AI speeds up individual work, but managers now have to judge it, govern it and prove it pays off.
Here is how leaders can close that gap.
Employees are turning out work faster with AI. Managing that work is getting messier. A manager today may be reviewing a report that software partly wrote, supervising tools they didn’t choose, and signing off on decisions shaped by systems they don’t fully understand.
A new survey puts numbers on the strain. monday.com commissioned Nielsen to question 1,701 directors and senior managers across seven countries. According to TechCrunch’s report of 21 October 2026, more than half of senior leaders said evaluating their employees’ work had become harder since AI arrived. Nearly half said they now spend more time reviewing and guiding what their teams produce.
Choosing the right AI tools is only part of the problem. Companies also have to rethink what productivity means, how they measure individual contribution, and how work gets organised when people and AI systems share the load.
Why AI Makes Performance Harder to Judge
AI can turn out a polished report, deck or analysis in minutes. A clean finished product, though, tells you very little about what the employee actually understood, decided or contributed.
Picture a manager reading an AI-assisted report. Did the employee spot the problem themselves? Did they check the evidence? Did they push back on the tool’s conclusions, or just accept its first answer? The document may look finished and still hide errors that take real effort to find.
That is a headache for traditional performance management, which leans on deliverables, deadlines and visible results. Once AI changes how those results are produced, managers need another way to see the thinking behind the work.
Many already are adjusting. In the monday.com survey, 78% of respondents said they had changed how they evaluate employees since adopting AI. They are putting more weight on skills such as writing good prompts, deciding which tasks to hand to AI and trying out new ways of working.
Those skills matter, but they shouldn’t become the whole scorecard. Someone who reaches an excellent result through sound judgement and careful AI use may add more value than a colleague who pumps out a mountain of unchecked material.
The priority for leaders is to judge the quality of decisions, the reliability of results and the value delivered. Output volume alone isn’t a measure of productivity.
The Hidden Work Behind AI Productivity
When companies tally up AI’s efficiency gains, they tend to miss the extra work it creates.
An employee may spend less time on a first draft but more time checking facts, fixing mistakes and reshaping the document for different audiences. Managers, meanwhile, have to decide what can be automated, set up review processes and step in when a system gets something wrong.
The survey shows how wide that gap can be. Some 94% of respondents said finding opportunities to improve work with AI was exciting. Yet 59% said finding and testing where AI actually adds value took more effort than it saved. And only 29% worked in organisations with fully optimised, centralised AI infrastructure.
A tool can speed up one task without improving the process around it. Take a marketing team using AI to write campaign copy. The drafting gets quicker, but someone still has to verify claims, secure approvals, check brand consistency and measure results. If none of that changes, the time saved on the draft barely dents the time it takes to get the campaign out the door.
Harvard Business Review’s 21 October 2026 article, The Deployment Failure Holding Back Your AI Investments, points to a related problem. AI tools can improve individual functions while leaving coordination between teams untouched.
So, measure productivity across the whole workflow, including review, coordination and correction. The speed of a single task tells you very little.
The Manager’s Role is Changing
Managers are now expected to understand both how people perform and the systems that shape that performance.
According to TechCrunch, 46% of surveyed managers use AI to help inform their decisions, and 39% of leaders use it to draft performance reviews. That saves time. It also raises hard questions about fairness, accountability and the evidence behind the judgement.
AI-written performance reviews deserve extra scrutiny. If a manager feeds in thin information, or past evaluations that carried bias, the system can repeat those flaws in its recommendations. A fluent write-up can sound objective without being properly grounded in evidence.
Research cited in the TechCrunch report has found bias in AI-based assessments of people, which strengthens the case for human oversight whenever AI touches employment decisions.
Leaders should set firm limits on where AI can be used in hiring, promotion, performance reviews and other consequential decisions. Managers should check the underlying evidence, let employees respond to assessments and keep the final call for themselves.
AI can help organise feedback or surface patterns in documented performance. It should never be the last word on someone’s ability, potential or worth.
Managers are increasingly the people who set standards, interpret evidence, develop their teams and use judgement where automated recommendations fall short.
The Governance Problem with AI Agents
Things get trickier once organisations deploy AI agents, which can carry out multistep tasks with varying degrees of independence.
In the survey, 58% of respondents expected to oversee more AI agents than human employees within the next year. Some 46% worried about losing control over what those agents do, and 35% feared being held responsible for AI-driven decisions.
