Something odd happens when you read what European companies actually do with artificial intelligence. Eurostat asked firms with ten or more employees what they used AI for in 2025. The most common answer, given by 11.8% of them, was analysing written language. Next came generating pictures, video and audio at 9.5%. Then generating written or spoken language at 8.8%. Then turning speech into text at 7.2%.

Every item on that list is a task that used to take an afternoon and now takes a minute. That is worth something. It is not the thing that decides whether a company is still here in three years.

That distance is the subject here. The forecasts for this technology are enormous, and the average firm's response to them is a better email. The space between those two facts is where businesses are being quietly repriced.

The forecasts are not small

The World Economic Forum projects 22% of jobs in disruption by 2030, with 170 million new roles created and 92 million displaced. The IMF puts roughly 40% of jobs worldwide as exposed to AI. Goldman Sachs counted 300 million.

92 million displaced across five years is about 18 million a year. Each one is a cost line being removed from somebody's business. The same force either removes a cost line from yours or hands you the margin a competitor just gave up.

Most firms are not in the race at all

Eurostat measured 20.0% of EU enterprises with ten or more employees using AI in 2025, up from 13.5% in 2024. Eight in ten use none of it.

Estonia runs ahead of that average. The OECD's 2026 survey of the Estonian economy put local firms at 23.4%, up from around 14% the year before. Still, three in four Estonian companies use none of it.

So the market splits three ways.

A small group is rebuilding how it operates. A larger group is typing prompts into a chat window and calling it transformation. And most of the market is doing nothing at all.

The tool is not the multiplier

McKinsey surveyed the same landscape in November 2025 and found the pattern behind that list.

Nearly two thirds of respondents said their organisations had not begun scaling AI across the enterprise. And only 39% reported an impact on EBIT at company level. Which means 61% were spending money and seeing nothing on the bottom line.

The reasons sit in the same survey.

80% set efficiency as the objective. The companies reporting the most value set growth and innovation as additional objectives. And the ones pulling ahead were redesigning workflows rather than buying tools.

A tool makes one task faster. A workflow changes how many customers are possible at all. Those are not the same size of prize, and they never were.

60% report nothing

BCG surveyed 1,250 senior executives for a report published in September 2025, and sorted them into three groups.

5% qualify as future-built. Another 35% are scaling and starting to see value. The remaining 60% report minimal gains in revenue or cost.

The leaders expect twice the revenue increase and 40% greater cost reductions than the laggards, in the areas where they apply AI. And 70% of AI's potential value sits in three functions: research and development, innovation, and digital marketing.

Read that last line twice if you sell anything for a living. The value is not in the back office. It is in the part of the business that finds customers.

What a multiplier looks like next to a tool

The difference is easier to see in concrete terms, from work we have shipped.

A tool writes your follow-up email faster. A system makes sure no follow-up is ever missed. For Serve Funding, a US lending platform, that second version meant matching each deal to qualified lenders on more than 160 criteria in real time. Work that used to take days now takes seconds, and millions in loan volume move through the platform every month.

Same technology. Different question. One asks how to do this task faster. The other asks how many of these we can do at all.

The first saves minutes. The second changes the size of the business.

Why the fluff work feels like progress

Because it is measurable, and it is safe.

A summary appears in three seconds. You can show it to a colleague. Nothing breaks. Nobody has to change how they work, and nobody has to agree about anything.

Rebuilding a workflow is the opposite. It is slow at the start. It drags out the parts of the business nobody wants to look at, like the fact that half the inbound leads never get a second call. It needs someone to own a number and be judged on it.

So most companies take the safe version. Which is exactly why the gap widens instead of closing.

The cost of waiting is not zero

Waiting feels neutral. It is not, because the other side of the market keeps moving.

A competitor growing at twice your rate does not stay one step ahead. The gap compounds. Next year they have more cash to spend on the same customers you want. The year after, they can underprice you and still hold their margin.

Kristalina Georgieva, the IMF's managing director, set out the requirement plainly in October. To get the benefit, she wrote, businesses must rethink, reinvent and re-engineer their product lines, processes and workflows.

She added a line worth keeping: "The benefits of the internet did not come with the ability to throw up a home page."

The internet did not pay anyone for owning a website. It paid the companies that rebuilt retail, media and logistics around it. The same distinction is being drawn right now, and the terms are being set in quarters, not decades.

The 14-day version

We built Nordspike around the second question.

We install a marketing system for you: the creative, the ads, the tracking, and the loop that turns a stranger into a booked call. Not a tool you have to learn. A department that runs whether or not you have time to look at it.

It goes live in 14 days. If it is not live, you do not pay.

The companies that will be fine either way are already doing this. Everyone else is choosing, every month, between the safe version and the version that changes the number.

Back to the list

That list of most common AI uses is still there. Analysing written language. Generating images. Generating text. Turning speech into text.

None of it is wrong. It is just small.

The forecast is 92 million jobs displaced by 2030 and 300 million exposed to the change. The companies that come out of that shakeout will not be the ones that wrote the best emails. They will be the ones that decided a workflow was worth rebuilding while everybody else was saving six minutes.