Where Are the Opportunities for New Startups in Generative AI?
Sharing our mental model on gen AI in SaaS
Over the past twelve months, like many others, I’ve been diving deep into generative AI and its implications for SaaS companies, as well as what it means for us as early-stage SaaS investors.
I’ve already shared some of my thoughts before (e.g., here, here, here, and here), and Louis Coppey summarized his learnings exploring GPT/LLMs in this great post.
Now, I want to put out our “mental map” on the topic. It’s far from perfect, it’s not polished, and it’s simplified in many ways. Think of it more as a rough snapshot of our current understanding and hypotheses. I’m sure we’ll revise it considerably, which is precisely why we’re publishing it: to gather feedback from as many people as possible.
What do you agree with? What do you disagree with or have additional perspectives on? We’re particularly interested in concrete examples that either confirm or refute our ideas. So if you have any such examples, bring them on!
Importantly, the map is not true to scale, so if it looks like most of the categories are red and yellow, that’s the wrong takeaway. If the map was true to scale, I believe the green parts of the map would be much, much larger. Also, note that we’re only at the application layer here, leaving the infrastructure layer for another post. :)
The map is pretty self-explanatory, but here are a few notes.
- Re. general-purpose tools, a number of startups are challenging the incumbents with very cool and innovative products, e.g. beautiful.ai and Tome for presentations, Rows for spreadsheets, and mem for note-taking. It’s great to see startups rethink categories and UIs with an “LLM first” lens, and some of these companies are growing very fast, but the jury is still out on whether any of them will get huge mainstream adoption.
- We’ve mentioned auto-pilots as a very interesting opportunity, i.e., the idea to sell the work results instead of the software. We’ve seen this in accounting some years ago (with Bench, and, more recently Pilot) and are very curious about this approach in other areas.
- For categories with an existing SaaS champion, one of the questions is what kind of data the existing player(s) have access to. For example, as a practice management solution, Clio sits on a wealth of data that enables it to build an AI solution that helps customers with all aspects of managing a law firm. However, Clio is not a contract drafting tool, so it doesn’t necessarily have a data advantage when it comes to contract work (although having 10,000s of customers will obviously help, should Clio decide to go for it!).
- One of our theses is that AI might accelerate the adoption of SaaS in laggard verticals — by delivering more value to the customer faster and with less effort for the user. But this is admittedly mostly an idea, and we haven’t seen great examples of it just yet — have you seen any examples of this actually happening?
- Just to make it super clear, that inconspicuous little corner in the bottom right of the map is something we’re extremely bullish about — and which would be much larger if the map was true to scale!