The question for executives is blunt: when an agent gets it wrong, who answers for it?
Give an agent access to business systems, customer data or internal workflows, and the risks climb quickly if its permissions, goals and escalation rules are vague. Mistakes can also travel across connected tools before anyone spots them.
Organisations need clear governance. Spell out what an agent can access, which actions it can take on its own and which decisions need human sign-off. High-impact actions need defined escalation routes, audit trails and a named person responsible for reviewing outcomes.
Count the cost of supervision too. If staff spend large chunks of time watching agents, fixing their errors and juggling overlapping tools, those hours belong in any calculation of what automation saves.
The aim is to hand over the right work without handing over accountability to a system that cannot carry organisational responsibility.
Five Ways Leaders Can Turn AI Adoption into Measurable Productivity
Companies don’t need to slow down every AI initiative. They need to be more disciplined about where the technology belongs and how its value gets measured.
1. Measure outcomes across complete workflows
Track how long a process takes from start to finish, including verification, corrections, approvals and coordination. Compare before and after AI adoption using the same yardsticks.
Useful indicators include error rates, turnaround time, customer satisfaction, rework and cost per completed task. More output only counts if it leads to better business results.
2. Set clear standards for AI-assisted work
Employees should know which tasks can use AI, what must be verified and when human review is mandatory. Managers should define acceptable quality instead of leaving each person to invent their own rules.
Consistent standards also make performance reviews fairer, because everyone is judged against the same expectations.
3. Train managers, not just employees
Most AI training focuses on how to use the tools. Managers also need to learn how to assess AI-assisted work, spot unreliable output, redesign workflows and weigh the risks of automated decisions.
Without that, staff will adopt AI faster than the organisation can handle the consequences.
4. Consolidate tools and control access
The survey found that 57% of respondents had more tools in circulation than before they adopted AI, and about 51% said their tools had become more interdependent.
Leaders should map the AI tools in use across departments, find the overlaps and approve specific systems for sensitive work. Firm access controls and integration standards cut down on needless complexity and make oversight far easier.
5. Keep humans accountable for consequential decisions
AI can support analysis, summarise evidence and suggest options. Decisions about promotions, customer rights, financial exposure or anything else with serious consequences need human judgement and accountability.
Managers should document the evidence behind important decisions and make sure a named person owns each one.
The Leadership Test for AI Productivity
AI adoption is as much a management challenge as a technology investment. Individual tools may speed people up, but companies need sound processes, consistent standards and reliable infrastructure to turn that speed into wider performance gains.
The monday.com survey does not show that AI has cut productivity across organisations. What it shows is that managers face new demands as adoption deepens, even while enthusiasm for the technology stays high.
For executives, the next phase calls for a change of focus. The question is no longer how much work AI can produce or how many tasks it can automate. It is whether the technology improves the quality, speed and economics of the work the organisation actually needs to do.
Companies that tackle evaluation, governance and workflow design together will be best placed to see where AI creates value, where it adds cost and where human judgement still matters most.
Author’s Note:
I believe one of the biggest misconceptions about AI in the workplace is that producing work faster automatically makes an organisation more productive. In reality, speed is only one part of performance. The real test is whether that speed translates into better decisions, stronger outcomes and less unnecessary work.
What stands out to me in this survey is the growing gap between individual efficiency and organisational readiness. Employees can use AI to complete tasks in less time, yet managers may spend those saved hours checking accuracy, evaluating contributions and correcting mistakes. If companies measure only how quickly work is produced, they risk overlooking the additional effort required to make that work reliable.
I also believe organisations need to rethink how they recognise employee performance. When AI becomes part of everyday work, the value of an employee should not be judged solely by the volume of output or how little time a task takes. Critical thinking, sound judgement, the ability to challenge an AI-generated answer and the willingness to take responsibility for a decision become equally important.
For leaders, the next step should be to stop treating AI adoption as a race to introduce more tools. It should be an exercise in redesigning how work gets done. That means setting clear expectations, training managers to evaluate AI-assisted work, measuring the time spent on verification and ensuring that accountability remains with people.
AI should give managers more room to lead, coach and develop their teams. If it instead leaves them with more tools to supervise and more output to inspect, the organisation has improved the speed of production without necessarily improving productivity.
The question leaders should ask is not how much work AI can produce, but how much meaningful work it helps people accomplish. That distinction will determine whether AI becomes a lasting source of organisational value or another layer of complexity for managers to absorb.
