AI is doing the work. So what are we really measuring in human performance?

This article was first published in the latest edition of People Matters Perspectives.

For most of corporate history, performance was measured by effort, activity, and individual contribution. Show up. Deliver your tasks. Hit your targets. Wait for your annual review.

This model worked because work itself was predictable. Roles were clearly defined, skills evolved slowly, and productivity could be reasonably inferred from output. Performance management systems were built for stability – stable jobs, stable teams, stable expectations.

That definition of performance is now collapsing. And fast.

As artificial intelligence takes over repetitive, predictable, and rule-based work, the very idea of “doing your job well” is being rewritten. Employees are no longer measured by how much they do, but by how much value they create beyond what machines can automate.

Performance, in this new reality, is heavier, sharper, and far more exposed.

When AI does the easy work, humans inherit the hard part

Across industries, AI agents are now writing code, reviewing compliance, generating content, analysing risk, screening candidates, and automating decisions that once required entire teams. 

What remains for humans is not volume, but the real heavy lifting in the spaces machines cannot yet occupy: judgment, strategy, creativity, and accountability.

In practice, this means:

• Smaller teams carrying larger mandates

• Employees expected to operate at the level of “five people + AI”

• Productivity defined by leverage, not effort

This is not a future scenario. It is already happening.

At Meta, leadership has openly pushed teams to think ‘5X, not 5%’ when it comes to productivity gains from AI. As automation expands across engineering, risk, and compliance functions, remaining employees are expected to move faster, decide better, and deliver outsized impact.

The uncomfortable question for HR leaders is no longer theoretical: How do you measure performance when the job itself has fundamentally changed?

The quiet death of the annual review

Annual reviews were built for stable roles, predictable outputs, and linear careers. None of those conditions exist anymore.

In AI-shaped organisations:

• Goals change mid-quarter

• Work is project-based and cross-functional

• Impact is often indirect, collaborative, and long-term

Waiting twelve months to assess performance is not just inefficient — it is irrelevant today.

More concerning is how performance systems are being repurposed. At Meta, revised review processes have reportedly been used to quietly reduce headcount, with managers instructed to label a fixed percentage of employees as “below expectations.” 

In such environments, performance management stops being a development tool and becomes a risk management instrument.

When reviews are used to mask workforce reductions, trust erodes. Innovation slows. Psychological safety disappears. Employees shift from building to surviving.

Performance reviews should be a compass. Too often, they are becoming a cudgel.

From outputs to outcomes — and now to judgement

The first evolution in performance management was the shift from outputs to outcomes. Completing tasks mattered less than what changed because the work was done.

AI is now forcing a second, deeper shift,  from outcomes to judgement. When machines can deliver outcomes efficiently, human performance is increasingly defined by:

• Quality of decisions under uncertainty

• Ability to frame the right problems

• Ethical judgement and risk awareness

• Creative synthesis across domains

• Long-term thinking over short-term optimisation

This is why traditional productivity metrics — hours worked, tickets closed, lines of code written — are breaking down. They capture activity, not value. In some cases, they actively distort behaviour.

Digital surveillance tools that track keystrokes and screen time may promise visibility, but evidence consistently shows they increase stress, reduce trust, and fail to capture the complexity of modern work. Watching people work is not the same as understanding their contribution.

Hyundai’s humanoids and the future of human performance

Hyundai’s investment in humanoid robots offers a glimpse of what lies ahead. As robots take on physical, repetitive, and precision-driven tasks, human workers are repositioned as orchestrators, problem-solvers, and system thinkers.

In such environments, performance is no longer about execution alone. It is about:

• How well humans collaborate with intelligent machines

• How they intervene when systems fail or behave unpredictably

• How they improve processes rather than merely follow them

The same logic applies to white-collar work. AI agents may write most of the code, but humans remain accountable for architecture, intent, risk, and consequences. Performance becomes less visible — but far more consequential.

Continuous feedback in an AI-mediated workplace

As roles stretch and expectations intensify, delayed feedback becomes dangerous. Employees operating at high strategic altitude need real-time calibration, not retrospective judgement.

This is where continuous feedback and OKRs matter — not as HR upgrades, but as survival mechanisms. If used well, they:

• Keep effort aligned with evolving priorities

• Make trade-offs visible

• Surface misalignment early

• Shift conversations from blame to learning

But there is a caveat. As AI enters performance feedback itself, employees are increasingly suspicious. Many believe their reviews are partially, or entirely, AI-generated, citing generic language and lack of personal context.

When feedback loses its human touch, trust suffers. And without trust, even the most sophisticated performance system becomes theatre.

Careers, rewards, and the shrinking margin for error

AI-driven efficiency is compressing career ladders. Fewer roles sit in the middle. The distance between “high impact” and “non-essential” is shrinking. In this environment:

• Careers are shaped more by skills than titles

• Progression depends on adaptability, not tenure

• Rewards must recognise complexity, not just outcomes

Recognition systems that reward presence, politics, and performance theatre will fail in a world where real impact is quieter, faster, and increasingly mediated by AI. High performers quietly carrying disproportionate strategic weight risk burnout and disengagement if their contribution is not understood or acknowledged.

The future of rewards lies in recognising:

• Decision quality

• System-level impact

• Knowledge sharing and capability building

• Long-term value creation

Not everything that matters can be counted, but it can be observed, discussed, and rewarded.

The real challenge for HR leaders

The hardest part of redefining performance is not technology. It is courage. Most organisations already have access to AI tools, analytics platforms, and performance software. What they struggle with is letting go of the comfort provided by familiar, but increasingly misleading, metrics.

Redefining performance in an AI-powered world requires HR leaders to make a series of uncomfortable shifts.

First, it means letting go of metrics that reward activity over impact. Measures built around hours, task completion, visibility, or utilisation feel objective, but they often say very little about value created. As AI absorbs execution-heavy work, these metrics become not just outdated, but actively distortive, rewarding busyness while missing real contribution.

Second, it requires resisting the temptation to use performance systems primarily as cost-control mechanisms. When performance management becomes a proxy for workforce reduction, ranking, or forced differentiation, trust erodes. Today, roles are expanding rather than narrowing, employees need clarity on how they add value, not anxiety about being measured out of relevance.

Third, they must invest in managerial judgement, not just platforms. No system can fully capture strategic thinking, ethical decision-making, or cross-functional influence. Managers need to be developed as evaluators of context, not just administrators of scores. This means coaching leaders to have better performance conversations, about choices made, risks taken, trade-offs navigated, and outcomes owned.

Fourth, performance systems must be designed to balance accountability with humanity. As employees increasingly do the work of multiple roles, amplified by AI agents, burnout risk rises. High performance cannot be defined solely by output or resilience. It must also account for sustainability, learning, and long-term contribution.

At its core, performance management in the age of AI is no longer about control. It is about clarity, trust, and intelligent enablement.

1. Clarity on what outcomes truly matter.

2. Trust in human judgement where algorithms fall short.

3. Enablement that allows people to focus on the work only humans can do.

A 100% goal-achievement score is now the bare minimum.

As machines take over the easy work, human performance will be defined by what cannot be automated: thinking, deciding, imagining, integrating, and taking responsibility when it matters most.

What differentiates real performance today is not simply hitting set targets, but expanding them, shaping business outcomes, influencing direction beyond one’s role, and building capability for what comes next. 

That is the dimension today’s performance metrics must learn to capture. And it is something no annual review was ever designed to measure.