AI startups face different challenges than the typical SaaS company. That was the message delivered by Rudina Seseri, founder and managing partner at Glasswing Ventures, last week at the TechCrunch Early Stage event in Boston.
Seseri explained that just because you connect with some AI APIs does not imply you are an AI company. “By native AI, I don’t mean that you slap on a shiny box with some call of OpenAI or Anthropic with a human-like UI and you’re an AI company,” Seseri said. “What I mean is a situation where it’s really algorithms and data that are at the core of creating the value that you’re delivering.”
Seseri says this implies there are significant differences in how customers and investors evaluate an AI company in comparison with a SaaS startup, so it is vital to know these differences. For starters, you possibly can put something that is far from complete into the world with SaaS. For many reasons, this can’t be done with artificial intelligence.
“Here’s the thing: When you code a SaaS product, you provide quality control and, in a sense, you get a beta version — it’s not a finished product, but you can release it and get going,” she said.
Artificial intelligence is a completely different animal: you possibly can’t just release something and hope for the best. This is because an AI product takes time for the model to achieve a point where it is mature enough to work for real customers and for them to trust it in a business context.
“The learning and training of the algorithm is steep at first, but it has to be good enough for the customer to want to buy, so it has to be good enough for you to create value,” she said. And that is hard for an early-stage startup to seek out.
And that makes finding early adopters harder. He says you wish to avoid long conversations where the buyer is simply attempting to learn more about AI. Startup founders don’t have time for such conversations. He says it is vital to focus on the product and help the buyer understand your value proposition, even if it’s not fully accomplished yet.
“Always articulate the problem you are solving and what metric – how do you measure it?” she said. Optimize for what’s essential to the buyer. “So you’re solving a problem that has implications for business decisions.” It’s okay to precise your vision, but all the time base the discussion on your business priorities and how they impact your algorithms.
How can AI startups win?
As you build your business, it’s essential think about how you possibly can take a defensible place in AI, which is especially difficult because the big players are always developing huge chunks of business ideas.
Seseri points out that in the cloud era, there was a base layer on which infrastructure players staked their claims; the middle layer where platform players lived; and at the top we have the application layer where the SaaS lived.
With the cloud got here several players such as Amazon, Microsoft, and Google that controlled the infrastructure. The AI base layer is where the big language models run, and several players have emerged such as OpenAI and Anthropic. While you may argue that these are startups, they are not in the true sense of the word because they are funded by the same big players that dominate the infrastructure market.
“If you’re going to compete for a new entry-level or, you know, LLM game, it’s going to be very difficult with multi-billion dollar capital requirements, and ultimately it’s likely to become a commodity,” she said.
At the top of the stack is the application layer, which may very well be used by hundreds of SaaS corporations in the cloud era. She said that big players like Amazon, Google and Microsoft were not in a position to take over the entire application layer business and there was room for startups to grow and grow into large, successful corporations.
There is also a middle layer where the plumbing is installed. He points to corporations like Snowflake that have managed to build successful middle-tier businesses by giving app players a place to place their data.
So where is he investing when it involves artificial intelligence? “I invest money in the application layer and very selectively in the middle layer. Because I think there’s a moat around algorithms, whether or not they’re proprietary algorithms or open source solutions – and data. You don’t have to be the owner of the data. But if I have to decide on, I would really like to have unique access to data and unique algorithms. If I have to decide on one, I’ll reach for the data,” she said.
Building an AI startup is definitely not easy, maybe even harder than a SaaS startup. But this is where the future lies, and corporations that are going to try it have to know what they’re up against and build it accordingly.