The joke that wasn’t really a joke.
A former team member once made a joke that stayed with me.“ChatGPT wrote my review this year.” It landed with a laugh and we moved on. I really couldn’t challenge it because they were right. The vendor offered a new AI editor that polished up the dullest of rocks into thoughtful prose. And in that moment of easy acknowledgment, something worth my paying attention emerged. The truth was masked as a joke and signaled a need to address.
I have spent enough time in and around organizations to know that moment wasn’t an anomaly. The performance review, already one of the most dreaded and least trusted form factors in the workplace, has found a new way to lose the one thing that ever made it valuable: the human thinking behind it.
Here is what the loop looks like in practice. A team member, reasonably skeptical that anyone will read their self-review carefully, uses AI to write or heavily edit it. Their manager, stretched thin and not entirely sure what to say, uses AI to do the same. Two polished documents are produced. A meeting gets scheduled. Two people sit across from each other holding artifacts that neither of them fully authored, staring into the abyss rather than having a dialogue that matches the style of the written words. While AI
didn’t create the problem, it just gave everyone a comfortable place to hide from thoughtful and authentic communication.
The problem underneath the problem.
The performance review’s focus should have never been the document itself, rather the conversations the document anchors. A review should be the summary of a year of ongoing, continuous dialogue, not a substitute for one. When the ongoing dialogue never happened, the document has no foundation.
While AI can polish the language, it cannot manufacture the team member/manager relationship or the trust required to deliver feedback that lands and sticks.
The manager capability gap is what makes this particularly dangerous. Consider the combination that shows up more often than anyone wants to admit: a manager who doesn’t provide guidance; doesn’t set clear goals; doesn’t make time for real 1:1 topics beyond task assignment; doesn’t build a consistent feedback loop; gives vague feedback, in the rare instances they give it; and then loses patience when their people aren’t performing at expected pace without ever connecting the dots to recognize that the underperformance they are documenting is partly—or largely—a product of the very environment
they created.
Layer an AI-generated review on top of that manager and whatdo you get? A polished document that frames the team member’s performance in sophisticated developmental language but delivered in a conversation the manager who supposedly wrote the review isn’t equipped to have. The document
implies thoughtfulness. The live experience tells a different story. That leaves the team member who must reconcile those two things in real time, across a table, with someone who can’t back any of it up. What was prior just a bad review experience turns into trust-destroying experience with better spelling
and punctuation.
Why everyone sees it and nobody says it.
There remains an uncomfortable organizational dynamic thatallows this to persist. The team member knows the manager is struggling to manage. HR knows it and pushes on it to no avail. The skip level, the person with actual authority to act, often looks the other way since the manager is a strong coder they don’t want to backfill, or sometimes the skip level is uncomfortable with the conversation as they themselves don’t have the tools or skills to moderate it. So, they overlook it all as the path of least resistance and hold their breath hoping for the best.
Everyone who sees the problem is either powerless to fix it or motivated to ignore it. Into that environment arrives an AI performance review tool that makes the problem invisible on paper. The manager’s output now looks indistinguishable from someone who did the work. The skip level has even
less reason to look closely. HR has even less ammunition to make their point. And the team member is sitting with from a polished document wondering why nothing ever changes. The tool didn’t create the dysfunction, rather it gives everyone who was already looking away an even better reason to keep looking away. And the longer the loop runs, the harder it becomes to unwind.
How we got here.
Manager capability doesn’t fall short by accident. From my years of observation, I believe there are structural reasons it lands where it does, and they are worth naming honestly.
- The first is mismatched motivation. People management is frequently positioned as the path to more money and a bigger title, not as a fundamentally different job that requires a fundamentally different skill set and disposition. The result is a meaningful portion of managers who want the promotion but did not sign up for the accountability that comes with it. Accountability being supporting someone else’s growth and delivering hard feedback when necessary. It also means being present enough to know their people and their motivations, incentives,
superpowers, and gaps. - The second is structure. The player-coach model, particularly common in engineering and technical functions, puts managers on the hook for their own output while leading a team. Something always must give and it is almost always the people management part, because that is the slower feedback loop. The code ships or it doesn’t, which is easy to measure. A team member’s development is harder to measure and easier to defer.
