BIS 2026: AI Cuts Task Time 20-50% But Hiring Slows

The world's most influential central bank body just put a number on what AI is doing to your job — and it isn't reassuring. According to the Bank for International Settlements, AI tools are already cutting the time it takes to do many work tasks by 20% to 50%. That's not a Silicon Valley pitch deck. It's the conclusion of the BIS Annual Economic Report 2026, published on 28 June 2026 by the institution that coordinates the world's central banks.

Here's the part that should make every knowledge worker pause: in the US sectors most exposed to AI, productivity is climbing — but so is a quiet drag on hiring. The BIS found that these same sectors have seen higher productivity gains partly at the expense of lower employment growth. In plain terms, companies are producing more with fewer new hires. And in earnings calls, firms are increasingly saying the quiet part out loud.

The 20-50% number, and who feels it first

The BIS didn't invent the headline efficiency figure — it aggregated it. Across task-level studies, the report says gains "often to the tune of between 20 and 50% in time savings" show up repeatedly in coding, consulting tasks, clerical work, writing, software development, and customer support.

Notice what those categories have in common. They are cognitive, desk-based, and language-heavy — exactly the kind of work that used to feel safe from automation. The BIS makes this explicit: unlike past waves of technology that automated muscle and routine, AI competes directly with human cognitive abilities [Fact].

If you write code for a living, this is your world already. Our data on software developers and computer programmers shows some of the highest AI exposure of any occupation, and the BIS coding evidence lines up with that. The same pressure lands on customer service representatives, management analysts, writers and authors, and bookkeeping, accounting, and auditing clerks — every role where the core task is processing information rather than moving physical objects.

Productivity up, hiring down: the trade-off the BIS names

The most important line in the report for workers isn't the efficiency number. It's the correlation. The BIS observed that "US sectors with higher exposure to AI have also seen higher productivity gains, partly at the expense of lower employment growth" [Fact].

That is the displacement mechanism in one sentence. When a task that took an hour now takes thirty minutes, a firm has a choice: do twice as much work, or do the same work with fewer people. The BIS says firms are increasingly signalling the second option. In earnings calls, a growing number of companies are discussing an "actual or potential reduction in labour input due to AI or automation," and "more firms are acknowledging potential productivity gains from AI, signalling their intent to automate an increasing share of production processes" [Fact].

Before you panic, the BIS adds a crucial caveat: "to date, such disruptive labour displacements have yet to occur at scale" [Fact]. The intent is visible in boardrooms; the mass layoffs are not yet visible in the aggregate data. But the report warns plainly that "labour displacement could intensify" as more capable AI tools spread [Claim].

What makes this earnings-call signal worth taking seriously is that it is a leading indicator, not a lagging one. Companies rarely announce automation plans before they have tested the tools internally — so when the number of firms citing AI-driven labour reduction rises quarter after quarter, it is a forecast of hiring decisions that have not yet hit payroll data. The BIS reads this as a shift in intent that precedes the shift in headcount. For a worker, the practical implication is uncomfortable but useful: the moment to build AI-adjacent skills is while the aggregate data still looks calm, not after the sector-level squeeze becomes obvious in the unemployment numbers. The report's own framing is that whether "AI advances create new jobs or expand demand for existing ones" remains genuinely uncertain [Claim] — which means the outcome is not fixed, and individual positioning still matters.

The counterintuitive twist: AI might not make the economy much richer — yet

Here's the finding almost no headline captured. Despite those eye-catching 20-50% task savings, the BIS estimate for the boost to _aggregate_ productivity over a long horizon is conservative — less than 1% [Estimate]. How can both be true?

Because saving time on a task is not the same as transforming an economy. Integration is slow, adoption is uneven, and a lot of the freed-up time gets absorbed by coordination, oversight, and simply reworking AI output. This is the gap between the demo and the balance sheet — and it is exactly why the "AI will 10x GDP" story and the "AI is coming for your job" story can both be partly right at the same time. The threat to specific jobs is concentrated and real; the economy-wide windfall is diffuse and slow.

The bubble question hanging over all of it

The BIS spent much of the chapter on a different worry: the money. The five largest hyperscalers are set to spend over a trillion US dollars on AI capital expenditure across 2025-2026, and the report notes this spending is "outpacing earnings and the free cash flow" of those firms [Fact].

The BIS reached for historical parallels that should give anyone pause — canal mania in the 1830s, British railways in the 1840s, electrification in the late 1920s, and the dotcom boom of the late 1990s. Each featured "a genuine technological breakthrough that attracted capital in excess of what commercial returns could ultimately justify," and each ended in recession [Fact]. In an adverse scenario, the BIS models the sector's net economic surplus turning negative [Estimate].

Why should a worker care about a possible investment bubble? Because your job security is now tied to it. AI investment has been "an important driver of growth surprises in the United States." If that spending pulls back sharply, the growth it propped up pulls back with it — and the layoffs that AI-driven productivity was already enabling could arrive alongside a broader downturn. There are also physical limits building up: the BIS flags "growing bottlenecks in electricity, advanced semiconductors and grid equipment," which is good news for electrical engineers and power plant operators, whose skills the AI build-out desperately needs.

What this means for you

The BIS is a cautious institution, and its policy advice is telling: invest in skills and capabilities to leverage AI across sectors, build the "digital backbone and energy infrastructure," and use competition policy so the gains "are diffused and shared widely." Translated for an individual career, that means three things.

First, if your work is cognitive and information-based, assume the 20-50% efficiency gain is coming to your role — and position yourself as the person who directs the AI rather than the person the AI replaces. Second, watch your sector's hiring trend, not just your own desk; the BIS says the employment squeeze shows up at the sector level first. Third, the infrastructure side of the AI boom — power, chips, grids — is hiring, and those jobs are far harder to automate than the ones creating them.

The BIS message is neither doom nor hype. It's that the productivity gains are real, the displacement risk is real, the economy-wide payoff is slower than advertised, and the whole thing rests on an investment boom whose sustainability is genuinely uncertain. Progress and peril, as the report's own chapter title puts it. Knowing which is which is now part of your job.

Sources

• Bank for International Settlements, _Annual Economic Report 2026, Chapter I: "Progress and peril"_ (28 June 2026): https://www.bis.org/publ/arpdf/ar2026e1.htm
• Full report PDF: https://www.bis.org/publ/arpdf/ar2026e1.pdf
• Statistical annex: https://www.bis.org/statistics/ar2026stats/ar2026e1_stats.xlsx

This analysis was produced with AI assistance. All figures and quotations are drawn directly from the BIS Annual Economic Report 2026 (Chapter I) and cross-checked against the published source. Occupation-level exposure and risk data referenced above come from AI Changing Work's own analysis of ONET task structures.\*