The Always-On Performance Review

Continuous reviews with AI are finally possible.

👋 I‘m Andrew, the CEO of Effy AI, an AI-powered performance management platform. Let’s connect on LinkedIn to continue this conversation!

---

For most of my career, “the performance review” has meant the same thing it meant 30 years ago. A calendar event. Once a year, maybe quarterly if you’re modern. The manager writes a few paragraphs. HR runs calibration. Promotions get decided. And then everyone forgets about it for another nine months.

That model is breaking.

Not because someone published a new framework. Not because a consulting firm declared the annual cycle dead. But because the underlying technology that made the annual cycle the only viable option has finally changed.

The performance review is becoming a living document. A continuous stream that updates every week, every one-on-one, every project shipped. Always current. Always honest. Never a frantic blank-page exercise the night before the deadline.

For the first time, this is actually possible. And AI is the reason.

I want to share what I’m seeing inside the HR teams already running performance management this way. Because I think it’s the most important shift in people management in twenty years, and most companies haven’t even noticed it’s happening yet.

The shift is already happening (just not evenly)

Most HR leaders I talk to aren’t asking “should we use AI for people management?” They’re asking how to catch up to what’s already happening on their own teams.

Managers are already using AI. Just not in a coordinated, sanctioned way.

The more digitally-fluent managers are pasting meeting notes into ChatGPT to summarize a quarter of work. They’re drafting performance reviews with Claude. They’re asking AI to help them prep for difficult conversations with underperformers. They’re using AI to make sense of 360 feedback before they deliver it.

But it’s fragmented. It’s the top 10-20% of managers, doing it on their own, with no shared standards and no organizational visibility. The rest of the team is still drowning in Google Docs.

So HR leaders are coming to us with a different question now. Not “can you replace our review tool?” but “can you help us bring AI into people management in a way that actually scales across the org?”

This is the conversation I’m having more than any other. And it’s making me rethink what Effy is even for.

The foundation is data, and the foundation is finally arriving

Every conversation about AI eventually comes back to data. And until recently, this was the wall most HR teams hit.

You can’t have AI summarize a quarter of someone’s performance if there’s nothing written down about that quarter. You can’t have AI prep a manager for a one-on-one if the previous one-on-ones happened in someone’s head.

But that’s changing fast. And the biggest shift I’m seeing isn’t a software shift. It’s a behavior shift.

Managers are using meeting note-takers. They’re documenting one-on-ones. They’re writing things down they used to keep in their heads, because AI tools work better when there’s context to feed them.

The unintended consequence is that the documented context of day-to-day people management has exploded. Five years ago, the only written record of an employee’s quarter was their performance review. Today, in companies that have adopted the new tools, there are notes from every one-on-one, transcripts from every project meeting, and a steady trail of written feedback in Slack and Notion.

This is the substrate that makes AI people management actually possible. Without it, you’re just bolting GPT onto a 1-5 rating scale and calling it AI.

What our pilots actually look like

I want to share what we’re testing with our pilot customers right now, because I think it gives a clearer picture of where this is going than any abstract take could.

These aren’t hypotheticals. These are running in production with HR teams who wanted to be early.

Pilot one: AI-drafted performance reviews

This one is the most obvious. For companies still running quarterly reviews, the tax on managers is brutal. A manager with seven direct reports can spend an entire weekend writing reviews. Most of them hate it. Some of them rush. Almost all of them deliver something that took 40-60 minutes per person and reads like it took 10.

We’re using AI to do the first draft. The system pulls together everything available from the previous 3-6 months. Project work, peer feedback, manager notes from one-on-ones, OKR progress, whatever the company has connected.

The output is a full draft review. The manager doesn’t write from a blank page anymore. They review the draft, correct it, add the things only they would know, and submit.

The time savings are real. We’re seeing managers go from 40-60 minutes per review to 5-10 minutes. But the bigger win is consistency. The new manager who’s writing reviews for the first time now has the same starting point as the senior manager who’s done a hundred of them.

Pilot two: AI handles calibration and the final rating

This one is more controversial, and I think it’s the most interesting one we’re testing.

