The UK’s Scale-Up Problem Starts Earlier Than We Think

By Steve Orr, CEO of Catalyst

The UK does not have one scale-up problem. It has two.

The first is increasingly visible in London and Cambridge. People are leaving leading AI labs, starting companies, raising large venture rounds and growing quickly. Brands such as Revolut have become internationally significant.

The challenge is how the UK keeps exceptional companies independent, financed and creating long-term value here. DeepMind was sold to Google early. Arm is still headquartered in Cambridge but is owned by SoftBank and listed in New York. These examples explain why government is so exercised by ownership, control and where the benefits of breakthrough UK innovation ultimately end up. Solving that needs deeper late-stage capital, more ambitious investors, stronger procurement and better reasons for founders not to sell too early.

But there is another problem that I don’t think has been given enough attention. Most parts of the UK are not producing companies spun out of frontier AI labs and immediately raising enormous rounds. Their challenge is earlier.

So, how do we create more companies with the fundamentals required to scale in the first place? That matters if we are serious about good growth in every postcode, including our base in Northern Ireland, not just the places where scale is already happening.

For decades, the UK has tried to address this with familiar interventions: accelerators, investor readiness, mentorship, procurement, talent and capital. All matter. But they will not turn a company into a scale-up if it was never created to scale.

Let me explain what I mean by that. The UK startup sector still relies heavily on Lean methods. Lean was a major advance, but in the AI age its emphasis on early MVPs is becoming outdated. When building is now so fast and cheap, the greater risk is building too soon on the back of partial or poor customer discovery.

Disciplined Entrepreneurship: an approach developed by MIT which Catalyst has adopted starts from a different principle: proof before product.

That means validating the customer, market, economics and go-to-market so thoroughly before building that there is strong evidence demand could outstrip supply.

This is much more than speaking to potential customers and hearing encouraging comments. A company must prove the right first market, a painful customer problem, the end user and economic buyer, willingness to pay, pricing, competition, route to market, sales cycles, gross margin and unit economics.

Most importantly, it must show that customers can be won repeatedly without the founder dragging every sale over the line.

At Catalyst, our view is simple: how a company is created will determine whether it scales. That applies equally to a two-person AI software company and an academic spin-out commercialising complex science.

AI makes this more urgent. The scarce capability is no longer simply technical development. It is knowing what is worth building and proving that a market will pull it from you. Go-to-market is becoming the decisive constraint.

Jobs created and investment raised remain important outcomes. Investment will still be critical for deep tech companies. But neither proves that a scalable market exists.

Sales are becoming the clearest test, supported by extensive customer and commercial validation. Increasingly, that is also the evidence the strongest investors will expect before committing serious capital.

Through our partnership with the Martin Trust Centre at MIT, Catalyst is using Disciplined Entrepreneurship to build a data-driven stage-gate model for deciding when a company has earned the next level of support.

A company does not progress because it completed a programme or impressed a panel. It progresses because the evidence shows that it has removed more risk and moved closer to repeatable demand.

Those stage gates create an evidence chain between intervention and outcome. They give funders a better view of whether a venture is becoming more investable and scalable before the lagging outcomes arrive.

They also make it possible to compare different forms of support. Traditional accelerators and old-style mentorship should now be tested against AI-enabled tools, intensive residencies and more structured methods.

Government funders can now build a much stronger view of the return on investment from different mixes of support. By connecting intervention data with stage-gate progress and eventual outcomes, they can better understand what contributed to stronger sales, faster growth, more investment, international market entry and the creation of more Innovation Driven Enterprises.

Northern Ireland is the perfect place to test this properly. We have strong research, technical talent and companies that compete internationally. We also have a small domestic market, a thinner ecosystem and little room to keep funding activity that produces no evidence of progress.

We can apply a more rigorous method. Use evidence to decide when companies are ready for deeper support. Compare the results with the programmes we have funded for years.

If one mix of interventions produces better outcomes for the same public money, back it. If another does not, redirect the funding. At least then we will be making decisions based on evidence rather than habit.

Taking decisions on this basis will lead to more of our businesses becoming meaningful, robust and sustainable scaleups in future.