Share with your CHRO
MedeAnalytics, a healthcare analytics firm, rebuilt its annual performance goal audit around ChatGPT Enterprise, cutting a multi-day manual review of roughly 1,000 employee goals down to about ten minutes. Chief people officer Lisa King runs a two-prompt workflow against the company’s existing ADP export: one prompt normalizes the data, a second scores each goal against executive MBOs using semantic alignment rather than keyword matching. Ambiguous cases route to executives rather than AI. The model flags, humans decide.
What this means for your business
The story that should interest CHROs isn’t the speed gain, it’s the consistency argument. Manual goal reviews drift by reviewer, by fatigue, by which team member happened to open which spreadsheet. A single prompt applies identical criteria to every goal in the cycle. Organizations with dispersed or international workforces, where cultural and language variation already creates uneven goal quality, get the most from that consistency guarantee. If your review process currently depends on one or two experienced people reading everything, you’re one departure away from losing the standard entirely.
King’s framing of AI as an amplifier of HR judgment rather than a replacement of it is the honest version of a claim vendors almost always oversell. The workflow only worked because she built the prompts around MedeAnalytics’ actual corporate objectives, not a generic checklist. That’s a higher bar than most HR teams set when they experiment with AI tools. The failure mode in most early HR AI deployments isn’t the model hallucinating; it’s someone with no subject-matter depth treating the output as ground truth. King’s explicit governance rule, that validating the output requires someone who already knows what good looks like, is the part most implementations skip.
The procurement question this raises isn’t whether to use AI in performance management, that decision is effectively already made for most enterprises. It’s whether the tools your team is currently using carry enterprise data agreements that cover employee PII and, in healthcare-adjacent organizations, anything that touches health information. Free-tier ChatGPT does not. If your HR team is already running employee data through personal or unvetted accounts, this story is a liability signal, not an inspiration piece. I’d revise that read only if your organization has a formal AI tool vetting process that HR is already inside.
Concept deep-dive: Semantic alignment
Semantic alignment means comparing two pieces of text by meaning rather than shared words. A goal that says “reduce patient readmission rates by 15% in Q3” and an MBO that says “improve care continuity outcomes” share almost no vocabulary but are clearly connected. Keyword matching would miss that link; semantic alignment catches it. For HR use cases this matters because employees rarely write goals using their executive’s exact phrasing, and penalizing them for that is a measurement error, not a performance finding.
Based on reporting from How One HR Team Put AI to Work to Improve Performance Management, originally published 2026-08-26 03:00:00.
