Commentary
Honest analysis of the AI landscape as it affects working professionals and business owners. Published fortnightly. No vendor marketing. No manufactured urgency. Just evidence-based assessment of what's happening and what it means for the people in the middle of it.
Ninety-Four Billion Pounds
The gap between what SME productivity could deliver and what SMEs are actually being offered
The UK government's SME Digital Adoption Taskforce produced a striking figure last year: a 1% productivity uplift across Britain's small and mid-sized businesses would add £94 billion annually to GDP. One percent. Not a transformation. Not a revolution. A modest, compounding improvement across the 5.5 million firms that employ 60% of the private sector workforce.
Hold that number in mind whilst considering what those firms are actually being offered to achieve it.
On 28 January, the government launched the AI Skills Hub with £27 million in funding. Within hours, practitioners identified what the programme actually is: a directory of technology vendors presented as a training resource. Of the founding partners, one has education as its core business. The rest sell software. An EdTech provider compared it to “launching a national motor driving skills programme and partnering exclusively with car manufacturers, with no driving instructors involved.”
The same week, Somerset County Council cut its Skills Bootcamp funding by 68%. From 1,100 learners to 380. The pattern is plain: money flows towards programmes that look like action. Actual capability-building infrastructure loses resources.
This is the context in which 5.5 million businesses are supposed to achieve their 1% uplift.
The data on where those businesses actually stand makes the gap more vivid. The British Chambers of Commerce found that 82% of business leaders believe AI will be essential to competitiveness. Only 25% have meaningfully integrated it. 35% of UK SMEs have started using AI in some form, but only 11% feel they're using it “to a great extent.” 43% have no AI adoption plans at all.
That 57-point chasm between believing AI matters and having done something about it is not a technology problem. The tools are increasingly accessible — a practical AI suite is implementable for approximately £69 a month. It's not a willingness problem either. These are people who adopted cloud accounting, online booking systems, and social media marketing in earlier waves. They're not hostile to technology. They adapt quickly when shown something useful.
It's a guidance problem. 51% of SME leaders say they lack sufficient AI knowledge to make informed decisions. 39% are paralysed by an inability to identify relevant use cases. They know they should be doing something. They cannot see what.
And what's being offered doesn't help them see. The enterprise playbooks — McKinsey frameworks, Deloitte transformation models — are designed for firms with CTO offices and dedicated budgets. A 25-person construction firm or a three-partner solicitors' practice has no use for a “digital transformation roadmap.” They need to know: given what I've got, this team, this budget, these clients, what should I actually do on Monday?
The vendor webinars answer that question, but the answer is always the vendor's product. Microsoft Copilot demonstrations, Salesforce AI webinars, HubSpot's AI features explained. Product marketing positioned as strategy. The firms that benefit most from these tools are the firms most like the vendor's existing customer base — which rarely includes a 15-person plumbing company.
The “Top 10 AI Tools for Small Business” article genre fills the remaining space. Lists of software products with no strategic framework for choosing between them. Often thinly disguised affiliate marketing. Helpful for someone who's already decided to adopt AI. Useless for someone trying to decide what to adopt and why.
So the owner-manager is left precisely where they started: knowing they should act, unable to see how, surrounded by content that either doesn't speak to their reality or speaks to it in order to sell them something.
The sectoral divide makes this worse. B2B service firms — finance, law, marketing — are at 46% AI adoption. They work with information. Their clients expect technological sophistication. The tools map more obviously onto their operations. Manufacturing sits at 19-26%. The trades lower still. The firms where AI's relevance is least obvious are the ones receiving the least useful guidance about where that relevance might actually lie.
What would £94 billion worth of productivity improvement actually require? Not vendor catalogues. Not enterprise frameworks scaled down to small business without being redesigned for it. It would require guidance that starts with the business as it actually exists — its real constraints, its real data, its real team — and works outward to the technology from there. It would require honesty about what AI can do for a specific type of firm right now, what's premature, and what's noise. It would require someone who understands that the owner-manager reading the guidance is also the person who has to implement it, fund it, explain it to the staff, and live with the consequences.
The £94 billion figure is aspirational. The £27 million being spent to pursue it is misdirected. The gap between the two is where millions of businesses are trying to make decisions with inadequate help.
