What buyers are actually doing
Talk to enterprise buyers about AI right now. They're inundated with offerings and can't evaluate them. They take the demo. They don't buy.
The data lines up. 79% of organizations report challenges adopting AI, up double digits from 2025. Only 29% see significant ROI from generative AI, despite real productivity gains at the individual level. The market is overwhelmed, not underserved.
This isn't a buyer-quality problem. It's a structural one. The pace of new AI offerings has outrun the capacity of any buyer to evaluate them. A typical CIO sits through five or ten AI pitches a week and most of them sound the same. The differentiation that matters to the founder isn't legible to the buyer. So they freeze.
Why horizontal compounds the problem
Horizontal AI makes paralysis worse for three reasons.
First, the pitch sounds like the last ten. Generic productivity gains, generic workflow improvements, generic ROI promises. Without a specific business process to anchor against, the buyer can't tell which tool is right for their organization.
Second, the integration burden falls on the buyer. Horizontal tools require the customer to figure out which workflows they apply to, which teams should use them, and how to connect them to existing systems. Most buyers don't have the technical depth or the change-management bandwidth to do that on their own. ROI sits 18 months out, by which point the budget is gone and the political capital with it.
Third, and most importantly, horizontal AI faces a durability problem that has changed in the last year.
The frontier-lab absorption risk
Frontier labs are natively absorbing horizontal use cases. Memory, canvas, file analysis, computer use. Anything that was a thin wrapper on top of a foundation model last year is a native feature of the foundation model now. Building horizontal AI means competing with OpenAI on its own roadmap.
Buyers know it. And the implication for their decision changes everything.
Enterprise buyers aren't choosing AI tools the way they chose SaaS tools a decade ago. They're making infrastructure decisions they expect to hold for ten to twenty years. The cost of switching, retraining, and reintegrating is high enough that they treat each adoption decision like a long-term commitment.
In that frame, adopting a horizontal AI tool that gets absorbed by a foundation lab within a year isn't just a wasted purchase. It's a career-level risk. The buyer who picked it has to explain to the CEO why they bought a tool that became obsolete in twelve months while the foundation model their company already pays for shipped the same capability for free.
So they wait.
Why vertical works
Vertical works because it removes most of these problems at the source.
The buyer recognizes the problem before the demo starts. They've been trying to solve it for years and they know what a working solution looks like. The pitch is in their language. The integration is shorter because the workflow is constrained. The ROI is legible within 90 days.
The data backs this up. Industry-trained systems hit 40% higher task accuracy in regulated workflows than horizontal alternatives. And the frontier labs aren't shipping dental practice management software or ad operations workflows. Vertical builders work with the model layer, not against it.
This is what we wrote about in The Moat Just Moved earlier this year. As foundation models eat into the application tier, the durable layers shift to the places models can't reach: deep workflow knowledge, regulated data access, vertical-specific compliance, and the trust relationships that come from solving a specific buyer's specific problem. Vertical AI lives in that space. Horizontal AI competes with the model layer for the same surface area.
Swivel is the version of this we see in our portfolio. Vertical AI for ad operations. Didn't start horizontal and pivot. Built deep into the workflow from day one. Buyers in that category know what an ad ops AI is supposed to do before the meeting starts. Sales cycles are short for the size of company they target, integrations are clean, and the ROI shows up in the first quarter of deployment.
What to do if you're already horizontal
This doesn't mean horizontal is dead. If you have conviction around a horizontal solution but are struggling to get traction with customers, lead heavily with a vertical use case and use it as a wedge to get the sales motion going.
The horizontal vision can be the destination. The vertical wedge is the path that actually starts moving.
Good wedge execution looks like this. Pick the one industry where the pain is most acute and where your product solves it cleanly. Build the version of your product that solves that problem end to end, not the abstract horizontal version. Win that buyer, generate the proof point, and use the case study to expand into adjacent industries. The horizontal capability stays in the product. It just isn't the lead.
Superposition is the canonical version of this. Recruiting is one of the largest AI categories right now. Most companies in the space went broad. Superposition started with one ICP: founders hiring for founding teams. The candidate sets are small enough to evaluate, the buyer is the founder, and the workflow is consistent across customers. That focus made the agent work. Most competitors found it harder to sell and harder to build. Superposition got both right because they narrowed first.
The product can extend to other roles, other company sizes, other ICPs over time. Narrowing first is what made the agent work and what gave the buyer something they could evaluate.
The pattern across our deal flow
Founders leading with vertical depth get traction. Founders leading with horizontal capability get meetings. Both can be technically excellent. One is converting.
This isn't a critique of horizontal founders. The technical work is often more impressive. The vision is often larger. But when a buyer can't evaluate which tool is right for them and is afraid the one they pick will be absorbed by OpenAI in twelve months, capability isn't enough. The founders winning right now are the ones who solved a specific problem for a specific buyer in a specific way that the foundation labs aren't going to ship.
The companies trying to sell the horizontal vision from day one are competing for attention with every other AI startup in the market. The companies leading with a vertical wedge and expanding from there are compounding from a position of trust.
In the current market, the second path is the one that converts.