Guide · Your team and AI
AI fatigue: why AI can make work more tiring, and what managers can do
AI fatigue is the tiredness people feel from working with AI. It has two different sources: the pressure to use AI, with new tools, mandates and keeping up; and the work AI creates, with more tasks, constant checking and work spilling into breaks. Each needs a different fix, and neither is solved by telling people to use AI more.
By Andrew Ward, Founder of ScorchsoftChecked 10 minute read
In short
- Work out which kind you’re looking at: pressure to use AI, or the extra work and checking AI brings. The fixes differ.
- Checking AI’s output is real work. In a BCG survey of 1,488 US workers (vendor research, March 2026), overseeing AI was the most mentally taxing way of using it.
- Never set AI usage targets. Judge finished work, not how much AI someone used.
- Use the HSE Management Standards you already have: demands, control, support, relationships, role and change cover the causes of AI fatigue too.
Written for: Managers, team leaders and HR leads in UK organisations where people use AI in their work.
What is AI fatigue?
AI fatigue is the tiredness or overload people describe from working with AI: keeping up with new tools, checking a steady stream of output, and doing more in the same hours. It describes how people feel at work, not a medical condition.
You will also see it called AI burnout. This guide uses the plainer word, fatigue, because it describes the experience without labelling anyone. If tiredness is affecting someone’s health, that is a matter for a GP or another qualified professional, and the last section says where to go.
It can seem odd that a tool sold as a time saver makes people tired. The reason is that AI speeds up some steps and not others. Drafting gets quicker; deciding, checking and fixing don’t. If a report took five days to put together and one to check, and AI cuts the putting-together to a day, checking is now half the job. The work that’s left is the most demanding kind.
Is it the pressure to use AI, or the work itself?
Usually one more than the other, and it is worth finding out which. Pressure fatigue comes from mandates, constant new tools and being watched for use. Work fatigue comes from more tasks, more checking and no natural pauses. Asking the team is the quickest way to tell.
| Fatigue from the pressure to use AI | Fatigue from the work AI creates | |
|---|---|---|
| What people say | “Another new tool.” “I’m told to use it but not shown how.” “I feel judged if I don’t.” | “I get through more, but I’m shattered.” “I’m checking all day.” “I end up doing it in the evening.” |
| Where it comes from | Mandates, targets, tool churn, unclear rules, fear of falling behind | Wider scope, parallel tasks, constant review of output, lost breaks |
| What helps | Fewer tools, clear rules, time to learn, permission to say no to AI for a task | An agreed pace, protected checking time, pressure valves, honest workload talks |
| What makes it worse | Usage targets, league tables, “AI-first” orders with no support | Raising expectations every time output rises |
UK surveys suggest the pressure side is widespread. In Henley Business School’s 2026 survey of 2,900 UK full-time workers, 61% felt overwhelmed by the pace of AI development, and 63% said they sometimes choose not to use AI tools that are available to them. Choosing not to use a tool for a task can be good judgement. Treat it as information, not resistance.
What does the evidence say about AI fatigue?
The evidence is early, mostly from the US, and partly from firms that sell AI or consulting. It points the same way: AI tends to expand work rather than shrink it, and overseeing AI output is tiring. Treat the figures as signals, not settled facts.
| Study | What it is | What it found |
|---|---|---|
| Ranganathan and Ye, “AI Doesn’t Reduce Work—It Intensifies It”, Harvard Business Review, 9 February 2026 | Eight months of observation and more than 40 interviews at one US technology company of about 200 people. Qualitative, one firm, work in progress. | People took on tasks from other roles, prompted AI in lunch breaks and evenings, and ran several AI jobs at once, often without being asked. Higher output raised expectations, which pushed the pace up again. |
| Bedard and others, “When Using AI Leads to ‘Brain Fry’”, Harvard Business Review, 5 March 2026 | A Boston Consulting Group (BCG) survey of 1,488 full-time US workers at large companies. Self-report, one point in time; vendor research. | About 14% of those using AI at work said they had experienced mental fatigue from using or overseeing AI beyond their capacity. Overseeing AI was the most mentally taxing kind of use. Intent to quit was 34% among those affected, against 25% among the rest. |
| Lenny’s Newsletter, How tech workers are feeling in 2026, 7 July 2026 | A self-selected online survey of tech workers who read a newsletter, not the UK workforce. Sample size not published on the free page. | The top AI worries were being expected to do more for the same pay (51%) and an unsustainable pace (46%). Losing their job to AI was chosen by 22%. |
Read together, they suggest that the tiring part of AI isn’t typing prompts. It is the extra scope, the constant judging of output, and the quiet rise in what counts as a normal day’s work. Those are things a team can manage.
