Microsoft Kills Its Feedback Tool After Employees Start Using AI to Write Each Other's Reviews
The News
Microsoft is shutting down Perspectives, the internal employee feedback platform it launched in 2018 to replace formal peer reviews with a lighter, growth-mindset approach. Employees have until 15 September to respond to active requests and export their history. The reason, according to former Microsoft employee and Jozra founder Merrill Fernando, is straightforward: "The last few years everyone just started using AI and Copilot to generate the feedback, so I think it mostly lost its meaning." Fernando said he would receive 20 to 30 requests every six months and spend entire weekends writing responses — until AI made the exercise frictionless and empty. Microsoft is not abandoning peer feedback but is shifting toward regular, unstructured conversations between managers and employees, alongside a simplified five-category performance rating model.
Why It Matters
This story matters far beyond Redmond. Every TA and HR leader running structured feedback processes — 360 reviews, peer assessments, hiring manager evaluations of candidates — needs to ask the same question: are people still writing these themselves? If the answer is "probably not," the process is producing synthetic data dressed up as insight. The uncomfortable truth Microsoft just admitted is that formal feedback tools create formal behaviours. When compliance replaces candour, and when AI removes the friction that at least forced people to think, the output becomes noise. This applies directly to structured interview feedback forms, candidate scorecards, and reference checks — all of which are susceptible to the same hollowing-out.
Taking Action
All action is situated in your corporate context. These are starting points, not prescriptions — choose what fits.
Connect
Who to talk with in your enterprise
Your head of people operations or HRIS lead — audit your feedback processes and tools. Are you collecting responses because they produce useful signals, or because the system requires them? Your L&D or manager development lead — if you're shifting from formal platforms to manager conversations, do your managers actually know how to give useful, real-time feedback?
Inspect
What to take a closer look at
Pick your last 20 completed interview scorecards or peer feedback submissions. Look for patterns: identical phrasing across different reviewers, formulaic structures, or language that reads like it came from a prompt. If you find them, the data feeding your hiring and performance decisions may already be compromised.
Reflect
What to think about more deeply
If AI-generated feedback is indistinguishable from human-written feedback, the problem isn't the AI. It's that the process was already producing generic, risk-averse, going-through-the-motions responses. So what was the tool actually measuring?