- The third is training. Most manager evelopment programs are built around risk mitigation. Think: What not to say and how not to get the company sued; what to document and how, and when when to contact HR. Don’t get me wrong here; these things matter. That said, they do
not teach situational judgment and leadership: how to read a person’s cues and how to deliver a message that lands differently depending on who is receiving it. Better guidance is needed on how to hold a difficult conversation without going soft or being over-the-top. This doesn’t come in a brown bag session, rather practice and time investment. While some managers are just fortunate to have the secret sauce in this regard, so many don’t, and most organizations do
not invest in building it since it doesn’t scale as easily as a lunch and learn or pre-taped course and multiple choice quiz on anti-harassment.
When you layer AI-assisted performance tools on top of all three of these gaps, you do not fix them, only obscure them.
Running both tracks at once.
This is not an argument against AI in performance management. Used well, these tools can reduce administrative burden, prompt more consistent feedback structures, and help managers who are trying to communicate better find the right language. The question is whether theorganization is treating the tool as a supplement to human judgment or a substitute for it.
Rather that run from AI in this instance, I believe the best approach is to run both tracks simultaneously so AI adoption in performance management can move forward. It must move alongside real, sustained investment in manager capability. Not instead of it.
Here is what that looks like in practice:
- Fix the pipeline upstream. Manager selection needs to be treated as a distinct hiring decision, not a default promotion track. The best individual contributor is not always the right people
leader. Organizations that conflate the two will keep producing the same gap. Promotion criteria should include demonstrated people skills, not just assumed ones based on individual contributions and demonstrated technical skill. - Make people leadership a real performance metric. If managers are not held accountable for the development and engagement of their teams as explicitly as they are for output and delivery,
people management will always be the thing that gets deferred. Managers and skip levels need to be in the habit of asking their team members how things are going and support them based on their responses. - Invest in situational development, not only compliance training. There are tools emerging, some of them AI-powered,that offer situational coaching specific to a manager’s known gaps and not just the compliance and liability topics. They can help a manager build muscle in the practice of holding a hard conversation with a struggling team member. They can also strengthen the skill of knowing when to push and when to support. That is the kind of development building a compliance training never will scratch. Don't just buy the tech and hope it is adopted, but roll it out thoughtfully, track useage, and pulse managers on its utility.
- Reclaim the performance conversation as ahuman moment. While AI can help a manager structure their thinking, prompt completeness, and refine their language, it should not be the author. The review is the beginning of a live and real conversation, and the conversation
is where the value to the team member and the manager lives. If a manager cannot sit across from their team member and speak honestly to what is in that document, the document is not doing its job nor is the manager. In short, the organization has a people problem dressed up as a process problem.
The window is open. For now.
The joke I mentioned earlier was funny for about two seconds, then it quickly sunk in as true. The truth is that organizations are moving fast to adopt AI in performance management while not keeping at pace
with the underlying capability problem the tool is now camouflaging.
The longer that gap persists, the harder it becomes toclose. The artificiality of the tool starts to feel normal. The polished documents become the expectation. The conversational muscle underneath them
atrophies even further. Eventually you have an organization that has optimized its performance management process while quietly abandoning its performance and feedback culture.
Investing in manager capability is not a nice-to-have sitting below the line when budgets get tight, as it serves as the infrastructure that determines whether any of the other people investments an
organization makes land. AI tools can support that infrastructure, not replace it.
The window to get ahead of this is open since the tech is still emerging. The organizations that treat manager development as a parallel investment to AI adoption, not a deferred one, will build something the others won’t be able to catch up to easily: a culture where the performance conversations are real, the feedback means something, and the review is the beginning of a dialogue rather than the end of one.
Two AIs walking into a performance review should only be a punchline, not a standard and acceptable operating procedure.
Photo credit: AI