How do the HSE Management Standards apply to AI fatigue?
The Health and Safety Executive’s Management Standards name six areas of work that cause stress if they aren’t managed: demands, control, support, relationships, role and change. AI touches all six, so you can use the risk assessment you already do rather than invent a new process.
The HSE says employers have a legal duty to protect workers from stress at work by assessing the risk and acting on it, in the same way as other health and safety risks, and that employers with five or more workers must write the risk assessment down. The Management Standards are its approach to doing that. The table maps each standard to AI.
| Standard | How AI can strain it | A fix that fits |
|---|---|---|
| Demands | Output targets rise with every saving; checking load grows; work spills into breaks | Agree a pace and say what saved time is for; protect time for checking |
| Control | People are told which tool to use and how often, with no say | Let people choose when AI helps a task and when it doesn’t; no usage targets |
| Support | Tools arrive with no training or time to learn | Time to learn, a named person to ask, and a safe way to report mistakes |
| Relationships | Colleagues judged, or mocked, for using or not using AI; workslop passed on for others to fix | A team rule: you own what you send, AI-made or not |
| Role | Nobody knows who checks AI’s work or who owns the result | Name an owner for each piece of AI-assisted work |
| Change | A steady stream of new tools and rules, announced without consultation | Fewer, slower changes; ask the team first; review after a few weeks; bring hidden tools into the open without blame (see shadow AI) |
Mapping AI to the standards doesn’t by itself meet any legal duty, and your organisation’s stress risk assessment may already cover some of this. Use the table to add AI to the conversation you already have, and take HR or health and safety advice where you need it.
What can managers do about AI fatigue this month?
Ask the team which kind of fatigue it is, agree a pace and the pressure valves people may use, and protect time for checking. Stop anything that rewards using AI for its own sake. None of this needs a budget.
Step 1: Ask, in ten minutes (owner: the manager)
At a team meeting or in an anonymous form, ask three questions: what about AI is tiring you, is it the pressure to use it or the work it creates, and what one change would help most. Share what you heard and what you’ll change.
Step 2: Agree a pace and the pressure valves (owner: the manager, with the team)
The AI Stairway™ calls a pace unsustainable when you can only keep it by working longer, checking less or hiding that you’re behind. Agree in advance what someone does when they hit that point: drop a lower-priority task, ask for a deadline to move, or hand back a job. Using a valve is the system working, not a failure. Module 7 of the course practises setting a pace your team can keep.
Step 3: Protect checking time (owner: each person, backed by the manager)
Run several AI jobs at once only if you can check them all. Book focus time for review, and count it as work. Module 5 teaches this as being kind to your attention. Unchecked output isn’t a saving; it is a cost passed to someone else.
Step 4: Remove the pressure signals (owner: the manager or owner)
Drop any usage league tables or targets. Cut the tools to the ones people actually need. Tell the team that choosing not to use AI for a task is allowed when it isn’t the right tool. If you have AI champions, make sure they have time for it too; see our AI champions guide.
Hayley coordinates client services at a 40-person accountancy practice. Since the practice approved an AI tool, she has been drafting client chasers in batches, running several at once and checking them in the evening to keep up. Her output has doubled, and so have expectations. Her manager, Gareth, notices her messages arriving at 10pm. Instead of praising the volume, he asks which kind of tiredness it is. It is the work: the checking. They agree that Hayley drafts no more chasers in a day than she can check in working hours, that checking time is booked in her calendar, and that if month-end piles up she can push non-urgent chasers to the next day without asking. Her output settles a little below its peak, and the late-night messages stop.
