It's unsurprising that most legacy performance management systems are using generative AI to help write feedback faster. The worst version of this just expands on bullet points to fill the space quickly. But better versions will help people write more constructive and actionable feedback, tuned to the unique evaluation criteria of the company, function, and level. AI (not just generative AI) will improve this process. But generative feedback is just a band-aid on a legacy process that has passed its expiration date. AI and data from work tools enable a new approach to performance, where you can analyze the actual work itself and provide real-time guidance without the lengthy legacy cycle inbetween. Feedback can be streamlined, in the moment, and more quantifiable than ever, which also helps remove some of the inherent bias. Nobody loves spending days writing up reviews, digging through past work to find highlights to prove your worth, or churning through reams of feedback and 1:1/OKR history to understand performance. Those days are numbered. We're excited to explore a new approach with our initial customers, and will have more to share soon.
Here's a place I hope that AI fails. This is Lattice, a tool for managers, their teams, and peers to give each other feedback. They are pushing these little "Assist" buttons into every field. Press it, and it will re-write and expand what you've written. The problem is that the goal of feedback is not to hit an impressive word count. It's not to make sure the feedback is well written or well rounded or "covers your bases". The goal of feedback is to reflect on what it is that will actually help someone level up in their job and their career. By giving people a shortcut to get to "submit" faster, it also shortcuts that reflection porcess. And it will deliver feedback that is at best unopinionated and unhelpful, and at worst incorrect. AI will have many incredible use cases. I hope this is not one of them.