One Company, Three Versions of Itself: What AI Readiness Really Looks Like for a Small Business

There is a tempting way to talk about AI and small business that goes something like this: the tools are cheap, the gains are obvious, so just start using them. It is tidy, it is optimistic, and it is mostly wrong. The evidence from the last two years tells a more interesting story — one where using AI and getting value from it turn out to be almost entirely different things.

To make that concrete, let's invent a company and run it three ways.

Meet Larke Home, a fictional but very ordinary European homeware brand. Twenty-two people. An online shop, one showroom, a small workshop that finishes a few signature pieces by hand. Healthy but not spectacular margins. The kind of business that exists in every market in Europe and rarely makes the news.

Now imagine three versions of Larke Home, competing in the same city, selling the same sofa throws and stoneware, facing the same customers — but each running on a completely different relationship with AI. Call them Larke A, Larke B, and Larke C. They look identical from the outside. Inside, they are diverging fast.

Larke A: the company that hasn't really started

Larke A runs the way most small companies still run. The founder answers customer emails between meetings, usually a day late. Marketing is "post when we remember." Stock decisions are made on gut feel and last year's spreadsheet. Someone once pasted a product description into ChatGPT, was mildly impressed, and never went back.

This is not an exotic failure. It is the statistical baseline. According to Eurostat, just 17% of small EU enterprises used any AI technology in 2025, compared with 30% of medium-sized firms and 55% of large ones — and even those headline numbers count a company as an "adopter" if it uses a single tool, however lightly. The gap between countries is enormous, too: Eurostat puts AI use among enterprises at 42% in Denmark but only 5% in Romania. Larke A is simply sitting on the wrong side of that line.

The cost of staying there is not dramatic. That is exactly what makes it dangerous. Nothing breaks. The founder is just permanently the bottleneck, the marketing stays generic, and the company quietly loses the customers it never realized it was slow to answer. Larke A doesn't have an AI problem. It has an attention problem, and AI is one of the things it isn't paying attention to.

Larke B: the company that's "using AI" — and that's the trap

Larke B looks far more modern, and its founder will happily tell you they're "doing AI." The marketing person drafts emails with a chatbot. A junior designer makes social graphics in an AI image tool. There's a support widget on the website that handles simple questions. Three or four subscriptions, billed monthly, scattered across the team.

This feels like progress, and in a narrow sense it is. The Thryv 2025 small-business survey found that AI use among SMBs jumped from 39% to 55% in a single year, with the steepest growth in companies of 10–100 employees — and content marketing as the single most popular use case. Larke B is squarely in that wave.

Here is the uncomfortable part. Larke B is also a near-perfect illustration of what McKinsey, in its State of AI in 2025 survey of nearly 2,000 respondents across 105 countries, identified as the central problem of the moment: adoption is almost universal — 88% of organizations now use AI in at least one function — but roughly two-thirds have not moved beyond experimentation and pilots, and only about 39% can link any measurable bottom-line impact to AI at all (for most of them, under 5%). McKinsey's blunt framing is that everyone has tried AI; very few have rebuilt how they work around it.

That is Larke B exactly. The AI is bolted onto the existing process. The chatbot drafts the same email the marketer would have written, a little faster. The support widget deflects a few tickets but routes the hard ones into the same overloaded inbox. Nobody measures whether any of it moved revenue, because nobody owns the question. There's no policy on what customer data goes into which tool — which, in an EU context, is not a small oversight. Larke B has spent money and motion to make its old workflow slightly quicker. It has bought efficiency in pockets and called it transformation.

The OECD's December 2025 discussion paper for the G7 on SME AI adoption captures why this happens. Surveyed small and medium firms most often report that generative AI's main benefit is improved employee performance — useful, but a long way from new products, higher revenue, or competing with larger rivals, which far fewer report. And around half of SMEs say their staff simply lack the skills to use the tools well. Larke B has the tools. It does not yet have the capability.

Larke C: the company that treats AI as a way of working

Larke C buys roughly the same software as Larke B. From the outside you could not tell them apart on a list of subscriptions. The difference is everything around the tools.

At Larke C, someone owns AI as part of the business strategy — not as an IT ticket. Before any tool went in, they asked a sharper question than "what can AI do?" They asked "where does a day of delay, a generic message, or a bad forecast actually cost us money?" — and pointed AI at those specific places.

