Company-Building during the AI Wave

For anyone starting a company in 2024, there are new questions to grapple with, such as “How is AI going to change this industry,” “are we in an AI bubble,” “are incumbents going to win with AI or are new startups going to win,” and “is AI for X a good startup idea even though dozens of other companies are going after the same goal?”

This post walks through the principles I’ve used when considering AI opportunities to bet on.

1. Act like it’s 1995.

I’ve generally been a skeptic of hyped industries (eg. crypto, web3, etc.). AI is the first hype wave that I’ve become a true believer in. Even as a true believer, I don’t believe that the most extreme outcomes are likely. Instead of AI putting humans out of work, my bet is more modest: I would bet on a 2–5x surge in American productivity over the next decade.

While this is a much smaller outcome than many AI advocates believe is likely, this would be a massive change for the US economy and society — and an even bigger outcome than the .com boom.

Based on this assumption, I look at the world like it’s 1995 for the .com boom — we are very early in a massive transformation in the economy, and while there is some froth and some ups and downs that will come in the coming years, fundamentally there is about to be an explosion in value for AI innovation.

As a result, I believe that:

1. This is a moment to be long on everything. It’s part of why I embarked on Shaper and think it makes sense to start lots of companies simultaneously.
2. This is the best time ever to start a new company. There is a massive amount of innovation about to take place, and value to be captured by innovators.

2. Fundamentals matter, and companies need to find defensibility.

Many companies are being started with the view of “X is possible, and if I do X, it will revolutionize an industry.” Even in cases where that may be true, it doesn’t necessarily result in a good startup idea; the fundamentals of “will this be a good business at steady state” still need to be true.

In particular, in the race to execute on exciting opportunities with AI, it seems like many companies aren’t asking themselves whether their business is defensible; in the long-term, these companies will not thrive, even if they were the first to bring AI into X industry. Being a first-mover can be valuable if it’s a means to an end to create a defensible moat, but economic returns will ultimately accrue to the portion of the value chain that creates differentiated value and has few substitutes.

As a corollary — the advantages that large incumbents in many industries have (clear moats, distribution networks, etc.) will continue to be advantages in a post-AI world, and large incumbents that are clever about utilizing AI will emerge stronger than ever.

3. Beware Silicon Valley dogma that is no longer applicable.

While fundamentals matter that are based on first principles, there are many pieces of startup dogma that I think are not necessarily true in the AI era.

For example, I think the “product” vs. “services” distinction is not especially relevant, and is not based on first principles. Margin certainly matters, and scalability certainly matters — which has historically been correlated with being a “product” business. But in the era of AI, there will be high-margin, highly-scalable “services” businesses as well.

Similarly, optimizing for “recurring revenue” in SAAS is overly dogmatic. Having strong long-term value from customers with minimal customer acquisition costs is a relevant goal from first principles, but recurring revenue isn’t the only way to get there; Google, Facebook, Apple, and Amazon don’t have recurring revenue models today on their core businesses. I’d predict that many top companies innovate on pricing as well.

4. Opportunity abounds in the complements

The AI wave won’t only benefit “AI companies;” it will also benefit complementary technologies and businesses.

For example, the area I’m most focused on is data. While data itself doesn’t change with AI, I predict the value of certain data sets to skyrocket based on the AI wave, because it is so valuable to AI companies.

I’d predict that many of the most valuable businesses created by the AI wave won’t be directly related to AI, but rather are complementary to AI.

5. Figuring out product market fit will be hard.

One of Datavant’s board members always asked the same valuable question in 100% of board meetings: “What percentage of your customers are getting fundamental value from you, and what percentage use you in their ‘innovation showcase’?”

This was always an extremely useful question. The reality is, large entreprises spend money on technology that creates value for them, but they also spend money on “innovation projects” that are ultimately proofs of concepts or vanity projects.

True product-market fit is based on creating fundamental value for customers. During a major boom, lots of money is spent on “innovation showcase” projects — and it’s possible to get a large amount of revenue without having true product-market fit. Differentiating “revenue growth” from “product-market fit” will be complex for entrepreneurs, but large economic outcomes will only come from companies that get true product-market fit.

6. Blitzscaling will make a comeback in some areas.

In the 2010s, blitzscaling was popular as a strategy, and often made sense as a strategy: because interest rates were low, spending capital today for growth tomorrow made sense. There were many companies that took this too far, and threw money away wastefully.

By 2022, this became frowned upon, and companies had a wave of belt-tightening as interest rates went up.

Even with higher interest rates, the strategy of blitzscaling still fundamentally makes sense in many AI businesses: in a moment of rapid technical change and large prizes for moving rapidly, going into “land grab” mode makes sense as a strategy. Companies will need to ask themselves how quickly an industry will change, how much first-mover advantage matters, and what the prize is at the end (eg. is there defensibility at steady-state).

As a result, there will be more companies in 2024 that pursue a blitzscaling strategy, and losing money in the near-term can be a viable strategy. (Hopefully, this doesn’t become synonymous with “wasteful spending” again— but there will be companies that take this too far)

7. Talent will be the bottleneck

As industries change quickly with AI, many companies will win by being nimble and quickest to capitalize on market changes. The biggest lever in corporate nimbleness/velocity is talent, and great talent will become even more valuable.

While “AI talent” is at a premium today, much of the value will actually come from generalist talent. Strong generalist engineers who can figure out how to incorporate AI into technical workflows will become increasingly valuable. Strong generalist managers who can figure out how to navigate a rapidly-changing industry will become increasingly valuable.

Companies that have access to the best generalist talent will disproportionately move faster, and disproportionately win. While this has always been true, I believe that the AI wave will amplify this effect.