Vertical AI (Platform -> Competitor), Voice AI & Dario Speaks
Capital Efficient #10
Welcome to the latest edition of Capital Efficient. Let’s get into it.
Weekly Radar
Compete vs. Enable
I’ve been thinking more about what vertical AI’s form factor is as the cost of code declines. I see two paths for founders building venture-backable businesses in vertical AI.
Path 1: Build the platform. Get to a category early, get some heat on you, and try to build one of the breakout vertical AI platforms. This generally means selling high ACV software that is “AI-native,” offering agents and copilots that make workers more efficient. A big reason for why these companies command such rich valuations is that the efficiency (i.e., labor replacement) they offer will tap into human capital budgets and drive outsize ACVs. Examples: Harvey, Hebbia, Legora.
Path 2: Be the competitor. If a vertical AI platform can truly drive outsize efficiency, why not just compete directly? A few reasons this is increasingly attractive:
The end-market is much larger. Compare all legal spend to all legal IT spend. This also gets around ACV sticker shock. The current vertical AI bet is that it taps into human capital budgets and can charge massive ACVs over time, but there’s a psychic limit that will take time to overcome for what people want to pay for software in an old-line business. Going direct sidesteps this entirely.
You hide your technical asset. If every law firm is using Harvey, they’re all equally productive. If you build your own secret sauce and compete directly, you could outcompete these AI-enabled firms by being AI-first. Your innovation is your moat, not a product you’re selling to everyone.
The risk cuts both ways. Yes, it’s risky to compete directly. But you can build a bigger top-line business, tap into markets that already have a kingmade vertical AI platform, and here’s the thing: if the AI isn’t good enough to compete directly, it means the labor replacement isn’t real, the ACVs never materialize, and none of these vertical AI platforms get over their pref stack anyway. You can’t have it both ways.
“But what about margins?” If AI can truly automate the work, an AI-native services business should have margin profiles approaching software, not traditional services. And if it can’t, the entire vertical AI platform thesis falls apart anyway, because the labor replacement that justifies those massive ACVs was never real. Same labor, same automation, different business model.
And the market structure is more forgiving. Vertical software usually sees clustering around one or two category winners. If you’re the 10th-best vertical AI platform in a category, you’re dead. But if you build an AI-native insurance wholesaler and come in 10th, you’re still doing hundreds of millions in revenue.
Examples of what I’m talking about:
Build a staffing agency vs. sell recruiters AI sourcing tools
Start a law firm vs. sell lawyers AI copilots
Run a creative agency vs. sell marketers AI content generators
Offer accounting services vs. sell accountants AI bookkeeping software
Start a market research firm vs. sell analysts AI research assistants
In each case, the AI-native competitor captures the full value of the efficiency gain rather than selling it to incumbents for a fraction.
With the cost of code falling, we’ve seen VCs gravitate away from SaaS and towards defense tech, robotics, and neo-labs. I think the next phase of the cycle is founders building competitive businesses, not software providers, across a whole range of categories. Think of it as a return to the early days of venture backing companies like FedEx: real businesses with real operations, not just a software layer. I’m personally excited for this next wave of AI businesses, and if you are building one, get in touch.
Deals
AI meets RIA: Zocks, an AI assistant for financial advisors, announced a $45MM Series B led by Lightspeed and QED. This comes on the heels of direct competitor Jump raising a $20MM Series A led by Battery last year. A lot of big vertical AI businesses like Zocks are going to start as AI notetakers and spiral outwards from there. We’ve seen this playbook work in healthcare with Abridge, which started as a clinical documentation tool and expanded into a broader AI platform for health systems. As the gen 1.0 vertical AI companies mature, expect to see a clustering around one or two industry defaults in the same way we saw in vertical SaaS (i.e., if you are a GC, you are on Procore or Autodesk Construction Cloud).
Voice AI for Field Sales: Bowery’s first Voice AI investment is now officially out of stealth. Enata is building a second brain for field sellers. Enata lets field reps capture visit notes and customer context by voice, building a living record of relationship history across their territory so nothing slips through the cracks. Unlike coaching tools designed for compliance monitoring, Enata is built for high performers who want a second brain to cover more ground. With a founding team (Justin, Marty, Smit) that combines technical expertise and field sales know-how, we couldn’t be more excited to partner with them. We led their first round of funding with our friends at Breakers VC and SV Angel. Check out their launch video below.
What I’m Reading
The Adolescence of Tech: The Oracle has come down from the mountain and it’s worth reading. Dario’s implication is that the fast takeoff just got faster and we should brace for impact (or liftoff, if we take the right steps now). If you are in any way interested in AI and its implications, this is a must read. This one’s long, so I’m not going to recommend anything else for this week, lest readers get distracted.