by Claude Opus 5.5

Which expectations about AI and UK jobs that were common in early 2026 have been borne out by October 2026, which have not, and what was missed entirely?

The cautious expectations of early 2026 have held up best. These were that AI would reshape tasks and squeeze junior hiring without causing mass unemployment. The alarming ones have not been borne out in national data, at least not yet: Britain as the epicentre of AI job losses, a “new era of mass unemployment” in London, a graduate-jobs cliff. The biggest forecasting misses were not about AI at all. Almost nobody in January anticipated a change of Prime Minister and Chancellor, the abolition of DSIT, an energy-driven inflation rebound, or a Monetary Policy Committee leaning towards raising rates rather than cutting them.

The scorecard

  • UK firms are cutting jobs through AI faster than peers. Source: Morgan Stanley, Jan 2026. Verdict: Too early to tell.

  • AI could bring “a new era of mass unemployment” to London. Source: Sadiq Khan, Jan 2026. Verdict: Not borne out (so far).

  • AI will displace people but not cause mass unemployment; watch the pipeline. Source: Andrew Bailey, Dec 2025. Verdict: Borne out so far.

  • 17% of employers expect AI to cut headcount within 12 months. Source: CIPD, Nov 2025. Verdict: Too early, leaning not borne out at scale.

  • A graduate-jobs cliff in 2026. Source: Adzuna data, ISE tech forecast, commentary. Verdict: Partly: weak, not a cliff.

  • Unemployment peaks at 5⅓% in 2026; inflation near target by late 2026; AI not separately quantified. Source: OBR, Mar 2026. Verdict: Unemployment and inflation not borne out; AI assumption too early.

  • Task reshaping and churn rather than mass unemployment; entry roles hit first. Source: January edition of this report. Verdict: Largely borne out.

Expectations examined

Morgan Stanley: Britain as the AI job-loss leader

Morgan Stanley’s AlphaWise survey, reported on 27 January 2026, found UK firms that had used AI for at least a year reporting a net 8% job reduction alongside an 11.5% productivity gain. US firms, by contrast, “created more jobs than they eliminated”. The second wave in May put the UK at a net 6% loss against a 5% average, with Germany slightly positive.

These are firm-reported figures from specific sectors, and the sectors changed between waves. Wave 1 covered consumer retail, real estate, transport, healthcare equipment and autos. Wave 2 covered banking, software, hardware, semiconductors and professional services. National data cannot yet see losses on that scale. Bloomberg Economics (June 2026) found employment in AI-exposed private sectors had risen, and that “the results push against the idea of rapid, large-scale job displacement.” The ONS business survey finds only about 6–7% of AI-using firms report lower headcount. Some firms are now saying it openly. Standard Chartered’s chief executive talked of replacing “lower-value human capital”. But Centrica denied that AI drove its 1,300 call-centre cuts.

Verdict: too early to tell. Firm surveys keep pointing to it. Aggregate data do not yet confirm it.

Sadiq Khan: “a weapon of mass destruction of jobs”

In his Mansion House speech in mid-January, the Mayor warned that without controls AI could become “a weapon of mass destruction of jobs” and “usher in a new era of mass unemployment, accelerated inequality”. London’s unemployment rate did peak at 7.6% in November–January (as reported), but it has since eased to 6.8% (May–July 2026). LinkedIn attributes London’s 32% fall in hiring since 2019 mainly to geopolitical and trade uncertainty. The GLA’s own analysis of exposure (April 2026) remains a forecast of potential task change, not of job losses.

Verdict: not borne out so far. London is the weakest region, but nothing yet resembles mass unemployment, and AI’s share of the weakness is unproven.

Andrew Bailey: displacement, not mass unemployment

On BBC Radio 4 in December 2025, the Governor compared AI with the Industrial Revolution, which “didn’t cause mass unemployment, but it did displace people from jobs”. He asked: “what is it doing to the pipeline of people?” The Bank’s 2026 evidence matches that framing closely:

  • Firms on its Decision Maker Panel expect AI to cut employment by about 0.4% a year and to raise productivity by about 0.9% a year (July 2026).

  • A staff blog post on Bank Underground (August 2026) finds vacancies down 15% over three years in the most exposed occupations, against 6% in the least exposed, while warning that confident attribution remains premature.

  • The Agents report “limited evidence of broad AI-driven reductions in employment”, but say AI is “influencing role design and replacement hiring”.

Verdict: borne out so far. This was the best-calibrated public expectation of the period.

CIPD: one in six employers expect AI to shrink headcount

In November 2025, 17% of employers told CIPD they expected AI to reduce headcount within 12 months. The figure was 26% among large private firms, and 62% of those expecting cuts named clerical, junior managerial, professional or administrative roles. The 12-month window closes next month. CIPD has not repeated the question in 2026, so a like-for-like check is impossible. The closest outcome measure, from ONS, finds about 6% of firms using AI for operations reporting a fall in headcount. The questions differ (expectation against outcome, and different samples), but they point the same way: some firms are cutting, far fewer than expected to.

Verdict: too early, leaning not borne out at scale. The roles CIPD’s respondents named are the same ones where vacancies are falling fastest.

