The archive.

Five UK political parties now have something to say about AI. I spent a week reading what Labour, Reform UK, the Conservatives, the Liberal Democrats and Plaid Cymru are actually proposing, past the headlines and the press releases. All five have a policy. Not one of them has a plan for a business that isn't a technology company.

Where each party actually stands

Labour is the only one of the five doing something right now, not just proposing it. Their AI Opportunities Action Plan covers a large increase in UK compute capacity, new AI Growth Zones and a national skills push. The concrete piece is AI Skills Boost, a free training platform launched this year with dozens of industry and public sector partners, aiming to train ten million workers by 2030, including a specific push to reach over two million people working in small and medium businesses. It is already showing up inside public services too, from AI-assisted NHS diagnostics to a chatbot helping people navigate government websites.

Reform UK's position sits at a different level entirely. AI is folded into a wider ambition for Britain to lead in emerging technology generally, alongside manufacturing and life sciences, built on a pro-innovation, deregulated environment meant to attract investment. Their clearest voice on this has warned that Britain risks becoming dependent on American AI firms unless it builds a serious domestic industry of its own. It reads as an argument about sovereignty as much as an argument about growth.

The Conservatives are the weakest signal of the five. Most of their record predates the last election: an AI safety summit and a light touch regulation paper that backed existing sector regulators over one central AI regulator. Nothing reads like an updated position on how AI should actually reach ordinary businesses.

The Liberal Democrats have the most distinct pitch of the opposition parties, and it is worth taking seriously. Their AI Digital Independence plan centres on routing critical national infrastructure through British-controlled technology rather than US tech giants, alongside a push to position the UK as a leader in ethical AI. Their most interesting idea is using citizens' assemblies to keep the public genuinely engaged in decisions about how AI gets used by the state. What is missing is anything that reaches down to the level of an ordinary business trying to adopt AI day to day. Sovereignty over infrastructure is a different problem from capability inside a company, and right now their platform speaks to the first, not the second.

Plaid Cymru is worth including separately, because Wales already has its own AI strategy running alongside whatever Westminster does. Their manifesto commits to supporting Welsh businesses through the opportunities in growth sectors including digital and AI, and the Welsh Government's own AI Cymru strategy sets a genuinely broader ambition, AI embedded across health, education, local government and the economy. What is missing, from Plaid specifically, is anything that speaks to the businesses I would actually call the backbone of rural Wales: not technology companies, no in-house digital function, often bilingual by necessity rather than strategy. A national ambition to embed AI across the economy is a strong starting position. It is not yet a plan for the accountant in Brecon or the farm outside Hay.

The gap none of them are filling

Labour's training platform is the right instinct, and it is real. But it answers one question. Can a person get access to AI training? It does not answer the harder one, which is the one that actually decides whether any of this turns into growth. Can an existing business, one that is not a technology company, actually change how it works because of this?

Inside the businesses I know, the accountants, the trades, the retailers, the farms, the small manufacturers, the gap has a specific shape. Leadership often does not know what good adoption looks like for their own operation, so investment lands on the wrong things or does not happen at all. Skills bought in from outside, a consultant, a course, a new hire, tend to leave the business the moment that person does. And more than either of those, the thing that actually decides it is how the workforce experiences the change. AI imposed from above as a cost cutting tool breeds quiet resistance. AI built with people, who can see it making their own work better, is what actually sticks.

What actually needs to happen

• Build capability inside the team, not just give people access. A skill only counts as an asset once it survives the consultant, the course or the new hire moving on.
• Bring the workforce in early, not last. People need a real stake in how AI changes their own work, not a rollout that happens to them.
• Point policy at ordinary businesses, not just the headline numbers. The economic prize from AI is won or lost inside businesses that are not technology companies, not in a national training statistic on its own.

None of the five parties are wrong that AI matters. They are just all aiming at the same layer, the national one, while the decision that actually counts happens somewhere much smaller. Inside a business, in the room where someone decides whether to trust what the tool just produced. That is where the next few years will actually be won or lost, whichever party is in government when it happens.