Bessemer Venture Partners: Vertical AI Shows Potential to Dwarf Legacy SaaS

Caty Rea, Kent Bennett, and Brian Feinstein of Bessemer Venture Partners had a great conversation about how Vertical AI is a 10x bigger opportunity than legacy Vertical SaaS.

Top Discussion takeaways:

• Vertical AI is expected to have a market capitalization at least 10 times the size of legacy Vertical SaaS by targeting high-cost repetitive language-based tasks across various industries, potentially transforming sectors like business and professional services, which are significant portions of the economy.
• Three main business models are emerging in Vertical AI: a) Copilots: Enhancing existing workers' capabilities b) Agents: Performing tasks previously done by internal headcount c) AI-enabled services: Productizing traditionally outsourced services
• Vertical AI companies are showing significant pricing power, commanding around 80% of the average contract value (ACV) of traditional vertical SaaS companies.
• This pricing power suggests that the total addressable market (TAM) for vertical software could potentially double with the addition of AI-enabled services.
• Vertical AI companies are not necessarily replacing existing software spend, but often replacing or augmenting services spend within organizations.
• Despite being service-like, many Vertical AI businesses are demonstrating software-like characteristics: Rapid growth rates (some at 400% at scale) High gross margins (60-70%+) Differentiated offerings
• The impact of AI on professional services firms may be less severe than initially thought, with some potentially benefiting from adopting AI technologies.
• Multimodality in AI (ability to process and generate various forms of data like text, voice, and video) is expected to open up new opportunities in vertical markets.
• Incumbents in vertical markets have advantages (distribution, customer relationships, brand), but their success in adopting AI will depend on execution.
• Many new vertical AI opportunities are emerging in areas where traditional software was less effective, particularly in language-heavy tasks.
• Vertical AI is enabling the capture and utilization of previously untapped data sources (e.g., voice conversations in sales), creating new defensible positions for companies.

From Bessemer Venture Partners

Highlighted Transcript:

Caty Rea

So I think at a high level, we're seeing three business models that these companies are entailing. So the first is copilots. And this was really the first wave of vertical AI. So I'd say this time a year ago, really, the thing that we were seeing was verticalized copilots. So we have a company sixfold in our portfolio. They're a copilot for insurance underwriters. It helps them parse all this data, better understand the risk profile. And so in that way, copilots really turbocharge an existing worker and help them gain efficiencies and superpowers in this way.

The second one is agents. I'd say this is really the wave that we're starting to see now and tying to the multimodal portion of state of the cloud. Many of these are multimodal. They're voice agents, they're text agents, they're coding agents, et cetera. And those do the work of a person. So that's typically internal headcount. So a sales rep, someone picking up the phone, same day, does inbound call reception for professional services, for home services businesses.

And then lastly, in the even up style, AI-enabled services. So these are typically third-party outsource service spend that's now able to be productized thanks to AI.

And so these are companies that are cropping up. They might be TPAs, they may be legal services, they may be broader just professional services, and they're able to tap into that services spend.

Kent Bennett

All right. So what would you say to somebody who says, Well, that's not software. I mean, Brian, you just said that's not software. And so in our industry, that's a bad word, right? This is, Oh, that's not software. That's not software. That's not a venture-backable business. That's not going to be efficient. Do we not know yet or what are we saying that tells us it is or isn't?

Brian Feinstein

I would point to a few things. One, these businesses are growing really, really fast. Just as fast as software businesses. Historically, I would argue faster than some of the best software businesses we've seen. We've seen a 400% growth rate at scale of some early businesses here.

Maybe taking a step back, if you think about the criticisms, let me outline the criticisms associated with services businesses from a venture perspective, and then what's different about these AI-enabled services. Criticism number one, hard to grow really fast organically with a services business. Your product is undifferentiated and you're constrained by your ability to hire labor.

These AI-enabled services businesses are growing really fast, most of them growing north of 100% a year because they have a differentiated product. They're better and cheaper, significantly better and cheaper than the traditional labor model. They're not constrained by people because most of the work is getting done by LLMs and software.

That's criticism number one mitigated.

Criticism number two is they have low margins. They're not that profitable.

Kent

You hear what people talk about, right??

Brian

Yes. 30, 40, maybe 50% gross margins, best case scenario. That's not what we're seeing with these AI-enabled services businesses. Not only are they better and cheaper, but they're also able to deliver better and cheaper value to their customers at 60 70 plus % software-like gross margin.?

Kent

State the obvious the model costs. They all have model costs, and there was a little bit of a chatter around how much were they a thin wrap around just a bunch of model cost. But the model costs are a fraction of of the cost basis.

Brian

Correct. It turns out that when you add workflow and know-how and go-to-market and customer success and last-mile operations on top of the LLM, you can capture a ton of value. Most of our AI-enabled services businesses are operating at 60 plus % gross margins, and those gross margins are only getting better.

Then finally, the complaint about services businesses is that they're undifferentiated. You're just hiring people and then adding a markup.

It goes without saying that these AI services businesses are highly differentiated because you can't just use OpenAI and add some bodies.

You need to build a lot of custom functionality, workflows, processes, expertise to deliver a service that is better and than the incumbents.

#Bessemer Venture Partners

#VerticalAI