AI was meant to take the chores. Workers say it is taking the work they love, too

AI was supposed to clear the busywork so people could spend more time on the parts of the job they care about. New research from Atlassian's Teamwork Lab suggests that, for many workers, the opposite is happening: AI is taking on enjoyable work as well as the chores.

What the survey found

Atlassian published the study on 4 November 2026. It is vendor research: a double-blind survey of 1,013 US full-time knowledge workers, with fieldwork from 28 September to 1 October 2026, and AI-use questions asked of 949 people who already use AI. Because Atlassian sells collaboration and AI products, treat the findings as useful but interested.

The headline figures are stark.

• 82% of AI users say AI is doing at least some of the work they enjoy, and 82% say the same about work they dislike.
• 10% say AI has taken on all the work they enjoy, against 3% for all the work they dislike.
• Among workers whose team made AI standard, 71% say AI handles a moderate or greater share of work they enjoy, against 24% of occasional users.
• Leaders at director level and above are more likely to say AI handles a moderate or greater share of enjoyable work (60%) than managers (42%) or individual contributors (36%).
• 95% say AI made at least one expert task easier for non-experts; only 28% say it lets non-experts make decisions that used to require a specialist.
• 54% expect AI to create more specialists, rising to 60% among heavy users, because expert judgement matters more when tools are widely available.
• On high-risk decisions, 78% want humans leading; only 3% would let AI make those calls alone.

Molly Sands of the Teamwork Lab summarised the tension: "We promised AI would clear the busywork so people could focus on what they love. Instead, it's taking the work they love, too. Nobody designed it that way. It's just what happens when no one decides what to protect."

What it means for leaders

Enjoyable work is not a soft extra. It is often where craft, judgement and pride live: the hard case a specialist wants to solve, the design choice a product person cares about, the conversation a manager is good at. When AI takes those first, people lose the reason they stay, and the organisation loses the skills it will need when the model is wrong.

The leadership risk is sharper still. The survey finds that the people with most power to set AI norms, directors and above, and teams that make AI standard, are the ones most likely to report AI taking enjoyable work. Leaders can hand off the parts of the job they personally find easy or boring, without noticing that those same parts are what keep someone else engaged, or that they still need a human accountable when the stakes are high.

There is also a useful distinction buried in the numbers. Making an expert task easier for a non-expert is not the same as letting that person take a decision that used to need a specialist. The first is common in this survey. The second is not. Treat them as different design choices, not as one continuous slide toward automation.

A service management principle that still applies

Esm does not treat this as a new AI puzzle. It applies a principle John Sansbury set out in The Journey to Service Excellence about process and automation: "A sub-optimal process that is automated may simply fail quicker or in unexpected ways."

If nobody has decided which enjoyable work to protect, which decisions stay with specialists, and who owns high-risk calls, then rolling out AI simply speeds up the muddle. Consistency, simplification and clear ownership come before automation, not after the tool has already reshaped the day.

Practical steps

1. Before the next AI rollout, ask the people doing the work which tasks they most want to keep, and which they would gladly hand over. Write both lists down. Do not assume leaders' preferences match the team's.
2. Separate "made easier" from "handed over". If AI drafts analysis for a non-expert, say who still makes the decision and how that person is accountable.
3. For every high-risk use, name the human who leads the call. The survey's 78% who want humans in charge of high-risk decisions is a clear signal; the 3% willing to leave it to AI alone is not a mandate.
4. Where a team makes AI standard, review after 30 and 90 days what enjoyable work has shifted, not only what got faster. Speed without craft is a weak bargain.
5. Measure correction and rework alongside any claimed time saved. Parallel surveys this week, from Miro on private AI sessions and quality, and from Fractl on cleanup time after AI work, point the same way: unreviewed output creates hidden cost and lost trust.

Where training fits

This is a leadership and service design problem before it is a model problem. Managers need practice in redesigning work with their teams, naming owners for decisions, and reporting outcomes rather than licences. Training helps when it is built around real services and gives people the confidence to protect craft on purpose, including when a vendor slide says "automate everything".

Honest limits

This is a US survey from a vendor with AI products to sell. Enjoyment is self-reported. It does not tell you which use cases are safe in your sector, and high-risk decisions in regulated work still need legal, risk and domain expertise. Deciding what to protect is a start, not a substitute for that advice.

The organisations that keep people who care about the work will be the ones that decide, out loud, which work AI may take and which work stays human on purpose.

Tags: AI adoption, leadership, enjoyable work, automation, service management, Atlassian

Sources

• https://www.atlassian.com/blog/ai-at-work/ai-was-supposed-to-take-on-the-drudgery-workers-say-its-taking-on-the-good-stuff-too

• https://miro.com/blog/state-of-intelligent-collaboration-2026/

• https://www.businesswire.com/news/home/20261103271772/en/Individual-AI-Productivity-Soars-but-Organizational-Impact-Stalls

• https://kvia.com/stacker-ai/2026/11/04/most-workers-say-getting-good-work-from-ai-takes-as-much-effort-as-training-a-new-hire/

• https://acora.com/insights/businesses-are-hitting-pause-on-ai-but-their-employees-arent/

• https://www.frac.tl/ai-slop-at-work/

• https://www.iteuropa.com/news/uk-firms-face-ai-reality-check-returns-fail-match-ambition