Gartner says performance management will become less — and more — human. Mid-market calibration is still theatre.

In late October 2025, Gartner named four trends talent management leaders should prepare for in 2026. One of them lands squarely on CPOs who still run annual review theatre: performance management will become less — and more — human. The press release — “Gartner Identifies Four Trends Talent Management Leaders Should Prepare for in 2026” (29 Oct 2025) — is clear about the split. Automate the administrative load. Do not automate the act of managing people.

Gartner’s line, via analyst Guadagni: the future of performance management processes is automation, but the future of managing performance can’t be. Managers are already experimenting with AI for reviews; most still have no formal training on what “good use” looks like. That gap is where mid-market stacks either get sharper — or quietly make bias faster.

What Gartner is actually saying

Restating the performance-management thread in plain language:

1. Less human on the admin — drafting goals, summarising feedback, preparing review packs, and supporting calibration with evidence should stop eating manager weekends.
2. More human on the judgment — empathy, accountability, coaching conversations, and final calls stay with people. Tools prepare the room; they don’t own the decision.
3. Train the AI layer — approved tools, bias mitigation, and explicit good/bad uses. Experimentation without guardrails is not a strategy.
4. Fix underperformance mechanisms — Gartner also flags that most organisations are weak at improving low-productivity employees; reinvention of PIPs needs clear goals and hard timelines, not vague “needs improvement” labels.

None of that starts with buying another chatbot. It starts with a review system that already holds scorecards, a real cycle, and calibration that debates evidence instead of personalities.

The hurdle: calibration as theatre

For mid-market People leaders, the Gartner split exposes a familiar failure mode. Reviews are still once-a-year. Managers write narratives from memory. Calibration is a PowerPoint room where the loudest voice wins. AI, when it appears, drafts generic prose that nobody trusts — then leadership blames “the tool.”

• Episodic reviews — twelve months of work compressed into a week of forms, so AI summarises noise, not a living record.
• Scorecards without a shared bar — managers invent criteria; calibration argues taste instead of evidence.
• Calibration without a data layer — no goal progress, no feedback themes, no consistent deliverables lens — only advocacy.
• AI bolted onto chaos — drafts get faster while fairness and coaching stay stuck.

Gartner’s “less and more human” only works if the human half has something real to stand on. Theatre plus automation is still theatre.

How Revolut People Reviews operationalises this

The primary product surface for this hurdle is Reviews — scorecards, a real cycle, and calibration built for evidence, not theatre.

• Structured scorecards — managers rate against published criteria so “less human” admin is form-filling against a bar, not inventing the bar each cycle.
• Quarterly (or frequent) cycles — shorter loops give calibration longitudinal signal instead of a yearly surprise; underperformance conversations can start before a PIP becomes inevitable.
• Calibration on shared evidence — the room debates the same scorecard and goal trail, which is how human judgment stays accountable instead of political.
• Outcomes that stick — promo, pay, and development signals leave the room tied to the grade, so the “more human” half has consequences, not just conversation.

That is the stack Gartner’s split assumes. Method context lives in the playbook summary and Week 1 — Why annual reviews fail; Reviews is how teams run the cycle instead of performing it once a year. Pair it with live Goals and calibration stops being a negotiation of vibes.

A 30-day start

1. Publish one scorecard template (deliverables + behaviour) for two critical teams — kill free-text-only reviews for that cohort.
2. Move those teams onto a quarterly cycle with a fixed calibration date on the calendar before forms open.
3. Require every calibration pack to include goal progress + scorecard scores (no narrative-only advocacy).
4. Name approved AI uses for managers (summarise peer feedback; never invent ratings) and ban the rest until training exists.
5. Rewrite one underperformance path with clear goals and a predetermined timeline — test it on the next real case, not in a policy wiki.

Gartner’s 2026 read is not “replace managers with models.” It is automate the grind so managers can actually manage. Mid-market stacks that still treat calibration as theatre will automate the theatre — and call it progress.