Year-end performance reviews: HBR says AI is polishing the wrong thing. Fix the stack underneath.

Q4 is when most People teams find out whether their stack actually works. Review forms open, managers dig through a year of Slack threads and spreadsheets for evidence, and someone in finance asks for salary changes by a date that was fixed months ago. It is also when a lot of UK and Nordic companies start shopping for a new HRIS or performance management system, usually halfway through a cycle that is already painful.

Harvard Business Review put its finger on the problem in May 2026. In “Gen AI Could Fix Performance Reviews—or Make Them Even Worse” (Chrysanthos Dellarocas, HBR, 15 May 2026), the argument is simple: companies are pouring AI into reviews, but mostly to write the same old review faster.

What HBR is actually saying

In plain language:

1. Big employers are already automating year-end reviews. HBR points to Citi’s Performance Assist drafting evaluations from data across the organisation, JPMorgan’s LLM Suite supporting year-end review writing, and BCG’s internal assistant, which reportedly cuts review-writing time by 40%.
2. Most of it is speed, not substance. So far, most organisations use these tools to produce more polished versions of traditional narrative reviews, more quickly.
3. Polish can hide weak evidence. Smoother language can mask thin evidence, subjective judgments and incomplete observations. A well-written review is not the same as a fair one.
4. The real opportunity is evidence. The better use is to surface what the person actually did: the decisions they made, the problems they solved and the influence they had.

That last point is the one buyers should hold onto. AI can only surface evidence that exists somewhere. If goals live in a slide deck, feedback lives in someone’s head and 1:1 notes live in a personal doc, no model will find them in December.

Why year-end exposes the stack

For mid-market teams, the year-end crunch usually looks like this:

• Goals set in January, untouched since. Progress was never tracked, so the review grades memory, not delivery.
• Feedback that never got written down. The last six weeks dominate because they are the only weeks anyone remembers.
• 1:1s that restart from zero. No shared notes, so coaching does not build up into evidence.
• Core HR in one tool, performance in another, pay in a spreadsheet. Headcount, leave and reporting lines sit in the HRIS; ratings sit in forms; salary changes get typed into payroll by hand after calibration.
• Calibration as advocacy. Without a shared evidence trail, the loudest manager wins.

Bolting an AI writer onto that setup gives you HBR’s worst case: faster, smoother reviews built on the same gaps.

Includes a price for your headcount + 3 fixes for your review cycle.

A buyer’s checklist: HRIS and performance management for UK teams

If you are comparing systems before year-end, these are the questions that separate a review tool from a performance system:

1. Do goals update during the year? Look for company → team → individual goals with live progress, ideally pulled from the tools teams already use, not typed in at review time.
2. Are reviews structured around evidence? Scorecards against published criteria, 360 input and calibration views that show goal progress next to ratings.
3. Is feedback continuous? Feedback and 1:1 notes should sit against the person and their goals all year, so the review is a summary, not an archaeology project.
4. Does the review lead somewhere? Promotions, development plans, PIPs and salary changes should come out of the cycle as tracked outcomes, not a separate spreadsheet.
5. Is core HR in the same place? Employee records, onboarding documents, org chart and time off. If they live in a different system, you are reconciling two sources of truth every cycle.
6. How does it handle payroll data? Most performance-led platforms don’t run payroll themselves. Ask how new hires, salary changes and leavers get to your payroll provider without retyping.
7. Will it fit local rules? UK, Nordic and EEA leave policies, workflows and entities differ. Check it can be configured per country without forking the performance framework.

Where Revolut People fits

Revolut People started as the in-house system Revolut uses to run performance across its own workforce, and it is built performance-first. Mapped to the checklist:

• Goals: company, team and individual goals with real-time tracking, plus integrations with Jira, Linear, Notion and ClickUp, so progress builds up during the year instead of being reconstructed in December.
• Reviews and calibration: structured scorecards and review templates that put evidence ahead of opinion, 360 feedback, calibration, and access to previous cycles for a complete view of each person’s goal progress.
• Continuous feedback and 1:1s: feedback and AI-assisted 1:1 summaries that highlight talking points, so the evidence HBR says matters gets captured as it happens.
• Outcomes: post-review plans for promotions, performance improvement plans and salary adjustments, so every person gets a clear outcome and every manager owns the next step.
• Core HR in the same platform: employee records and self-service profiles, offers and onboarding documents, org charts, and time off with configurable leave policies and one-click approvals.
• Payroll and existing HRIS: Revolut People doesn’t run payroll; it syncs employee and salary changes with the payroll provider you already use, and integrates with HRIS tools such as BambooHR, HiBob, Personio and Deel if you keep one.

That is the stack HBR’s argument assumes: one place where goals, feedback, 1:1s and reviews build up all year, next to the HR record. AI then helps you find what someone actually did, rather than making a thin review sound good. Method context is in the playbook summary, Why annual reviews fail and Gartner on calibration.

If you’re deciding before year-end

1. Run this cycle with what you have. Switching tools mid-review rarely helps. Write down every place your managers went looking for evidence this time.
2. Pick two pain points to fix first. Usually goals that go stale and feedback that never gets written down.
3. Decide the stack before next year’s goals are set. Goals set in a new system in January give you twelve months of evidence by the next year-end.
4. Agree AI rules up front. Use AI to summarise and surface evidence; don’t let it invent ratings or polish narratives that have nothing behind them.
5. Close the loop to pay. Map how review outcomes reach salary changes and payroll before calibration week, not after.

Free tools: estimate what your review cycle costs with the performance review cost calculator, or see if your HRIS is holding performance back with the HRIS switch checker.

FAQ

Do I need a separate HRIS and performance management system?

Not necessarily. Running core HR in one tool and reviews in spreadsheets or a second tool splits the evidence a year-end review needs. Revolut People combines performance (goals, reviews, calibration, feedback, 1:1s) with core HR (records, onboarding, time off) in one platform, and integrates with BambooHR, HiBob, Personio and Deel if you’d rather keep your HRIS.

Does Revolut People run payroll?

No. It keeps employee and compensation data (new hires, salary changes, leavers) and syncs it with the payroll provider you already use.

Is Revolut People available in the UK?

Yes. It’s available to businesses with a registered entity in the UK, the EEA or the US.

Should we switch before or after this year’s review cycle?

Run this cycle with what you have, but decide before next year’s goals are set, so a full year of goals, feedback and 1:1 notes builds up in one place before the next review.