The future of investing in AI

This article originally appeared in Forbes.

‍

It’s a challenging time to raise startup capital. Total investment in startups has fallen substantially—down 80% in the first quarter of this year compared to the same period in 2022. Of the founders who have succeeded in raising, almost one-fifth are having to take a down round.

But there is one industry bucking the trend: AI. According to PitchBook—even excluding OpenAI's latest $10 billion round—AI venture funding in 2023 has far surpassed what it was in 2022 and is well over halfway to what it reached in 2021, when it peaked at $9.1 billion.

From a broad view of the market, investors have tightened their investment criteria across the board after what was the most exuberant era of venture fundraising and dealmaking in history. A record amount of capital was raised at the start of the decade, reaching $128.3 billion in 2021, up 1.5 times compared to the year before. Much of that capital still remains to be deployed.

AI is here as the latest in a series of headline-stealing technologies to capture investor demand. In the past few years, it was crypto and the metaverse (to questionable value both for consumers and investors). But with the explosion of LLMs and other generative models, enabling new apps like ChatGPT and Midjourney to become some of the fastest-growing in history, AI has demonstrated its staying power as a transformational technology.

These tools have changed what is fundamentally possible with software and created a new wave of ideas, startups and talent flocking to build in the space.

As an investor, it’s almost impossible to ignore any hype cycle. The fear of missing out is real even in uncertain times when high-risk opportunities are usually otherwise avoided. AI has opened up new possibilities for disruption across every industry. This is the psychology that has been driving the divergence we’re seeing in AI investing compared to the rest of the private market.

These are three observations about how the opportunity in AI is evolving.

1. The ability to accrue unique data sets is becoming a defensible moat

LLMs have made it possible—even easy, in some cases—to build a huge number of new features and products that otherwise would have been very hard to do in-house building and training models from scratch. But there’s a distinct advantage and defensibility for companies whose LLM-powered features are built on unique datasets they own.

Consider a company like Box. Their Box AI offering is enabled by the files and content their enterprise customers are already storing on their base platform. By feeding that data into a language model, Box AI can offer features like a Q&A chatbot about those files/content that only Box is uniquely capable of building, given other companies don’t have access to the same underlying data.

Generally, any company that offers a system of record (like file storage or CRM) has a huge opportunity to deliver defensible value with generative AI. Investors should be looking for and understanding data moats when considering investments in new AI companies.

2. How companies adopt AI internally can be as important as what features they build with it

There’s a lot of talk among investors about what new features and businesses can (and will) be built using AI. However, there isn’t as much discussion about how portfolio companies could be leveraging AI internally to improve their internal operating efficiencies. In my view, this is as central to a company’s AI strategy as the product roadmap.

Companies that successfully adopt AI productivity tools will be able to scale much more efficiently because of the improved results when humans and technology work in tandem. A tool like ChatGPT is able to process a huge amount of unstructured data—for example, user feedback from platforms like Discord—to understand what people care about and what you, as a builder, should know about. But then, ultimately, it's still up to humans to make decisions.

Investors should make it part of both their portfolio company support and due diligence to learn about how startups are using AI to improve their efficiency metrics and team productivity and to reduce costs. Especially in an uncertain macro environment like we are now, AI can make a significant difference to the bottom line.

3. There is an untapped opportunity in AI-powered services

Right now, we’re seeing a lot of the AI conversation focus on AI-powered software features and products. But I think there will be an equally large opportunity in the creation and reinvention of entire services sectors using AI.

Industries like banking, law and accounting have traditionally provided value through the delivery of services, which is a highly human-intensive operation. Most of the value of those industries accrued to the biggest services firms themselves.

But how many of those human roles can be automated now with AI, and what might an AI-first law or accounting firm look like? This will be an interesting question for both incumbents and founders in these spaces.

This overall hype cycle? It won’t last—they never do. As is the nature of startups, there will be more failures than successes. But, AI is lowering the barrier to entry in many industries. Opportunities are being created at a faster pace, there is nascent opportunity in areas like AI-reinvented services and companies with unique data sets have the potential to achieve some truly transformational outcomes.

Ultimately, it’s up to investors to find and fund these opportunities—and it’s vital that they consider a company’s AI strategy in all deals going forward.