Eighteen Months
What Suleyman's prediction misses about who actually bears the risk
Mustafa Suleyman told the Financial Times last week that “most, if not all, professional tasks” involving sitting at a computer would be “fully automated by an AI within the next 12 to 18 months.” He named accounting, legal, marketing, and project management explicitly.
Set aside, for a moment, whether he's right about the timeline. The more revealing problem is who disappears from view in a sentence like that.
“Professional tasks” is doing enormous work in Suleyman's formulation. It flattens a workforce of tens of millions into a single category — as though the financial analyst in the City and the administrative assistant in Swindon face the same disruption, with the same resources, on the same timescale. They don't. And the data that shows why they don't has been available for three weeks, largely unremarked upon in the coverage of Suleyman's prediction.
In January 2026, the Brookings Institution and the Centre for the Governance of AI published research examining not just who is exposed to AI automation but who has the capacity to adapt to it. The distinction matters more than almost anything else in this conversation.
Of 37.1 million US workers in the top quartile of AI exposure, 26.5 million also have above-median adaptive capacity. They have savings. Their skills transfer across industries. They work in cities with dense job markets. They maintain professional networks that surface opportunities. They are exposed — but they are equipped.
The remaining 6.1 million are not. These workers face high exposure to automation and low capacity to respond. Limited savings. Skills specific to the roles being automated. Thin professional networks. Concentrated in smaller metropolitan areas with fewer alternative employers. 86% are women. They are disproportionately in clerical and administrative roles — the work that sits closest to the automation frontier and furthest from the resources that make adaptation possible.
Financial analysts score in the 99th percentile for adaptive capacity. Office clerks score in the 22nd. Both appear in the same sentence when Suleyman says “professional tasks.” Their situations could not be more different.
This isn't a footnote to the automation story. It's the automation story, or it should be. When the Microsoft AI CEO predicts sweeping disruption of professional work, the implicit audience is the professional class — the people reading the Financial Times, attending Davos, following AI developments with informed concern. They hear “your tasks will be automated” and feel a chill. But the Brookings data suggests most of them have the resources to navigate what's coming. Not painlessly, but with options.
The people for whom Suleyman's timeline would be genuinely devastating — the 6.1 million with high exposure and low adaptive capacity — are not reading the Financial Times. They are not the implied audience for predictions made at technology conferences. They are largely invisible in a conversation that is, in principle, about them.
Evercore's analysis adds a dimension that the “18 months” framing obscures entirely. AI does not tend to eliminate jobs in one stroke. It automates 30-40% of tasks within a role, leaving the role smaller but still present. The legal associate who spent 60% of her time on document review now spends 30%, because AI handles the first pass. The role contracts. It doesn't vanish.
But contraction has consequences that don't make headlines the way elimination would. Fewer entry-level positions. Compressed career progression. Stanford's Digital Economy Lab found entry-level hiring in AI-exposed jobs already down 13% since large language models proliferated. The people at the start of their careers are being quietly denied the start. Employee concern about AI-related job loss has risen from 28% in 2024 to 40% in 2026. That near-doubling isn't panic. It's a rational response to a career ladder that is visibly losing rungs.
None of this means Suleyman is wrong about the direction. The technical capability is advancing rapidly. The models released this month can do things that would have seemed implausible eighteen months ago. The question is not whether AI will reshape professional work — it will — but whether “eighteen months” is a forecast or a positioning statement from the CEO of a company that sells the tools he's predicting will be indispensable.
The distinction between “can automate,” “will automate,” and “has automated” is where honest analysis lives. The technology can do much of what Suleyman claims. But deployment requires restructuring workflows, retraining staff, managing resistance, absorbing initial productivity drops. For large enterprises with transformation budgets, this is achievable. For the 5.5 million UK SMEs that employ 60% of the private sector workforce, it's a different proposition entirely.
What would actually help — what would serve the 6.1 million as well as the 26.5 million — is guidance that begins with the person's specific situation rather than the technology's general capability. Where do you sit on the exposure spectrum? What are your actual resources for adaptation? What should you do first, and what can you safely set aside?
These questions don't generate headlines. But they're the ones that matter, and they're the ones that Suleyman's eighteen months does nothing to answer.
Stay informed
Fortnightly analysis. No vendor marketing. Unsubscribe whenever you like.