Most AI adoption numbers floating around come from paid internet surveys where it's unclear whether respondents are actually executives or providing informed answers. The McKinsey 2025 report, for example, estimated 88% AI adoption. The Census BTOS survey pegged it at around 9%. That's a 10x difference, and it highlights how much survey design matters.

The NBER study solved this three ways:

• Verified executives: Respondents were initially recruited by phone to confirm their position and company, then moved into the survey panel. No paid anonymous clicks.
• Institutional backing: The surveys were sponsored by the Federal Reserve Bank of Atlanta, the Bank of England, and the Deutsche Bundesbank - which meant senior leaders actually responded.
• Standardized questions: All four countries used identical questions timed to run in the same three-month window (November 2025 - January 2026).

The result is data that closely tracks GDP growth and employment figures going back a decade - meaning when these executives say something about their businesses, their answers line up with what actually happens.

Across the four countries, 69% of businesses currently use some AI technology. The US leads at 78%, followed by the UK (71%), Germany (65%), and Australia (59%).

The most common use case is text generation through large language models, used by 41% of firms. Data processing with machine learning and visual content creation each sit around 28-29%. Robotics (9%) and autonomous vehicles (3%) remain niche.

The study also found clear patterns in who's adopting:

• Larger, more productive, and higher-paying firms are significantly more likely to use AI
• Firms with younger leadership adopt AI at higher rates
• Older firms and firms with older directors are less likely to be using AI
• Finance, insurance, and professional services sectors lead adoption in the UK

Looking ahead, 75% of businesses expect to be using some AI technology within three years. The technology firms expect to use most going forward is data processing with machine learning - a shift from the current LLM-heavy pattern that suggests businesses are planning to move beyond text generation into data-driven operations.

Here's a number that surprised me: the average senior executive - we're talking CEOs, CFOs, and finance directors - uses AI for about 1.5 hours per week. A quarter of these leaders don't use it at all.

The breakdown looks like this: 28% of executives report zero AI use during a typical work week. The largest group (41%) uses AI for up to one hour per week. About 24% use it between 1-5 hours, and just 7% spend more than 5 hours per week with AI tools.

There's been real movement here, though. In the UK, the share of executives who don't use AI at all dropped from 45% in early 2025 to 25% by late 2025 - a 50% increase in average weekly AI use in less than a year. CEOs are also more likely to use AI than CFOs and other senior executives, which makes sense given the broader strategic questions AI addresses versus the more structured work of financial management.

The age factor is consistent: younger leaders use AI more. Companies with younger directors show higher personal AI use by executives, and younger firms show the same pattern.

This is the finding that should cause some serious reflection. Across all four countries, 89% of firms report that AI has had no measurable effect on their sales per employee over the past three years. The average productivity boost is just 0.29%.

Compare that to individual task-level studies that find 15-30% productivity gains in specific settings. The gap between what AI can do in a controlled experiment and what it's actually delivering for businesses at scale is enormous.

There's some variation by sector. In the UK, information and communications firms reported the largest productivity gains at +0.8%, followed by professional and scientific firms at +0.5%. Construction and wholesale/retail reported near-zero impact.

The forward-looking numbers tell a different story. Executives expect AI to boost productivity by 1.4% over the next three years. US firms are the most optimistic at +2.3%, followed by UK firms at +1.9%. German and Australian executives are more conservative at around +0.9% each.

If those predictions hold, we'd be looking at a meaningful reversal of the productivity growth slowdown that's plagued advanced economies for years. But given the track record of executives predicting AI impacts versus what actually materializes, some healthy skepticism seems appropriate.

The employment picture mirrors productivity. Over 90% of firms across all four countries say AI has had zero effect on headcount over the past three years. The average impact rounds to essentially zero.

But ask executives what they expect over the next three years, and the picture shifts. Firms predict AI will reduce employment by 0.7% on average - the equivalent of roughly 1.75 million fewer jobs across the four countries surveyed. The largest expected cuts are in the UK (-1.4%) and US (-1.2%).

A key detail: about two-thirds of expected employment reductions would come from hiring fewer new workers rather than laying off existing staff. That distinction matters for SMBs thinking about their workforce planning.

By sector, the accommodation/food and wholesale/retail industries expect the largest employment reductions - around -1.8% to -2% over three years in the UK. Construction, health, and recreational services expect smaller impacts. This lines up with what you'd expect: jobs with more routine, process-driven tasks are more exposed to AI-driven automation than those requiring physical presence and hands-on work.

The study also surveyed about 3,000 US employees through a separate survey, and the contrast with executive views is striking.

Employees use AI at roughly the same rate as executives - about 1.8 hours per week. But their expectations about AI's future impact on employment are completely opposite. Employees predict AI will increase employment by 0.5% over the next three years. Executives predict a decrease of 0.7% across all firms (1.2% for US firms specifically).

On productivity, the gap runs in the other direction. Employees expect a 0.9% productivity gain, while executives predict 1.4% (2.3% for US executives). So executives are more optimistic about productivity gains but also more likely to see those gains as coming at the cost of headcount.

This disconnect isn't academic. It has real implications for how businesses handle AI communication internally. If your team thinks AI will create opportunities while you're planning to reduce hiring, that's a recipe for trust problems down the road. Having an honest, early conversation about AI's role in your business - both the opportunities and the constraints - is worth the discomfort.

If you're running a construction company, law firm, or oil and gas operation in the Houston metro, this data should inform your AI strategy in several ways.

Don't panic about falling behind on AI. 69% adoption sounds high, but when 89% of those businesses report zero productivity impact, the real question isn't "are we using AI?" It's "are we using AI in ways that generate measurable value?" We see this pattern at least twice a month with Houston businesses - they've bought the tools, paid for the subscriptions, and have nothing to show for it because nobody thought about the integration piece.

Focus on specific use cases, not general adoption. The firms seeing actual results aren't the ones that gave everyone a ChatGPT subscription and called it a day. They're the ones who identified specific workflows where AI tools solve concrete problems: processing accounting data, drafting standard documents for engineering firms, or automating repetitive reporting tasks.

The real opportunity is in data processing, not text generation. While LLM text generation is the most common AI use today, firms expect data processing using machine learning to be the dominant use case going forward. For businesses in manufacturing, energy services, and wealth management, that shift toward data-driven AI is where the real productivity gains will come from.

Get your cybersecurity house in order before expanding AI use. Every AI tool your team uses is another potential data exposure point. We've already seen cases where employees are pasting client data into public AI tools without realizing the security implications. An AI usage policy and proper security guardrails aren't optional - they're a prerequisite.