In the latest episode of Orbit, Hg’s podcast series, Hg’s Head of Research David Toms sat down with Chief Investment Officer Matthew Brockman to discuss what AI means for enterprise software, for Hg’s portfolio, and for long-term investors. Here are ten key points from their conversation.

1. Markets are telling one story about software and AI; we're seeing another

There’s a growing gap in how public markets are treating different types of technology companies. AI model businesses like Anthropic and OpenAI are attracting enormous valuations and investor enthusiasm. At the same time, established software companies, many of which are still hitting their profit forecasts, have seen sharp falls in their share prices.

“You’ve got this sharp sell-off in the public markets for existing software, complemented with continued excitement, perhaps peak excitement, for what the model capability is and where that’s going to influence enterprise software over the long term.”

Our view is that this paints a misleading picture. AI doesn’t simply replace existing software; it dramatically expands what software can do. If AI genuinely opens up new markets, then at least some established companies, with their customer relationships, data, and expertise, will also capture these opportunities.

2. AI has got meaningfully better, fast

The latest generation of AI models released in late 2025 marked a significant step up in real-world performance. This wasn’t just a theoretical improvement; people using these tools in day-to-day work noticed a meaningful difference, particularly in software development and business applications.

“You can now run an instance of Claude for a day and it produces very high-quality code. Even six months ago that was an hour. And before that it was five minutes. The productivity effect is very pronounced.”

At Hg, we had expected this level of capability to arrive. The question was when. That ‘when’ has now become ‘now’, and we’re in a great position to take advantage.

3. Enterprise software is protected by more than just habit

A common narrative suggests that AI could quickly replace the software businesses use every day. But the reality is far more complex. Enterprise software isn’t just a product, it’s deeply embedded in how organisations operate. It handles regulated processes, stores irreplaceable data, and underpins workflows that have been built up over decades.

“You’ve still ultimately got to go and deal with complex customers who want to be regulatory compliant, who have had business processes, who employ thousands of people who need organisational change, who have existing data migrations. There’s a bunch of stuff in there which is just a natural moat.”

We describe these advantages as the ‘four Ds’: proprietary data, deep domain knowledge, established distribution (trusted customer relationships), and deterministic processes (where you need precise, predictable outcomes, not probabilistic guesses). Together, these give established software providers a significant head start when building AI-powered products.

4. AI will increasingly do the work, not just help with it

Until now, most business software has been a ‘system of record’, essentially a sophisticated way of storing and tracking information. AI is changing that. Software is starting to become a ‘system of action’, where AI agents can manage parts of the workflow themselves, under human oversight.

“One of the things that customers are very quickly looking to do is say, look, I have several hundred people dealing with processing invoices. I can pretty much automate that with your product, but I need you to deliver it for me because I need reliability, I need quality, I need the existing processes and existing data.”

This is the real economic shift. Software companies are no longer just competing for the roughly $1 trillion global software market. They’re starting to address the $50 trillion-plus human labour market by automating tasks that were previously done manually. The opportunity is enormous, but it requires established providers to move quickly.

5. Incumbents have a head start

A persistent worry is that nimble AI start-ups will simply outpace established players. But the companies Hg invests in already own the data, understand the workflows, and have trusted customer relationships. That gives them a powerful advantage when building AI-powered products, as long as they don’t sit still.

“If you’ve got that incumbency, I would be very surprised if most of our companies, if they act quickly and with deliberate intent, don’t become the natural agent provider. Why would a customer not be working with them in that way?”

The proof is already emerging. Invoice ingestion, for example, can technically be done for free using any general-purpose AI tool. Yet customers are choosing to pay their existing software provider for the same capability, because it’s integrated into their core system, governed by existing processes and, most importantly, trusted.

6. The time for experimenting is over; now it’s time to invest

A year ago, the message to Hg’s portfolio companies was about experimentation: try things, understand the potential, see what’s possible. That message has changed. Companies now need to invest serious capital into building AI-first products, reshaping their organisations, and rethinking how they serve customers.

"I was literally with a company this morning who are saying, well, actually we are going to spend a few extra million euros this year on AI build in a way that probably three or six months ago, they'd been very reluctant to do."

This is a significant commitment. Across the Hg portfolio, we’re talking hundreds of millions of dollars of investment in AI-related product development, capability building, and organisational change.

7. ‘Vibe coding’ is useful, but it’s not a replacement

‘Vibe coding’ is the popular term for using AI to quickly build software prototypes. It’s impressive: someone can now build a working demo of an application in an afternoon. But there’s a big difference between a prototype and a production-grade system that handles real customers, real regulations, and real data.

“I know that there are CPOs and product leads and CTOs in our organisations using vibe coding to wireframe something for customers that can evidence what a product could be. That is not in production. That is leading to a decision around a product to be properly developed.”

In other words, for now, it’s a tool that helps vertical software companies move faster, rather than a threat that replaces them.

8. AI spending hasn’t yet replaced traditional software budgets

One question investors often ask is whether money spent on AI tools like ChatGPT or Claude comes at the expense of spending on existing software. So far, the effect has been modest, perhaps a couple of percentage points of growth. Most AI budgets have been incremental, funded from research and development budgets or new allocations, rather than by cutting existing software subscriptions.

“I don’t think it was a straight substitution effect. A lot of these budgets were incremental. They were coming out of R&D, they were coming out of other product sets, they were coming out of different ways in which you might find the funding.”

However, for the AI model companies to justify their current valuations, they will ultimately need to access labour budgets, not just IT budgets. If that happens, it supports the view that software will become significantly more valuable, because it's doing more of the work. That's precisely the opportunity Hg's portfolio companies are positioned to capture.

9. In mission-critical software, the customer relationship stays with today’s software vendors – if they innovate

A key question is where the customer relationship sits in an AI-powered world. Will customers interact with their existing software provider, or bypass them and use a general AI assistant like ChatGPT? For the kind of companies Hg invests in, which typically serve small and mid-sized businesses with mission-critical back-office applications, Matthew believes the relationship is most likely to stay with the incumbent.

"I think in the vast majority of the companies that we invest in, which are largely SME, small enterprise and deal with sort of back-office compliance style of applications, it will sit with the software layer."

But this isn't guaranteed. This is software that handles tax filings, payroll processing, and regulatory compliance, areas where accuracy matters absolutely and where the existing provider already understands the rules. That's a strong position from which to build AI capabilities.

10. Long-term investment discipline is what matters most

Hg is approaching this moment with a long-term mindset. Rather than trying to time market cycles, the focus is on patiently building strong technology companies over five to ten years, much as Hg did during the transition to cloud-based software a decade ago.

“We’re not someone who’s going to try and call cycles. We are going to try and patiently build technology companies. I think there’s a huge opportunity in the next five or ten years for us to build some very successful technology companies.”

This includes significant support from Hg, notably through Hg Catalyst, our in-house AI product incubator that works alongside portfolio companies to accelerate AI product development.

Being a private investor gives Hg the flexibility to invest in these AI capabilities without the short-term pressures of public earnings targets. As David Toms, Hg’s Head of Research, regularly reminds us: long-term value creation in software has historically come from earnings growth, not from changes in market multiples. That principle hasn’t changed.

For further reading

Matthew Brockman, Everything, everywhere, but not all at once (February 2026)

David Toms, Six possible things before breakfast (February 2026)