Vertical AI doesn't replace labor. It reframes it.

The shift to Vertical AI and the conversation around it is almost entirely about TAM.

The math is genuinely staggering. NEA frames it as roughly $11T in U.S. labor spend versus ~$450B in enterprise software. Bessemer projects Vertical AI's market cap will be at least 10x the size of legacy Vertical SaaS. Menlo's read: software is finally moving from systems that record work to systems that do work. The pricing model is shifting from per-seat to per-outcome, from IT budget to payroll budget.

I agree with all of it. We are building Agentic AI environments across multiple verticals, and the budget conversations with our portfolio companies have already moved from "what does this software cost" to "what does this work cost." The TAM expansion is real.

But the TAM story is the surface story. And if we stop there, we are going to mis-position what Vertical AI actually is.

Vertical AI is not labor replacement. It is labor reframing.

The market discourse has collapsed into a binary — AI is either coming for jobs or it is not. That framing is sensationalist, fear-based, and, more importantly, wrong about how durable Vertical AI actually gets built and sold inside an enterprise. A Vertical AI deployment that functions purely as a headcount substitution play is a thin product. It will get unbundled, repriced, and outcompeted within an iteration cycle.

The Vertical AI deployments that compound do something different. They take the workflows and decision trees that consume the bulk of an employee's day — the judgment-intensive but ultimately repeatable execution layer — and they hand that time back. What employees do with the time is where the real value lives. They get to revisit the "what ifs" that have been sitting on a shelf in the vault of their mind, accumulated over years of being inside an industry. The pricing pressure they always wanted to test. The vendor relationship they always thought was structured wrong. The fulfillment workflow everyone in the industry treats as normal but they have privately known is broken.

That is not productivity. That is creativity unlocked. And it is only possible because the operator knowledge those employees carry is genuinely valuable — and was previously trapped underneath a workload that did not let it surface.

The condition for the unlock is captured context.

Here is where most of the market is still under-investing.

Vertical AI does not work — does not produce industry-specific judgment, does not earn trust, does not compound — without a structured organizational source of truth. A permissioned, function-mapped, cadenced repository that mirrors the org chart and treats context capture as a permanent operating behavior rather than a one-time documentation sprint.

Two organizations can deploy the exact same Agentic stack and get radically different outputs. The variable is almost never the model. It is the depth and structure of the context the model has been allowed to learn from. The model is the crop. The captured context is the soil. And without the soil, you are deploying horizontal AI dressed up in vertical clothing — the market will eventually see through that.

This is the point most of the "AI is taking your job" discourse misses. The organizations that successfully deploy Vertical AI are not the ones that get to fire half their workforce. They are the ones that build a culture where every function — sales, fulfillment, vendor management, compliance, finance, logistics — captures context as a normal rhythm. The agents become more capable because the institutional knowledge of the company is now structured and queryable. The employees become more valuable because their tacit knowledge is now an asset that compounds across the organization rather than walking out the door at 5pm.

This is a culture shift. It is unglamorous, deeply human work.

Why this matters for how Vertical AI defends its moat.

The Vertical AI category is already getting crowded, and the loudest defensibility narratives are the weakest ones. Proprietary models are the shakiest moat — Harvey scrapped its custom legal model in 2025 when frontier reasoning models started outperforming it on Harvey's own benchmark. Workflow embedment matters more than weights. Customer data matters more than weights. Distribution matters more than weights.

But the deepest moat — and the one almost no one is talking about as a strategic asset — is the cadence of context capture inside the customer's own organization. When a Vertical AI platform becomes the system through which institutional knowledge is structured, permissioned, and operationalized, switching costs stop being about feature parity.

The vSaaS thesis I have been writing about for the last several years has always been that the moat is not the model — the moat is the mud on your boots. Vertical AI does not change that. It deepens it. The operator-knowledge is still the fire. The Agentic layer is still the accelerant. The Source of Truth is what makes the accelerant ignite.

And the employees who carry that knowledge are not being replaced. They are being repositioned — from executors of workflows to authors of context. That is a higher-leverage role, not a lower one.

The reframe.

The Vertical AI story is genuinely about a TAM shift from software spend to labor spend. That part of the discourse is correct. But the story underneath is more interesting and, frankly, more honest.

Vertical AI is not the end of the knowledge worker. It is the moment the knowledge worker stops being valued for throughput and starts being valued for judgment, creativity, and accumulated industry intuition. The workflows go to the agent. The "what ifs" come back to the human.

That is not productivity software. That is a creativity unlock. And the companies that frame it that way — internally, externally, in their pricing, in their change-management, in their culture — are the ones that will build the durable Vertical AI moats.

The TAM is bigger because the work is bigger. The value is bigger because the humans get to do work that actually requires being human.

The moat is the mud on your boots. The Source of Truth is how you bottle it. And Vertical AI, done right, is how you finally give the operator the space to use everything they have learned.