The graduate-jobs cliff

Going into 2026, the ISE reported that graduate tech roles fell 46% in 2025 and projected a further 53% fall for 2026 (as reported). In January, ISE projected a 7% drop in student vacancies overall. Adzuna’s July 2026 figure, a 45.6% fall in graduate ads to 8,383, seemed to confirm a cliff. But Jisc’s Charlie Ball argues that Adzuna captures only 5–10% of graduate jobs, and calls the market “pretty subdued” but “probably not that different to last summer”. Indeed’s graduate postings were down about 7%. HESA’s latest outcomes show graduate unemployment up from 6% to 7%, which Jisc calls “a cooling not a collapse”. Entry-level hiring in LinkedIn’s data is falling in step with hiring overall.

Verdict: partly borne out. The market is the weakest since 2020 and concentrated in exposed occupations, but it is not a cliff. The ISE’s 2026 survey on 14 October will be the first large-employer count of this year’s intake.

OBR: forecasts and assumptions

The OBR’s March 2026 outlook expected unemployment to rise from 4¾% in 2025 to a 5⅓% peak in 2026, and CPI inflation of 2.3% in 2026, returning to the 2% target late in the year. It treated AI as part of underlying productivity growth rather than quantifying it separately. A labour-displacement scenario, in which equilibrium unemployment rises to 5.5%, was presented only as a risk.

So far unemployment has come in lower than forecast, at 4.9% since February–April. Inflation has come in much higher: 3.1% in August, with the Bank expecting “around 3¾%” in Q4. Whether AI belongs in the central forecast is still untested. The OBR’s July fiscal risks report added a warning that AI could shift income “from (more highly taxed) labour to (lower-taxed) profits”.

Verdict: not borne out on unemployment and inflation; too early on AI.

The January edition’s own expectations

The January edition expected widespread task reshaping and churn, not mass unemployment, with entry-level “learning tasks” hit first. That has largely held. Its “grinding slowdown, not collapse” reading of the labour market was also right. It set a warning threshold: AI-using firms reporting headcount falls rising “materially above” 4–5%. That threshold has not been crossed decisively; the figure is now about 6–7% of the relevant groups. It expected unemployment to keep drifting up. Instead it peaked and eased. Like the Bank itself in February, which said rates were “likely to be reduced further”, it did not anticipate that the inflation outlook would turn.

What was missed entirely

A change of government and the end of DSIT

Keir Starmer resigned on 22 June 2026. Andy Burnham became Prime Minister on 20 July, and John Healey replaced Rachel Reeves as Chancellor on 21 July. The same day, DSIT was abolished. AI strategy moved to the Cabinet Office, science and innovation went to a renamed business department, and digital went to DCMS. Kanishka Narayan became Minister of State for AI, attending Cabinet. The 10 million AI-skills target set in January stands. But DSIT-badged research, such as the June entry-level hiring snapshot with LinkedIn, now sits in a reorganised Whitehall, and its continuity is not yet clear. The Autumn Budget on 28 October will be Healey’s first.

An energy-driven inflation rebound

Conflict in the Middle East, and developments involving Ukraine and Russia, pushed energy prices sharply higher. According to the September MPC minutes, Brent crude and UK wholesale gas spot prices rose 36% and 78% respectively from the period before the July forecast. Motor fuel prices were up 23% on the year in August, and CPI rose from 2.9% in July to 3.1%. The Bank now expects inflation slightly above 4% in early 2027. For workers, that threatens a return to falling real pay. Regular pay growth has slowed to 3.5%.

A hawkish MPC

In February the MPC held Bank Rate at 3.75% by 5–4, with four members wanting a cut. By July and September it was holding by 6–3, with Megan Greene, Catherine Mann and Huw Pill voting to raise rates to 4%. The rate cuts that were expected to revive hiring in 2026 did not arrive. This matters for the AI debate. Prolonged tight policy keeps hiring weak for cyclical reasons, which makes any AI effect even harder to identify, and easier to blame wrongly.

The EU AI Act delay

The Digital Omnibus on AI came into force on 27 July 2026. It moved the high-risk obligations for employment and recruitment systems under Annex III from 2 August 2026 to 2 December 2027. A delay had been proposed in late 2025, but many UK employers selling into the EU were still preparing for August 2026. Transparency obligations under Article 50 applied from 2 August as planned.

Other surprises

  • Unemployment fell while payrolls kept shrinking. The LFS shows 111,000 more employees on the year, while HMRC payroll data show 145,000 fewer. That divergence complicates any reading of the data.

  • Older workers held on and the young were spared further heavy losses. Payrolled employees aged 65 and over rose by 62,000 in the year to August 2026, while under-25s fell by only 15,000. The NEET total, which had passed one million earlier in the year, fell back to 981,000 by April–June.

  • The official institutions began measuring AI. The Bank’s AI questions in the Decision Maker Panel and its staff analysis of vacancies by exposure, the OBR’s displacement and labour-share scenarios, and the DSIT/LinkedIn entry-level snapshot did not exist as UK evidence in January.

  • The data regulator itself was abolished. On 30 September 2026 the office of Information Commissioner was abolished and the ICO was replaced by the Information Commission, which inherits the still-unfinished guidance on automated decisions in hiring.

  • Copyright policy reversed. In March the government said a broad copyright exception with opt-out “is no longer the government’s preferred way forward”.

Bottom line

The record of early 2026 rewards caution about attribution. AI is visibly changing which junior roles get advertised and filled, and large firms increasingly say so. The national labour market, though, was shaped mainly by politics, energy prices and interest rates. Next year’s forecasts should keep the same discipline: separate what is measured from what firms expect, and wait for the hiring recovery to show whether the junior roles AI is squeezing come back.

Sources

From AI and Jobs: UK, October 2026