When AI Time Savings Turn Into Workload Creep
A new LSE and Protiviti study of 3,001 professionals finds AI saves time, but the freed hours often become extra work, and AI use doesn't reliably lift perceived performance.

Most teams that roll out AI tools expect the same outcome: tasks finish faster, and the spare hours go somewhere useful. A new global study suggests that second part is where things go wrong. The hours don't always disappear into better work. They often get filled with more work.

The research, titled 'Beyond the Technology: Who Really Captures AI's Gains at Work', comes from The Inclusion Initiative at the London School of Economics and Political Science (LSE), working with global consulting firm Protiviti. It's the third yearly survey in the series, and it collected responses from 3,001 professionals across the world between March and May 2026. The findings are worth a close read if you're the person deciding how AI fits into a team's day.
The pressure runs both ways
Start with the social side, because it shapes adoption as much as the tools do. Almost half of respondents (48pc) said they worry they'll be judged negatively for not using AI enough at work. That figure is more than double the share (20pc) who worry about criticism for using it. Put simply, far more people fear being seen as behind than being seen as leaning on a tool too heavily.

Usage has moved fast, too. Almost three-quarters of respondents now say they use AI at work, up from 29pc in 2024. When adoption climbs that quickly, the cultural expectation often arrives before the policy, training or workflow design that would make the tool useful. Teams in this position typically end up with uneven habits: a few enthusiasts who automate everything, and a larger group using AI quietly to avoid looking behind.
What 'workload creep' actually looks like
The report's central idea is what it calls 'workload creep'. The pattern works like this: AI makes a task quicker, but the time saved doesn't stay saved. It turns into new responsibilities, extra projects or a higher baseline for output. The productivity gain is real, but it's absorbed by the system around it.

Dr Grace Lordan, founding director of The Inclusion Initiative at LSE, put the dynamic plainly: "AI is becoming an essential part of how we work. But expectations are rising with it. Employees feel pressure to show they are keeping pace with AI while also taking on more." She added that leaders need to be clear about where AI adds value and how freed-up time should be used. Without that direction, she said, greater efficiency may simply raise the bar for what staff are expected to deliver. "Where those time savings go matters as much as the size of the savings."
That last line is the one I'd pin above every rollout plan. A time-saving tool with no decision about the reclaimed hours is, in effect, a decision to give them to whoever asks next.

The benefits people report
None of this means respondents found AI unhelpful. The benefits they described were substantial. Of those surveyed, 82pc said AI helps them complete tasks more quickly. 70pc reported reduced mental effort, 63pc said it helps automate manual tasks, and 62pc said it improves the quality of their work.
Among those who saved time with AI, half said they used that time to take on new responsibilities or projects. That's a sensible choice for an individual employee trying to build a case for promotion. It's a less comfortable outcome for a manager who assumed the freed hours would go to backlog, training or slack.

Usage isn't the same as performance
Here's the finding I'd take most seriously: more AI use doesn't reliably translate into higher perceived performance. The research found no consistent link between perceived job performance and AI adoption, training, in-house access, or integration into standardised workflows. In other words, rolling out the licence, running the training course and standardising the process didn't, on its own, produce a measurable performance lift in the eyes of the people involved.
The report points instead to characteristics that are not about software. Higher performance remained linked to conscientiousness, resilience, confidence in solving problems and willingness to learn. That doesn't make tooling irrelevant, but it does suggest that buying the tool is the easy part. Building the habits and judgement around it is harder.

Matt Duncan, managing director at Protiviti, framed the gap as a choice about capacity. "AI is creating meaningful productivity gains, but the bigger opportunity is how organisations use the capacity it creates. Simply absorbing those gains into existing workloads leaves much of AI's potential untapped." He said leaders should decide where that capacity creates the most value, whether through higher-value work, better quality or room for people to solve problems and innovate. "The goal shouldn't just be doing more work faster, but using AI to change what people are able to accomplish."
What IT and operations teams can take from this
For administrators and MSP owners, the report reads less like a verdict on AI and more like a brief on measurement. If your success metric is tickets closed or documents drafted per person, you'll likely see the numbers climb and conclude the rollout worked. Yet the study suggests those numbers can rise while perceived performance and quality stay flat, and while the load on your most capable people quietly grows.
A few questions are worth asking before the next change window. Who decides what happens to time saved by a new automation? Are people being measured on output that AI now inflates? Does training cover judgement about when to use the tool, or only how to click through it? And are staff who avoid AI being left behind, or being pushed toward it without clear guidance?
This kind of research is unlikely to settle the debate about AI and work. But the core message is practical and fairly portable: the productivity gain is the easy part to measure. Whether it lands as better work or simply more work depends on decisions your organisation makes after the tool is switched on. The full report, including the survey methodology, is available from Protiviti's publication page, and it's worth reading before you set the next round of targets.