The one thing not to do
Don’t set AI usage targets or rank people by how much they use it. How much AI someone uses isn’t work done. Targets add the pressure kind of fatigue to the work kind, and they reward volume over judgement, which is how slop gets made.
Where can people get support?
If tiredness or stress from work is affecting someone’s health, sleep or life outside work, they should talk to someone: their employee assistance programme if there is one, their GP, or a charity such as Mind. In an emergency, call 999.
- Your employee assistance programme (EAP). Many employers offer confidential support by phone. Ask HR or check your handbook.
- Your GP, for stress or exhaustion that isn’t easing.
- Mind, whose page on work and stress (opens in a new tab) covers workload and work-life balance.
- HSE’s guidance for workers on work-related stress (opens in a new tab), including how to raise it with your employer.
- Urgent help. NHS 111 online or call 111 for urgent mental health help; 999 if anyone is in danger; Samaritans (opens in a new tab) on 116 123, free, 24 hours a day.
If the tiredness is mixed with worry about what AI means for jobs or for keeping up, our guides to AI anxiety at work and Will AI take my job? may help.
Questions people ask
Is AI fatigue the same as burnout?
People use the two terms for the same experience. This guide uses “fatigue” because it describes tiredness at work without labelling anyone. If exhaustion is affecting someone’s health, a GP or another qualified professional is the right person to talk to.
Why does AI make some people more tired, not less?
Because it speeds up some steps, such as drafting, and not others, such as deciding and checking. The slower steps take over more of the day, expectations rise with output, and research from 2026 found work spilling into breaks and evenings. The time saved doesn’t always come back to the person.
How many AI tools at once is too many?
There is no reliable number, and it depends on the work. A sensible rule is to run only as many AI tasks at once as you can properly check, and to keep the set of tools your team uses as small as it can be.
Should we stop using AI if the team is tired?
Not necessarily. Find out which kind of fatigue it is first. If it’s the pressure, ease the pressure: fewer tools, no targets, clearer rules. If it’s the work, change the pace and protect checking time. Stopping altogether rarely deals with either cause.
Does AI fatigue count as work-related stress?
It can be one of its causes. HSE describes work-related stress as arising from six areas, including demands, control and change, and AI can affect all of them. Include AI in your stress risk assessment, and take HR or health and safety advice on your own situation.
Training for the whole team
The AI Stairway™ gives everyone the same careful habits, then helps your team agree its rules and test one AI idea properly. Everyone can start with the free introductory module.
AI training for teamsSources
Every figure on this page comes from one of these. Links open in a new tab.
- Work-related stress and how to manage it: stress risk assessment (opens in a new tab). Health and Safety Executive, updated 20 January 2026.
- Work-related stress and how to manage it: causes of stress at work (opens in a new tab). Health and Safety Executive, updated 4 June 2024.
- Management Standards (opens in a new tab). Health and Safety Executive, updated 13 January 2025.
- AI Doesn’t Reduce Work—It Intensifies It (Ranganathan and Ye) (opens in a new tab). Harvard Business Review, 9 February 2026.
- AI promised to free up workers’ time. UC Berkeley Haas researchers found the opposite (opens in a new tab). Berkeley Haas, 18 February 2026.
- When Using AI Leads to “Brain Fry” (Bedard, Kropp, Hsu, Karaman, Hawes and Kellerman) (opens in a new tab). Harvard Business Review, with Boston Consulting Group (vendor research), 5 March 2026.
- When using AI leads to “brain fry” (BCG’s note on the survey of 1,488 US workers at large companies) (opens in a new tab). Boston Consulting Group (vendor research), 5 March 2026.
- ‘AI brain fry’ is real — and it’s making workers more exhausted, not more productive, new study finds (press coverage of the BCG survey, open to read) (opens in a new tab). Fortune, 10 March 2026.
- Hype versus reality: a 12-month AI pulse check (opens in a new tab). Henley Business School, 4 June 2026.
- How tech workers are feeling in 2026 (self-selected survey of tech workers) (opens in a new tab). Lenny’s Newsletter, 7 July 2026.
- Work and stress (opens in a new tab). Mind, read 29 October 2026.