So the customer-service assistant isn't a deflection widget; it's connected to order and returns data, resolves the routine cases end to end, and hands the genuinely tricky ones to a human with the context already summarized. Customer service, incidentally, is where most small firms find their first real return — Talkdesk found that roughly 51% of AI-using US small businesses start there, precisely because the payback is immediate and measurable.

The marketing engine at Larke C doesn't just draft faster; it segments the customer base and personalizes what each group sees, which is why McKinsey finds revenue gains from AI cluster in marketing and sales, strategy, and product development rather than in raw cost-cutting. Demand forecasting feeds the workshop and the stock orders, so the company stops guessing at the spring run. And critically, all of it is measured — they know which changes earned their keep and which didn't.

Two things make Larke C rare rather than impossible. First, leadership actually drives it: McKinsey's high performers — the roughly 6% of organizations capturing significant value — are about three times more likely to have senior leaders openly championing AI and to have redesigned workflows rather than decorated old ones. Second, Larke C treats governance as part of the product, not a brake on it. It knows which tools process personal data, where that data sits, and how the EU's GDPR and the phased AI Act apply to what it's doing. In Europe this isn't bureaucratic caution; it's what lets the company use AI on customer data confidently instead of nervously — and it's a genuine differentiator when a customer or a larger B2B partner asks the hard question.

The OECD makes the same point from the data side: the SMEs pulling ahead are the digitally mature ones — about 51% of them integrate AI into operations, versus 39% of the average firm — because they had the data, the habits, and the leadership to build on. The JPMorgan Chase Institute's 2026 analysis of small-business AI use tells the historical version of the same story: the firms that won were disproportionately the growth-oriented ones that moved deliberately from isolated pilots into operational integration, rather than the casual experimenters who dabbled and drifted.

What actually separates the three

Notice what didn't distinguish Larke C: it wasn't a bigger budget, a data-science team, or some exotic tool the others couldn't buy. All three companies had access to the same market of cheap, capable software. The differences were almost entirely organizational:

• Larke A hadn't started, and the cost of that was invisible — which is why it persisted.
• Larke B adopted tools but not a way of working, and got pockets of efficiency it couldn't measure or compound.
• Larke C started from business problems, redesigned the work around them, put a human and a governance layer in the right places, and tracked the result.

This is the real meaning of "AI readiness." It is not a measure of how many tools you've bought or how curious your team is. It is a measure of whether your strategy, your workflows, your data, your leadership, and your compliance posture are arranged so that AI compounds instead of just blinking on a dashboard. The McKinsey, OECD, and Eurostat evidence all converge on the same quiet conclusion: the technology is now the easy part. The organization is the hard part — and the part that decides whether you end up a Larke B or a Larke C.

If you recognized your own company in one of these three — and most owners recognize Larke B, usually with a slightly sinking feeling — that recognition is the genuinely useful output. The point of an honest AI readiness assessment isn't a score. It's to tell you which version of the company you're currently running, and the shortest credible path to the next one.

Sources

• Eurostat, Use of artificial intelligence in enterprises and "1 in 5 EU enterprises use AI technologies" (2025 reference year, published December 2025). EU enterprise AI use rose to 20.0% in 2025 from 13.5% in 2024; by size class, 17% of small, 30.4% of medium and 55.0% of large enterprises; highest in Denmark (42.0%), lowest in Romania (5.2%).
• McKinsey & Company, The State of AI in 2025: Agents, Innovation, and Transformation (November 2025), survey of ~1,993 respondents across 105 countries. 88% of organizations use AI in at least one function; nearly two-thirds have not scaled beyond pilots; 39% link any EBIT impact to AI (mostly under 5%); ~6% qualify as high performers; revenue gains concentrate in marketing & sales, strategy, and product development.
• OECD, AI adoption by small and medium-sized enterprises: OECD discussion paper for the G7 (December 2025). SME adoption lags larger firms; improved employee performance is the most-cited benefit; ~50% of SMEs report a staff skills gap; digitally mature SMEs (51%) adopt far more than the average firm (39%).
• JPMorgan Chase Institute, Understanding the use of AI among small businesses (April 2026). Many firms stalled in experimentation after generative AI's release; winners moved deliberately from pilots to operational integration.
• Thryv, 2025 AI and Small Business Survey. SMB AI use rose from 39% (2024) to 55% (2025), strongest in 10–100-employee firms; content marketing the most popular use case.
• Talkdesk (reported 2025–2026). ~51% of AI-using US small businesses apply it first to customer service.

Larke Home is a fictional, composite example created for illustration; any resemblance to a real company is coincidental.