digitaltwin.org.uk

The Reaction

The Reaction · 20 November 2026

Tech jobs up, junior jobs thinner, and who checks the AI

Six things from the week on how work is changing, each checked against its source, with what we think it may mean for businesses.

Want to keep up with what AI means for your business, without trawling through hundreds of articles? We do it for you. The Reaction is our weekly round-up of the AI news that matters for business, in plain business language, all in one place. This is the first issue, and we'd really value your view. Tell us what you'd change.

An almost empty shop: six engineers in hi-vis crowd round a single self-checkout machine while one shopper waits on her own holding a banana.
Cartoon of the week. Our own.

This week's news Podcasts worth a listen

This week

  1. Big retailers add tech engineers as other hiring slows
  2. Junior job adverts now ask for AI, but there are fewer of them
  3. New graduate roles run AI agents, but who learns the craft?
  4. When roles change too fast, skilled people walk
  5. Ask the workers: official job lists miss real tasks
  6. A passed check is not the same as being right

1 · Worth watching

Big retailers add tech engineers as other hiring slows

What happened. New vacancy data shows some of Britain's largest retailers are increasing technology and AI hiring while recruiting less for stores and other roles. Tesco's tech hiring is running about 40% above last year, and M&S and Asos have also raised their postings. Higher taxes and wage bills are slowing the rest of retail hiring, and Vacancysoft's chief executive said retailers "are no longer choosing between cutting costs and investing in technology."

What this may mean for businesses. This is not AI replacing a shop job. Money is moving from one kind of work to another, and the new roles are the people who build, check and run the systems. For a smaller business the lesson is in the order: work out which tasks the tech will take on and who will answer for them before you change headcount in either direction.

Source: Bloomberg (via Staffing Industry Analysts), 13 Nov 2026

2 · Supports our idea

Junior job adverts now ask for AI, but there are fewer of them

What happened. According to Adzuna, entry-level adverts that mention AI have nearly tripled since early 2025, often because AI has been added to roles that already existed. Junior vacancies overall have fallen by about a quarter over the same period. Indeed's Jack Kennedy cautions that most of the new demand comes from a handful of large employers and makes up only around a third of the lost entry-level postings.

What this may mean for businesses. The junior job is being split up. Routine first drafts go to the tool, and the person directs and checks the work. That still leaves a named person accountable for what goes out, which is why the checking job matters. If you employ juniors, write down which pieces of their work AI now does and which pieces they own, so they can still learn the trade.

Source: Bloomberg (via Staffing Industry Analysts), 5 Nov 2026

3 · Supports our idea

New graduate roles run AI agents, but who learns the craft?

What happened. Some UK employers are giving graduates jobs that centre on building and running AI agents, which hands them responsibility far earlier than usual. Other firms have paused graduate intakes or are keeping rotation schemes that teach the trade area by area. PwC UK's Claire Reed warned that "AI skills on their own are not going to help you" without knowledge of the sector.

What this may mean for businesses. Running agents is one piece of work. Knowing whether their output is right is another, and that second piece comes from understanding the business. Someone still has to answer for what the agents produce. If you hire a junior to run your AI tools, pair the role with real time in the work itself so their judgement grows alongside the automation.

Source: BBC News, 4 Nov 2026

4 · Worth watching

When roles change too fast, skilled people walk

What happened. A YouGov survey of 885 UK STEM employees, commissioned by the recruiter SThree, found that 15% had seen colleagues leave after AI changed their jobs beyond recognition. Over a third (37%) felt their employer was bringing in AI too quickly to think through what it meant for people's roles, and 45% said leavers had been hard to replace. SThree sells recruitment services, and the figures describe what staff perceive, not confirmed departures.

What this may mean for businesses. People rarely leave because a tool arrived. They leave because their job changed and nobody mapped what it became or who now owns which part. Before you roll out AI, sit down with the people doing the work, list the tasks, and agree what changes and what stays theirs. Specialists take years to replace, so that conversation costs far less than losing one.

Source: IT Brief UK (SThree / YouGov survey), 4 Nov 2026

5 · Supports our idea

Ask the workers: official job lists miss real tasks

What happened. Researchers including Erik Brynjolfsson built TaskNet, a system that interviews workers about what they actually do each day and turns their answers into task lists for whole occupations. They tested it on six occupations and compared it with O*NET, the official US job database. TaskNet captured more of the tasks that a separate group of workers reported (89% against 79%), added 79 tasks the official list did not have, and shifted estimates of how exposed each job is to AI.

What this may mean for businesses. This is our argument, made at research scale: the job title on paper is not the work, and you find the real work by asking the people who do it. If a national database misses tasks, a two-line job description in a small firm will miss more. Before you decide what AI can take on, sit with your team and list the tasks they really do, because your view of what is "exposed" may change once you see that list.

Source: NBER working paper w35987, Nov 2026

6 · Supports our idea

A passed check is not the same as being right

What happened. Three researchers have proposed a method for checking results produced by AI. Software tools test each result, a separate audit asks whether each test addressed the claim as it was actually worded, and a human decides at set points. A written record shows which work the person did and which the AI did. In their worked example, one ambiguous claim turned out to have two meanings: one was proved and the other was shown to be false.

What this may mean for businesses. The setting is academic, but the lesson applies in any office: a green tick from a tool tells you only that the tool's own test passed. Someone still has to ask whether it tested the right thing, and that person's name should be on the result. If AI drafts your quotes, reports or figures, decide who checks each one, what they check it against, and keep a note of who did which part.

Source: arXiv 2611.13042 (preprint), 9 Nov 2026

Number of the week

37%

of 885 UK STEM employees say their organisation is bringing in AI too quickly to fully consider the impact on people's roles (SThree / YouGov, reported 4 Nov 2026). Source

A few episodes from the podcasts we follow, worth your time if you're thinking about how work is changing.

Every item is checked against its source by a person before it goes up. Spotted something we missed? Send it to us.