PART I: The Market Shift and What it Demands Operationally
You’ve seen the headlines. The $285 billion selloff. The “SaaSpocalypse.” Jefferies traders coining new vocabulary for the carnage. The noise level is high enough that adding to it feels like a waste of your time and mine. So we won’t.
What we will share is our honest view on where the real opportunity sits in this shift, and specifically what Vertical AI companies need to do right now to be on the winning side of it.
The Selloff Got One Thing Right
To be fair to the bears, the specific fears driving the selloff deserve a fair hearing. Steel-manning them:
- AI labs are actively moving up the application and action layer. Anthropic’s legal and cybersecurity plugins demonstrated this when they sent specialized software stocks sliding.
- Near-zero-cost “vibe coding” and homegrown AI tools are beginning to erode demand for some categories of third-party software.
- New entrants can now replicate product features in days rather than months, compressing competitive advantages that used to take years to build.
- The mismatch between AI agent-based value delivery and legacy seat-based pricing models is creating real revenue uncertainty across the industry.
All four fears trace back to a single underlying threat: the barrier to building and executing plain vanilla workflows is approaching zero. If your product is, at its core, a model plus a wrapper plus a workflow, the moat was never structural — it was temporal. Anthropic didn’t need a multi-year enterprise sales cycle to take a run at Thomson Reuters. It needed a few weeks and a GitHub repo.
But “Vertical AI” Is a Different Conversation
Pure workflow wrappers and thin AI apps are not the same as purpose-built Vertical AI companies serving specific industries with deep workflow integration, proprietary data, and real switching costs. What protects a true Vertical AI company compounds over time:
- Brand trust earned through consistent, accurate performance in high-stakes workflows
- Deep semantic understanding of what domain, knowledge that takes years of iteration to build
- Best practices that encode not just what to do but the right way to do it for a given industry
- An ecosystem moat built through channel partners, system integrators, and industry analysts
- Deep data plumbing and integration with the legacy systems that already anchor customer operations
- Last-mile product efficacy that horizontal platforms cannot match without years of domain-specific investment
The threat is real but asymmetric. Anthropic has signaled genuine intent to compete in legal; a vertical AI company serving CPG finance planning or industrial supply chain faces a different picture. “Microsoft or Google could build this” is almost always technically accurate. The more important question is whether they will, and whether their generic version will ever match a purpose-built vertical platform.
McKinsey has sized the new value creation potential from the AI-driven “execution layer” of software at $2.9 trillion. The best-positioned Vertical AI companies are still commanding premium valuations despite the broader selloff. Winners here don’t just survive the selloff narrative. They get re-rated above it. Capturing a piece of that prize requires getting several things right.
What It Takes to Win
Lay the Groundwork on Data
AI value is downstream of data quality. The companies winning this race are aggressively capturing customer and industry context competitors can’t easily replicate. The ones that defer it are creating a ceiling on what their AI can actually do. Key priorities:
- Clean data plumbing (definitions, lineage, permissions) before you can build agents that do meaningful work
- Continuous capture of operational context, not just transactional data
- Treat this as foundational infrastructure, not a future roadmap item
Get Your Internal Organization Ready
Skills are misaligned. Incentive structures still reward legacy processes. We see it in nearly every company we look at. For earlier-stage companies, the immediate priority isn’t wholesale restructuring:
- Every employee should be using AI to its current potential in their functional area, as even modest productivity gains compound
- Set a concrete goal: 1.5x individual productivity, and measure traction against it
- Adopt new KPIs to measure AI’s impact internally, not just within the core product
Evolve the Business Model
The subscription seat model made sense when value was delivered through user activity. It maps poorly to a world where agents do the work. Founders who get ahead of this will find it easier to capture the value they’re delivering. The emerging alternatives:
- Usage-based and outcome-based structures tied to what the AI actually delivers
- AI-FTE-based and unit-based models that reflect agent-driven value creation
The right answer varies by the type of value you’re creating. There’s no universal replacement for seat pricing.
In Part II, we’ll cover the four components of durable moat for Vertical AI companies and why the window to establish that position is measured in quarters, not years.