AI rebounding: why some employers are asking back the people AI replaced
· 5 min read
New SHL research finds some UK and US workers replaced by AI were later asked back. Why it happens, and how leaders can understand the work before automating it.
In brief
Some employers are now asking people to come back after replacing them with AI. A survey of 2,036 working adults in the UK and US, commissioned by the talent assessment firm SHL, found that 16 per cent said this had happened to them personally, including 12.8 per cent in the UK and almost one in four of those aged 18 to 24. Fewer than half said their manager fully understood their skills beyond their job title, and only about a quarter said jobs were being redesigned around the new tools. The figures are reported by workers themselves and come from a vendor, so they are a signal rather than a precise count.
The lesson for leaders is simple. A job description lists tasks, but the person in the role usually does much more: catching exceptions, knowing customers and spotting when an output is wrong. Remove the role on the strength of its task list and that knowledge goes too, often unnoticed until problems appear.
This piece explains how leaders can study the real work first, using tools from the AI Enablement Toolkit for Leaders, where training helps managers understand what their people contribute, and where the toolkit stops, since it is not employment or legal advice.
Picture the phone call. A few months ago a role went, because an AI tool could do the work. Now someone is ringing the person who held it to ask whether they would come back.
New research suggests that call is more common than most leaders would guess. It also suggests why it happens: the work was automated before anyone understood it.
What the research found
SHL, a talent assessment firm, commissioned Propeller Insights to survey 2,036 employed adults in the UK and US in June and July 2026. Its report, [Talent Mobility at Work in 2026: The Hidden Workforce](https://www.shl.com/assets/premium-content/talent-mobility-at-work-in-2026-report-en-v3.pdf), was covered at the weekend by [The Globe and Mail](https://www.theglobeandmail.com/business/careers/article-reversing-course-workers-replaced-by-ai-are-being-asked-back/). This is vendor research and the answers are self-reported, so the figures are best read as a signal rather than a measurement.
The headline finding is what SHL calls "AI rebounding". 16.0% of respondents say they were personally replaced by AI and later asked back by the same employer. In the US the figure is 19.1%; in the UK it is 12.8%. Among 18 to 24 year olds it rises to 23.8%.
The more useful numbers sit underneath. 48.7% of workers say their employer is introducing AI tools, but only 24.1% say jobs or processes are being redesigned around what AI can do. Only 41.8% say their manager fully understands their skills beyond their job title, and in the UK that falls to just over one in three.
If they are asked back, people expect the deal to change. 50.8% would want a significant pay rise, 42.2% would want written job security guarantees and 39.7% would want full transparency about how AI would be used in their role. SHL's Lucy Beaumont summed up the pattern: "The pattern we see is automate first, understand the work second. That's backwards."
The same weekend brought related signals. A YouGov survey of 817 STEM workers in Germany, commissioned by the recruiter SThree and [reported by dpa-AFX](https://www.finanznachrichten.de/nachrichten-2026-10/69926371-wandel-im-berufsalltag-ki-wird-zum-kuendigungsgrund-016.htm), found 38.4% had seen colleagues leave for AI-related reasons, most often because the work had changed beyond recognition. And Apollo's chief economist Torsten Slok [argued on 24 October](https://www.apollo.com/wealth/insights-news/insights/daily-spark/no-signs-of-ai-payoff-outside-the-tech-sector) that "in business, intelligence is mostly judgment."
What it means for leaders
A job description lists tasks. A person in the job does more than that. They catch the exception nobody wrote down. They know which customer needs a call rather than an email. They hold the workaround that keeps a fragile process running, and they are often the only one who can tell when an output is wrong.
When a role is removed on the strength of its task list, all of that goes with it. Sometimes nobody notices for months. Then the errors, complaints and rework arrive, and the cheapest fix is to ask the person back, on terms that now cost more and with trust that has already been spent.
None of this argues against using AI to change work. It argues for understanding the work before changing it. John Sansbury made the point about change in general, long before generative AI, in The Journey to Service Excellence:
"When making such a change, the temptation is to start from scratch. However, this is potentially wasteful if you have elements of the current environment that are working perfectly well and they are discarded without thought."
