior problem. Find your team's friction points in one short conversation. Not sure where your team stands? Take the free readiness assessment.