BTOM Consultants: AI in Performance Management

Most managers are stretched thin, and gathering scattered information eats a huge share of their week. One-on-one notes sit in one app. Project updates land in email, then Slack, spread across three places by the time anyone needs them. Goals shift mid-cycle without much warning. By the time review season rolls around, a manager is often piecing together six months of context from memory, which is a rough way to write anything that holds up. This is one of the main reasons why everyone hates the performance review process – it is hard on managers, hard on employees, and too often not helpful to either.

This is where AI pulls real weight. It can organize approved notes or turn a messy pile of bullet points into something coherent that saves actual time. It can flag a question worth raising in an upcoming one-on-one, or sketch a rough development plan so a manager has a starting point when the page is blank. All of this is prep work, and good prep gives a manager’s judgment something solid to work from.

Manager engagement dropped from 30% in 2023 to 27% in 2024, while individual contributor engagement held flat at 18% (Source: Gallup, 2025). That’s part of why the thirty minutes a manager saves writing a review is worth something real. That time can go back into reviewing a project more closely, or finally scheduling the follow-up conversation that’s been sliding for two quarters running.

For someone new to the role, the payoff runs bigger. A first-time manager writing their first review usually has no real model for what a good one looks like, because nobody sat them down and showed them. Seeing organized language on a page gives them something to react to and push against. Some first-time manager training programs now use AI performance reviews as practice material, giving new managers a draft to critique before they write their own. That’s a legitimate use of the technology. The trouble starts the moment a manager mistakes a well-organized draft for evidence that the actual management work happened underneath it, and that mistake is an easy one to make.

A generated review can sound measured and fair even when the manager behind it barely understands what happened over six months. AI works only with what it’s fed. Thin notes, warning signs nobody addressed, expectations that were never stated clearly out loud: all of that limits what comes out the other end. The tool reorganizes an incomplete record into cleaner sentences. That’s the ceiling.

Take a line like “needs to demonstrate greater ownership across cross-functional initiatives.” Fine on the surface, until you ask which initiative, or what action was actually missing, or whether anyone told this person that was even expected of them before the review landed. Those questions only get answered through real conversations held across the year, the kind of ongoing employee feedback that never makes it into a single tidy paragraph.

This is a big part of why so many first-time managers lean hard on templates. A template puts distance between a manager and an uncomfortable conversation, and AI makes that distance feel more legitimate because the output sounds custom-built, even when it draws on the same generic phrasing every manager gets. A sentence a manager can’t explain in plain language, or back up with something specific, doesn’t belong in the review.

Two people can miss the exact same target for completely different reasons, and lumping them together is where a lot of reviews go wrong. One person might be missing a skill and need direct coaching. A teammate hitting the same wall could understand the job fine and still lack the time or authority to execute it the way it’s written on paper. Another person could be dealing with something hard personally and just need a stretch of patience, and there’s always the person who’s gotten support and clear expectations already and simply hasn’t met them.

Telling those situations apart is human judgment in employee performance, and it only comes from being present with people over time. It shows up in one-on-ones done well. It also shapes the follow-up question a manager asks after really listening, and the development plan built around one specific person. Manager coaching is supposed to build exactly that: an instinct for which situation is which.

First-time manager training has some catching up to do here. In the UK, 82% of managers stepping into their first management role received no formal leadership or management training at all, based on research covering more than 4,500 workers and managers (Source: Chartered Management Institute, 2023). A manager who’s never learned to deliver feedback might start treating AI like a stand-in for that skill, even though the tool can only shape wording. Picking the right moment, reading how someone reacts in the room, deciding whether to push or hold the line: those skills come into play only once the actual conversation is happening.

For employees, the conversation is the part that counts, and the written review is a record of what the manager believed walking in. Tone, timing, which questions got asked, the silences that stretch a beat too long, how someone reacts when the feedback actually stings. A better draft can’t fix any of that.

Listening during that conversation can change the outcome while it’s happening. The employee might mention that two leaders gave them conflicting instructions and nobody caught it. They could describe a workflow problem nobody above them has noticed yet. Sometimes they simply confirm the expectation was clear and admit, flatly, that they didn’t meet it. A good manager stays open to whatever comes up without softening the standard they walked in with, and holding both of those at once is harder than it sounds.

What happens after the conversation ends matters just as much, maybe more. Skip the follow-through and the whole thing turns into a yearly ritual, same script, slightly different words each time around. Follow-through looks like checking in two weeks later to see what changed, noticing that someone has gone quiet in meetings because they felt dismissed, and deciding whether slow but real improvement deserves more patience or whether missed commitments call for something firmer. It only happens if the manager stays in it well after the meeting’s over.

AI’s usefulness narrows fast once you’re at that stage. Before a review, an approved tool can organize non-sensitive notes and check whether each piece of feedback is backed by a real example. The manager still has to verify every claim and rewrite the draft into language they’d actually say out loud, because a review only lands if the person delivering it sounds like themselves. That’s the honest answer to how managers should use AI for performance reviews without quietly handing the whole process over to it.

During the meeting, the manager’s full attention matters most. Reading a generated script word for word can flatten an already tense conversation fast, so the manager needs to listen closely enough to drop the script the second something real comes up. Once the meeting wraps, AI can help summarize agreed next steps within whatever privacy rules apply, and the manager still owns whether that record is accurate and whether the follow-through happens weeks later.

The pressure to sort this out keeps building. Plenty of workplaces are asking employees to start supervising AI agents at the same time many first-time managers are still learning how to lead people. Leaders already expect their teams to be training and managing AI agents within five years, and 51% of managers think AI training or upskilling will become a core part of their team’s job in that same window (Source: Microsoft, 2025). That’s good reason to work out now what management actually requires, and where AI performance reviews fit inside that picture.

AI amplifies the habits a manager already brings to the job, including the habit of avoiding hard conversations. A manager who listens well, sets clear expectations, and follows through will feel the time savings right away.