In most companies, the painful part of reviews isn’t writing them. It’s calibration. The hours of meetings where leaders sit in a room arguing whether someone is a 3 or a 4. The bias. The political dynamics. The pressure to round toward the middle so nobody has to defend an outlier rating.

What we’re testing is different. Managers and peers describe what the person did. They give context, examples, and observations. But they don’t pick the rating.

The AI does.

Same prompt, same criteria, same calibration logic applied across the entire org. No fatigue. No bias toward people the manager likes. No avoiding awkward conversations by rounding to a 3.

This is uncomfortable for a lot of HR leaders to talk about, but I’ll be honest: I think it’s where this is going. Humans are demonstrably bad at consistent rating. AI isn’t perfect, but it’s consistent. And consistency is the thing calibration has always been trying (and failing) to achieve.

Pilot three: AI-prepped one-on-ones

This one is my favorite, because it’s the most immediately useful for the day-to-day.

The system knows when the next one-on-one is scheduled. It knows what was agreed in the last one. It has access to project updates, feedback received, and meeting context from the period in between. And it understands the employee’s personality profile and recent reviews.

Before the meeting, the manager gets a short brief. Suggested topics for discussion. Things that have changed since last time. Items that were promised but haven’t been delivered. Potential coaching moments based on what’s happening in the person’s projects.

The manager doesn’t have to use any of it. But suddenly, walking into a one-on-one cold isn’t the default anymore.

For new managers especially, this is transformative. The biggest predictor of one-on-one quality used to be whether the manager had years of experience and strong intuition. Now, every manager has that scaffolding from day one.

Pilot four: AI-coached feedback delivery

This builds on pilot three. Inside the one-on-one prep, managers also get suggested feedback.

Drafted examples of how to deliver praise for something specific. Examples of how to surface a concern about a behavior pattern. Models of constructive feedback that a senior manager would give, adapted for the situation.

Again, the manager doesn’t have to use any of it. But it gives new managers a feedback library they didn’t have before. And it makes feedback delivery more consistent across the team.

The combination of pilots three and four is what excites me most. Because together, they don’t just save time. They actually raise the floor of what good people management looks like inside a company.

What this is really about

I’ll be honest about my motivation here.

I started Effy because I watched too many performance review processes feel like a compliance exercise instead of an actual conversation. I watched too many employees get a 3 they didn’t deserve. I watched too many new managers struggle through their first review cycle with no support.

The dream isn’t AI replacing managers. The dream is AI making the average manager as good as the best manager. And making feedback more honest, more consistent, and more frequent in the process.

That’s the real prize. Not efficiency. Not time savings. Just: better managers. At scale. Without needing to hire a leadership coach for every manager in your company.

But I’m also not going to pretend this stops at “assistant.” If I’m being honest with you, and with myself, the more we build these systems, the closer we get to something that genuinely could replace large portions of what a manager does.

Coaching? An AI that’s read every coaching book and observed thousands of one-on-ones can probably coach better than a stressed manager who hasn’t had time to read in a year.

Feedback delivery? An AI that doesn’t have ego, doesn’t have political pressure, and doesn’t have a bad day can probably deliver clearer, kinder feedback than a human under pressure.

Rating and calibration? Already covered.

Career development? Probably next.

So is the AI assistant just a stepping stone toward an AI manager?

I don’t know. Maybe I’m joking when I ask that. Or maybe I’m not.

What I do know is that the conversation about AI in people management isn’t really about tools anymore. It’s about what the role of the manager will actually be in five years. Whether managers will spend their time coaching humans on the things AI can’t do (relationships, judgment under ambiguity, hard calls), and let AI handle the rest. Or whether companies just flatten layers and let the AI run point.

I have my bets. But I’d rather hear yours.

An invitation

I’m publishing this because I want to be part of this conversation, not just an observer of it.

If you’re an HR leader who’s thinking about how AI fits into your people management stack, I want to hear from you.

What part of this excites you? What part makes you uncomfortable? What do you think we’re getting wrong?

Reply to this post. DM me on LinkedIn. Or drop a comment below. I read everything.

The next two years of people management are going to look very different from the last twenty. I’d rather figure it out with you than write essays about it alone.