I’m a VC who bets on AI. What keeps me up at night isn’t the idea of these companies failing—quite the opposite

I’m going to be honest about something most VCs won’t say out loud: what keeps me up at night isn’t that the AI companies I back will fail. It’s that AI as a technology will succeed so much that it makes the business models of the companies I back completely irrelevant.

We are living through a genuine step change in the industry. The advent of AI-assisted application development, known colloquially as “vibe coding,” has meant applications that once took years to build can now be replicated in weeks. And as big tech companies bundle AI features into a wider product suite, standalone chatbots lose their pricing power overnight.

The human management software space — startups building digital tools automating and managing employee-related tasks such as payroll and performance — is set to go the same way as firms like Microsoft roll out similar products as part of a wider bundle. Both developments underline an uncomfortable truth for venture capitalists right now: we risk overpaying for companies that look genuinely special today but could become generic — or worse, easily replicated by AI coding tools — within a year.

Just take a look at the pressure the SaaS subscription model — which has long been the backbone of a generation of VC-backed companies — is under in the public markets. Amid the so-called “SaaSpocalypse” that’s engulfed capital markets since the beginning of the year, software stocks have lost over $1 trillion in market cap. That’s largely been stoked by fears AI agents can now perform entire workflows that previously required multiple SaaS subscriptions.

In other words, none of what I’m saying is theoretical. It’s actually already happening.

So how am I approaching investing in this environment? My answer is to stop treating AI as a vertical and start treating it as a layer — and to focus on the underlying structure the technology cannot easily replace. Specifically, I look for three things. First, a company must own customer trust and distribution. Second, it needs to be embedded in systems where real money moves. And third, it must accumulate proprietary data that compounds efficiency over time.

Practically, this also means I’m more comfortable making multiple bets across a specific vertical rather than placing a single large bet on an AI-first business. Fraud compliance and security is one such vertical.

Exposure across the full stack is important — from KYC and identity verification, as with our portfolio company Smile ID, to transaction fraud monitoring, as with Orca. Over time I expect these to overlap, and the company with the deepest data will win regardless of which underlying model it uses.

Similarly, I am watching agentic platforms closely: tools that come to market generalized but gradually become embedded in a specific industry’s workflow. Companies building for specific niches will be harder to displace.

But if I am looking for the clearest bright spot in AI investment right now, I keep coming back to emerging markets — and to markets like Africa specifically.

The continent comprises 54 countries, each with its own regulatory environment, currency, and infrastructure. It’s because of this fragmentation that global players in AI have largely ignored it in favor of the U.S. and Europe. While that’s long been framed as a disadvantage to scaling a business, it also gives local companies a chance to develop while larger tech companies focus on geographies elsewhere.

Consider commodities trading. A platform like Norrsken22-backed Sabi doesn’t just match buyers and sellers. It also coordinates supply, transport, quality checks, storage, and financing across borders where none of those systems reliably talk to each other.

AI is being deployed to handle tracking, tracing, and pricing information flows across that entire chain. The company that solves this end-to-end becomes the default operating system for its market, and the data it accumulates in doing so is extraordinarily difficult for any competitor to replicate.

That describes the pattern I’m watching for. These are companies that do not merely sell software but run the process, accumulate regional data that does not exist elsewhere, and earn trust in markets where it’s trust is scarce.

That’s not to say there isn’t a lot of uncertainty in this industry right now. But I have found it clarifying to stop asking “which AI company should I back?” Instead, I’ve started asking which problems the major players are ignoring — and which companies are building solutions so deeply embedded in those problems that AI won’t make them obsolete.

In emerging markets like Africa, those companies are everywhere. Investors just have to be willing to look.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com