His remedy is a current state analysis: find out what is working, what should be kept and what should change, and use it as the baseline for judging progress. Esm's reading is that the same discipline applies when the change is AI. The book is about service management, not AI, so this is our application of his principle rather than his view on automation.
Where the toolkit helps
The AI Enablement Toolkit for Leaders is built for this moment, because it puts seeing the work before redesigning it.
- Tool 7 (see the work) follows real cases rather than the documented happy path, and asks the people doing the job what they know about the customer or the process that the system does not. - Tool 8 (customer value-stream map) shows where judgement is needed because the process cannot cope with variation. Those points are where a person is hardest to replace. - Tool 9 (work inventory and opportunity triage) offers six choices, including stop, simplify and keep human. Automation is one option, not the default. - Tool 14 (protect the apprenticeship) matters because younger workers report the most rebounding. Removing junior tasks without a new route to judgement is the anti-pattern the toolkit calls "Automating away the school". - Tool 12 (trust-by-design check) includes a pre-mortem prompt for the moment colleagues discover a job impact that was softened or hidden. - Tool 18 (proof-of-value scorecard) keeps cash saved, capacity released and risk apart, so a headcount saving is not booked before the work has been proven to run without the person.
Where training fits
The gap SHL describes is partly a capability gap. Managers who do not fully understand their people's skills cannot judge what a role really contributes. Esm's training and enablement helps leaders run the conversations and observation that Tools 3 and 7 depend on, and practise redesigning work with the people who do it rather than around them.
The honest limits
The toolkit is a management method. It is not employment law, consultation or redundancy advice, and anyone restructuring roles needs current specialist guidance. It has no skills inventory or skills-mapping tool, so it will not tell you everything a person knows; it gives you a disciplined way to find out. And SHL's figures are self-reported by workers, from a firm that sells talent assessment. They do not prove how often employers really reverse AI decisions.
What they do show is a pattern worth checking in your own organisation before the next role goes: is the plan based on what the person does, or only on what the job description says?
Tags: ai-adoption, workforce-planning, ai-rebounding, leadership, service-management
Sources
- https://www.shl.com/assets/premium-content/talent-mobility-at-work-in-2026-report-en-v3.pdf
- https://www.theglobeandmail.com/business/careers/article-reversing-course-workers-replaced-by-ai-are-being-asked-back/
- https://www.finanznachrichten.de/nachrichten-2026-10/69926371-wandel-im-berufsalltag-ki-wird-zum-kuendigungsgrund-016.htm
- https://www.apollo.com/wealth/insights-news/insights/daily-spark/no-signs-of-ai-payoff-outside-the-tech-sector
Join the Exchange free
Join free to unlock the complete knowledge library.
Founder membership is free for life
More from Esm
Related articles
AI Job Cuts and the Unfilled Role: Where Should Freed Capacity Go?
Barclays' AI-assisted job cuts and a quieter freeze on junior hiring raise one leadership question: when AI frees capacity, who decides where it goes before the spreadsheet decides for you?
AI and Middle Managers: Who Checks the Work When the Middle Thins?
New Conference Board research shows managers are being asked to check a flood of AI output just as many organisations expect to need fewer of them. What it means for leaders, and the toolkit tools that help.
The AI confidence gap: why UK leaders and their teams see different futures
New Microsoft research finds 84% of UK leaders are confident about Britain's AI future but only 41% of employees agree. Here is what that gap means for leaders, and four practical ways to close it.
Members
You Cannot AI What You Cannot See
Every university leader is under pressure to show what artificial intelligence will do for the institution. Here is an uncomfortable truth that a recent value stream mapping workshop surfaced, and it has nothing to do with the cleverness…
Members
Seven People, Seven Different Ways: Why a Map Is Not Enough
A consultant walks into a council housing repairs contact centre. Seven people take calls all day. My roof is leaking, my door is jammed, my toilet is blocked. He asks the manager to describe the procedure for handling these calls. The m…
“I just need to know”: why consistency beats speed
Her words, recorded in The Journey to Service Excellence, belong above every service desk in higher education: “I don't care whether it's one week or three weeks, I just need to know so that I can provide enough notice to be able to plan…