Two questions decide most pre-seed AI pitch outcomes for us. Who is your customer, exactly? And what is the single workflow you own end-to-end? If the founder answers "everyone" and "all of them," we pass. If they answer "regional HVAC owner-operators in the South" and "every inbound call from quote to scheduled job," we lean in. That's the difference between horizontal and vertical AI, and it's the most consequential distinction in agentic investing right now.
The horizontal trap
Horizontal AI products promise to do everything for everyone — write any document, summarize any meeting, code any feature, automate any workflow. The pitch is irresistible because the TAM looks infinite. The problem is that the foundation labs themselves are building the same product, with privileged access to their own models, distribution to hundreds of millions of users, and pricing power that no startup can match.
Every horizontal pitch we hear ends up answering the same question eventually: what happens when ChatGPT, Claude, or Gemini ships your feature next quarter? The honest answers are usually some combination of "we'll be faster" or "we'll be cheaper" or "we'll have better UX." None of these survive contact with a model lab that is willing to operate the feature at zero margin to keep users on its platform.
Why vertical agents win
Vertical agents commit to a specific industry, customer profile, and workflow. That commitment is the moat. The foundation labs cannot economically build a phone-answering agent for HVAC, an intake agent for dental practices, an outreach agent for staffing firms, and a billing agent for trucking companies. The integration depth, vertical-specific workflows, and customer relationships are not their game.
The four sources of vertical moat
- •Workflow integration depth — the agent is wired into the customer's calendar, payments, communications, CRM, and trade-specific tools.
- •Proprietary workflow data — every interaction makes the agent better at this customer's specific edge cases. Generic models cannot replicate this.
- •Domain-specific evaluation — the founder knows what good looks like in this industry better than any horizontal team will.
- •Vertical distribution — trade associations, franchise groups, channel partners, and warm referral networks the next vendor cannot replicate quickly.
Picking the right vertical
Not every vertical is investable. The strongest verticals share three traits.
1. Messy workflow
The current state-of-the-art is a phone, a spreadsheet, and a pile of sticky notes. If the incumbent solution is already a polished SaaS product, the vertical is harder to disrupt because the customers have already paid the integration cost once.
2. Expensive labor
The workflow today is performed by a human who costs at least $40K per year, fully loaded. Below that threshold, the math for an agentic replacement gets thin. Above it, the willingness to pay for automation is immediate and obvious.
3. Predictable outcome
There has to be a clearly measurable result — booking made, invoice collected, lead qualified, claim filed. Workflows where the outcome is ambiguous ("better strategy," "deeper insight") do not produce the outcome accountability that agentic businesses need to grow.
The unit economics gap
Vertical agents also win on the spreadsheet. Because the workflow is narrow, the agent ships faster, breaks less in production, and onboards customers in days rather than quarters. Because the customer is specific, distribution is cheaper and conversion is higher. Because the outcome is measurable, the founder can charge on outcomes rather than seats — which typically yields 2–5x the lifetime revenue of a comparable horizontal SaaS.
Typical vertical agent unit economics we underwrite to
- •ACV: $500–$5,000 per month, with outcome add-ons that can double that.
- •Gross margin: 70–85% once inference cost optimization is dialed in.
- •Payback period: under six months.
- •Net revenue retention: 110–130% as agents take on adjacent workflows.
- •CAC: under one month of ACV through vertical channels.
Where horizontal still works
Horizontal can still win at the very top of the stack — frontier models, foundational developer infrastructure, the agent operating system that other agents are built on. These are big-check, late-stage games for funds with billion-dollar pools. They are not pre-seed bets, and they are not where Fifth Turn Capital plays.
If you are starting a company today and you can choose, choose vertical. The TAM looks smaller in the deck and bigger in the bank account.
