The AI Jobs Paradox: 69% of Firms Use AI, But 90% Report Zero Employment Impact

Sixty-nine percent. That is the share of businesses actively using AI right now, according to the largest cross-country executive survey on the topic ever conducted. [Fact] (NBER Working Paper 34836)

Now here is the number that should make you pause: 90% of those same executives say AI has had zero measurable impact on their employment levels or productivity over the past three years. [Fact] (NBER w34836)

Two-thirds of firms have adopted the technology. Almost none of them can point to concrete workforce changes because of it. So what is actually going on?

The Biggest AI Employment Survey Yet

The study behind these numbers — NBER Working Paper 34836 — is not a think-tank opinion piece or a tech company's marketing survey. It is a rigorous academic paper by thirteen researchers spanning Stanford, the University of Chicago, the Bank of England, and the Atlanta Fed. [Fact] They surveyed approximately 6,000 senior business executives across the United States, the United Kingdom, Germany, and Australia between 2023 and 2025. [Fact]

What makes this study uniquely valuable is that it asks the people who actually make hiring and investment decisions — not AI researchers, not pundits, not employees worried about their jobs. These are CFOs, CEOs, and division heads with direct line of sight to both the technology adoption and the headcount numbers.

And their answers paint a picture that contradicts almost every AI employment headline you have read in the past year.

The Adoption-Impact Disconnect

The adoption numbers are impressive. 69% of firms are actively using AI, and the executives themselves spend an average of 1.5 hours per week working directly with AI tools. [Fact] (NBER w34836) This is not fringe experimentation — it is mainstream business practice across four major economies.

But when asked about actual results, the story flatlines. 90% of executives report zero impact on both employment and productivity from AI adoption over the last three years. [Fact] Not small impact. Not marginal impact. Zero.

This is not what the narrative has been telling us. Goldman Sachs estimates that AI is eliminating 16,000 US jobs per month. Consulting firms have been publishing dramatic automation exposure scores. Our own data shows significant AI exposure across hundreds of occupations — data entry keyers at 82% exposure, administrative assistants at 58%, customer service representatives at 65%. [Fact]

So how do you square widespread adoption and high theoretical exposure with executives saying they see nothing happening?

The answer may be timing. AI exposure measures what could be automated. Executive surveys measure what has been automated. The gap between those two numbers is the gap between potential and reality — and right now, that gap is enormous.

What Executives Think Comes Next

Here is where it gets interesting. The same executives who report zero past impact are considerably less sanguine about the future.

Over the next three years, they predict AI will reduce employment at their firms by -0.7%, boost productivity by +1.4%, and increase output by +0.8%. [Fact] (NBER w34836) Those are modest numbers individually, but applied across four economies they represent millions of affected workers.

The productivity prediction is particularly noteworthy. A +1.4% productivity gain would be a meaningful acceleration over the anemic productivity growth that has plagued advanced economies for over a decade. [Estimate] If executives are right, AI's main economic effect will not be mass unemployment — it will be getting more output from fewer workers.

This aligns with what the Brookings Institution's pro-worker AI framework has been arguing: the real question is not whether AI replaces jobs, but whether the productivity gains flow to workers or only to firm owners.

The Expectations Gap: Bosses vs. Workers

Perhaps the most striking finding is the disconnect between what executives expect and what employees expect.

Executives predict -0.7% employment at their firms over three years. [Fact] But when workers at those same firms are surveyed, they expect +0.5% employment growth. [Fact] (NBER w34836)

That is a 1.2 percentage point gap between what bosses are planning and what workers are anticipating. [Estimate] Employers are quietly budgeting for fewer people. Employees are expecting business as usual — or even growth.

This gap matters enormously for anyone in an AI-exposed occupation. If you are a software developer, an accountant, or a financial manager, your employer may already be planning a future with fewer people in your role, even as you expect your team to stay the same size or grow.

The study found that this planning gap is especially pronounced at younger, more productive firms — exactly the kind of growth companies where talented workers tend to gravitate. [Fact] These firms adopt AI faster, use it more intensively, and anticipate larger workforce adjustments. [Fact]

The Paradox Explained

How do we reconcile 69% adoption with 90% zero impact? Three factors likely explain the disconnect:

First, adoption does not equal transformation. Most firms are using AI for incremental productivity tools — drafting emails, summarizing documents, basic data analysis. These are useful but do not fundamentally restructure workflows or eliminate positions. [Claim] The gap between "using ChatGPT" and "redesigning your business processes around AI" is vast.

Second, organizations change slowly. Even when the technology exists to automate a task, firms need to restructure teams, update processes, retrain managers, and navigate internal politics before headcount actually changes. This institutional friction acts as a buffer — one that may be eroding as AI capabilities improve.

Third, we may be in the calm before the storm. The executives themselves are signaling this. Zero impact in the past three years, but meaningful reductions in the next three. The implication is that firms have been in experimentation mode and are now moving toward implementation mode.

What This Means for Your Career

If you are reading this and wondering what it means for your specific occupation, here is the honest read:

The short-term news is better than you feared. 90% of firms reporting zero impact means that for most workers right now, AI is changing how they work more than whether they work. If you are in customer service or data entry, your day-to-day may involve more AI tools, but your position likely still exists.

The medium-term news demands preparation. A predicted -0.7% employment decline sounds small until you realize that executives — the people who sign the hiring requisitions — are the ones saying it. And the workers expecting growth are going to be caught off-guard.

The biggest risk is complacency. The 90% zero-impact figure is not reassuring — it is a lagging indicator. By the time impact shows up in executive surveys, the strategic decisions have already been made. The time to position yourself in an AI-augmented role is before the reductions materialize, not after.

If you want to see exactly how AI affects your specific occupation, check the detailed data pages for roles like software developers, administrative assistants, accountants, or financial analysts.

The AI jobs paradox will eventually resolve. The question is whether you will be on the right side of that resolution.

---

This analysis was produced with AI assistance. All statistics are sourced from the cited NBER Working Paper 34836 and cross-referenced with existing research. For detailed AI impact data on any of the 1,000+ occupations covered, visit the individual occupation pages.

Sources

1. Yotzov, I., Barrero, J.M., Bloom, N., et al. "Firm Data on AI" (NBER Working Paper 34836, February 2026, revised March 2026). nber.org
2. Goldman Sachs, "How will AI affect the US labor market" (2026). Referenced analysis
3. Acemoglu, Autor & Johnson, "Pro-Worker AI" (Brookings, 2026). Referenced analysis

Update History

• 2026-04-08: Initial publication based on NBER Working Paper 34836.