You're probably using AI for performance reviews all wrong. I was too. Here's the painful story of how I learned to do it right.
It was 2022, and the AI hype was deafening. Every other startup in the Valley was "making use of AI" to do something or other. We, at RemoteTeam, were no different. We were already a successful company, on our way to being acquired by Gusto, but I’m always looking for an edge. The idea of using AI to make our performance reviews better, faster, and fairer? It was intoxicating.
I thought it would be easy. We’d just plug in some AI tool, feed it our employee data, and it would spit out perfectly objective, insightful reviews. No more manager bias. No more awkward conversations. Just pure, unadulterated data-driven feedback. What could go wrong?
As it turns out, everything.
The First Attempt: A Glorious Failure
I remember the day we rolled out our first AI-powered performance review system. I was so proud. I’d spent a small fortune on a fancy new platform that promised to revolutionize our HR processes. It had a slick dashboard, all the right buzzwords, and a sales team that could sell ice to an Eskimo.
I gathered the whole team for an all-hands meeting. I walked them through the new system, my voice brimming with excitement. "This," I declared, "is the future of work!"
Two weeks later, the future of work was a dumpster fire.
The reviews were a disaster. They were generic, impersonal, and completely missed the mark. The AI had no context. It didn’t understand the nuances of our projects. It didn’t know that Sarah in engineering had single-handedly saved a critical launch with a brilliant piece of code, or that Tom in marketing had been battling a personal crisis while still managing to deliver a killer campaign.
The AI just saw numbers. Lines of code. Tickets closed. Sales quotas met. It was a soulless, robotic assessment of human beings who were pouring their hearts and souls into their work.
The team was demoralized. I had managers coming to me, their faces a mixture of frustration and despair. "I can't use this," one of them told me, holding up a printout of a review. "It's telling me my top performer is a slacker because she didn't close as many tickets as the new guy. It doesn't know she spent the last month mentoring him!"
I had to admit it. I’d screwed up. I’d been so seduced by the promise of AI that I’d forgotten the most important ingredient in any performance review: humanity.
The Second Attempt: A Glimmer of Hope
I almost gave up on AI for performance reviews altogether. I was ready to go back to the old way of doing things, with all its flaws and biases. But then, something happened.
I was having coffee with a friend, another founder who’d been through a similar experience. He listened patiently as I vented my frustrations. Then he said something that changed everything.
"You're thinking about it all wrong," he said. "You're trying to replace your managers with AI. You should be trying to empower them with it."
It was a simple shift in perspective, but it was a profound one. I’d been so focused on automation that I’d missed the opportunity for augmentation.
I went back to the drawing board. I threw out the expensive, overhyped platform. I started talking to my managers. I asked them what they needed. What were their biggest pain points? What would actually help them do their jobs better?
The answer was surprisingly simple. They didn’t need an AI to write their reviews for them. They needed an AI to help them gather the data, identify the patterns, and surface the insights they were missing.
So that’s what we built.
The Third Attempt: The Machine
We didn’t buy a new tool. We built our own. We called it "The Machine," and it was the complete opposite of our first attempt. It wasn’t a black box. It was a transparent, collaborative platform that put our managers in the driver’s seat.
Here’s how it worked:
- Data Aggregation: The Machine pulled in data from all the tools we were already using: Slack, Jira, GitHub, Google Docs, you name it. It didn’t just look at the numbers. It looked at the conversations, the code reviews, the project updates. It built a holistic picture of each employee’s contributions.
- Pattern Recognition: The Machine didn’t make judgments. It surfaced patterns. It would highlight things like, "This employee consistently receives positive feedback from their peers in code reviews," or "This employee has been a key contributor to three major projects this quarter."
- Manager Empowerment: The Machine didn’t write the reviews. It provided the raw materials. It gave our managers a dashboard with all the data, all the patterns, all the insights. It was up to them to interpret the data, to add their own context, and to have a real, human conversation with their team members.
It was a big deal.
Our managers loved it. They felt empowered, not replaced. They were having richer, more meaningful conversations with their teams. And our employees? They finally felt seen. They were getting feedback that was specific, actionable, and fair.
We even open-sourced a part of our internal tool. You can check it out on our company blog.
The Results: By the Numbers
I’m a numbers guy, so let me give you some numbers. After we implemented The Machine:
- Employee satisfaction with our performance review process jumped by 42%.
- Time spent by managers on writing reviews decreased by 60%.
- We saw a 15% increase in employee retention in the first year.
But the numbers only tell part of the story. The real impact was on our culture. We went from a culture of fear and anxiety around performance reviews to a culture of growth and development. We were finally living up to our values.
The Takeaway: Don't Replace, Empower
Look, I get it. The allure of AI is strong. It’s tempting to think you can just automate away all your problems. But when it comes to people, there are no shortcuts.
My advice to any founder or leader who is thinking about using AI for performance reviews is this: don’t try to replace your managers. Empower them. Give them the tools they need to be better leaders. Use AI to augment their intelligence, not to replace their humanity.
And if you're interested in learning more about how we build our teams, you can read my post on how I became the number one angel investor.
It’s a lesson I learned the hard way, but it’s a lesson I’ll never forget. And it’s a lesson that has made all the difference.
Frequently Asked Questions
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.