Does AI change what we know about building startups?

The archetype of successful founders has fundamentally inverted between the SaaS era and the AI era.

This is the emerging consensus in venture.

But has the ideal founder profile really changed?

Yes and no. Let me explain.

Successful SaaS founders worked outside-in:

• Started with customer workflows and rebuilt them better in the cloud. Think Salesforce following Siebel, Workday following PeopleSoft, etc.

• The playbook was clear: take an existing enterprise workflow, make it cloud-native, win on user experience and pricing model.

• The technology was familiar territory. The real challenge was distribution and user experience.

AI flips this completely, successful AI founders work inside-out:

• Start with deep knowledge of model frontiers, then find the highest value applications.

• I.e. early winners live at the frontier of understanding which models excel at what and work from technology out, not customer problem in.

This isn’t about domain expertise becoming irrelevant. It's about the optimal starting point.

The argument is that in SaaS, you started with the customer problem because the technology was predictable. In AI, you have to start with the technology because the capabilities themselves are unpredictable. Then comes the customer problem.

This doesn’t mean don’t talk to customers - technically brilliant teams will still fail if they don’t understand their customers.

On the other end of the spectrum, domain experts who only chase customer requests to build incremental improvements vs pushing the boundaries of what’s possible (i.e. faster horses vs cars) will fail too.

Starting with technical possibilities does two things to domain experts:

1. Forces you to think bigger

2. Gives you a critical head start if you figure out what’s possible before others do - at a time when the data moat compounds (every interaction trains your model, every edge case strengthens your product) - granted you execute fast enough.

Nail this and you may be able to join the club of AI companies scaling from zero to $100 million in 12-18 months.

Unfortunately, winning is now harder than ever. Partially because founders now need to be able to hold two complex models in their head:

• The rapidly evolving capability landscape of AI models, and

• The nuanced needs of real markets

Both types of teams can win as long as they:

• Team of strong domain experts: Make sure to stay at the forefront of what’s possible with AI and challenge yourself to start there vs customer feedback.

• Teams with strong AI expertise: Get deep enough domain expertise on a space before the domain experts realize what’s possible.

That all said, I’m not a fan of generic statements like the ones I just made.

However, I do think it’s critical to understand how the world is changing and what it takes to win in this rapidly evolving environment.

In line with that, all of what I just wrote may irrelevant in 6 months already.