Bank of England: AI added 0.5% productivity. The rest is hope

You know the slide. "AI makes you 40% more productive." It's in practically every vendor deck. Now a central bank has counted. For an entire economy.

The number is smaller. Much smaller.

What the Bank of England says

The November 5 Monetary Policy Report (Box D) includes these points:

• Firms in the Decision Maker Panel, the Bank's regular business survey, estimate that AI has raised their productivity by around 0.5% over the past three years. That's roughly 0.1 to 0.2 percentage points of productivity growth a year since early 2024.
• In aggregate, the effect on employment so far is close to zero.
• Over the next three years, firms expect AI to add about 0.9 percentage points a year to productivity growth. Roughly half via slower employment growth, half via faster output growth.
• Novel AI-related activity accounts for just under a fifth of UK GDP growth since the start of 2026. But mostly from developing and implementing AI, not from broad productivity gains.
• Most of the recent productivity pickup looks cyclical, according to the Bank: post-pandemic adjustment and higher labour costs rather than AI.

The shovel sellers are getting paid

Here's what I find interesting: AI-driven growth is real. It just shows up somewhere other than the slides. Those building AI, pouring concrete for data centers and implementing systems are growing. Those using AI see little in the numbers so far. Classic gold rush.

One detail people skip: the 0.5% is the firms' own estimate. Not a measurement. When even the self-assessment is that low, that tells you something.

The catch: productivity here means "hire less"

The second-order point is in the forecast. Half the expected gain comes from slower employment growth. That doesn't mean mass layoffs. It means roles that don't get backfilled. And junior jobs that never get posted.

For marketing teams, that's the real question in my view. If AI does the entry-level work, where do your senior people come from in five years? Optimize only for cost and you save a salary today, then pay for it with an empty bench tomorrow.

What I'd do now

• Measure before, don't estimate after. Pick one number per AI use case: hours per campaign, time to approval, cost per asset. Otherwise you end up with a self-assessment, just like the survey.
• Set an output goal, not just a cost goal. Don't only ask "how many hours do we save?" Ask "what do we do with them?" More tests, more channels, better creative.
• Plan your junior pipeline on purpose. A junior role that works with AI is an investment in your 2030 team.

Bottom line: So far, AI mostly makes the people selling it rich. Whether it makes you more productive, you'll only know if you measure.