The $700 Billion Displacement: Why Outcome-as-Agentic-Solution Will Hollow Out the SaaS Market

The software industry sells access. It is about to discover that access was never what customers were buying.

The thing that was bought, in every successful SaaS deployment of the last fifteen years, was an outcome. A verified identity. A cleared transaction. An approved loan. A closed ticket. The seat was a proxy, because access to the tool was the rate-limit that determined how many outcomes a customer could produce. Pay for more seats, get more outcomes. The fiction held because the proxy was tight enough to function.

The fiction is breaking.

A CIO told me last month that his AML platform costs more per year than the team that operates it. The team that operates it costs more than the team that built his core banking system. He said it the way someone tells you that the boiler has been making a noise. I have started counting how often I hear that shape of sentence.

The unit of value is breaking

What is wrong in that CIO's budget is not SaaS as a category. SaaS as a category has been useful. What is wrong is the unit of value. We have spent fifteen years paying for access to tools while the work those tools were supposed to enable has remained human, expensive, and slow. The licence fee was never really the cost. It was the entry ticket to a longer line.

Per-seat pricing made sense in an era when access to a tool was the rate-limit. If a bank wanted more onboarding capacity, the binding constraint was analyst headcount, and the SaaS provider was paid in proportion to that headcount because that was, approximately, the shape of the value being delivered. The provider sold access. The customer hired the people. Everyone agreed on the fiction that the licence was the product.

Agentic systems remove the rate-limit. Once an autonomous agent can plan, call APIs, and complete a multi-step workflow inside the buyer's policy envelope, the seat is no longer the binding constraint. The outcome is. And outcomes price differently from seats.

This is the thesis. The next five years of enterprise software will be a slow, expensive, inevitable transition from charging for tool access to charging for completed work — and the SaaS industry's pricing strategies are not prepared for it.

Why the seat metric is going to fail in public

A pricing model survives until someone can credibly offer the same value priced another way. The reason per-seat survived this long is not that anyone loved it. It is that nobody had a better proxy. Outcome pricing was always intellectually superior and operationally impossible — until the operational part stopped being impossible.

What I see in regulated buyers — banks especially — is a kind of forensic accounting that did not exist three years ago. Procurement teams are now asking how many of the licensed seats actually produced a decision in the last quarter. How many produced a decision that wasn't already automatable. How much of the cost is the tool, how much is the operator, and how much is the integration tax. The answers are not flattering to the vendors.

I worked recently with a regulated team that ran the numbers on a screening platform they had been licensing for six years. The platform cost was a third of the line. The operating team was another third. The third third — the bit that should not have existed — was the cost of stitching the platform's output into the bank's own decision flow. The vendor had effectively been charging for a component while the buyer paid for the system around it. Once the buyer modelled the alternative — an orchestration layer that drove the same screening providers as utilities and produced a decision instead of a feed — the licence value collapsed.

This is what I mean by "the proxy is breaking." Per-seat pricing was a useful fiction when the buyer's mental model was "I am paying for software." Once the mental model becomes "I am paying for an outcome," the same software has to compete on a different axis. And on that axis, a lot of the incumbents have nothing to say.

The orchestration layer is the new product

Outcome-as-a-service is not a new product category. It is an inversion. The interesting move is not building a single agent that performs a task end-to-end. It is building the orchestration layer that turns existing vendors into utilities and produces a delivered result.

This is good news for buyers and bad news for any vendor whose business model assumes their tool is the destination. In an outcome architecture, the screening vendor, the bureau, the KYC provider, and the document classifier are interchangeable suppliers behind a thin agentic façade. The buyer pays per verified outcome. The orchestration layer routes traffic to whichever supplier is cheapest, fastest, or most accurate for that case. The supplier's leverage drops to whatever differentiation actually moves the outcome metric.

In a few sectors this is already visible. The platforms we have built for regulated buyers do not ship a tool. They ship the result the tool was supposed to enable — an onboarding decision, a credit assessment, a verified payment instruction — and they treat the underlying vendors as plug-replaceable. The interesting line in the budget is the outcome rate, not the seat count.

Most of the SaaS market does not yet price this way. Many of them do not even instrument it. They cannot tell you, with any precision, how many seats produced an outcome last quarter, because seats were never the variable that drove their internal economics. The transition is not just a pricing change. It is a measurement problem they have never had to solve.

The counter-argument is real, and partly right

I want to take seriously the strongest version of the objection. The case against outcome pricing is that it pushes risk in the wrong direction. The buyer can verify access to a tool — the licence either turns on or it does not. The buyer cannot easily verify the quality of an autonomous outcome. If the agent makes a decision that looks correct but is wrong, the buyer pays for it and discovers the problem in a regulatory finding eight months later. Outcome pricing rewards plausibility, not correctness.

This argument is not wrong. It is the reason regulated buyers, in particular, will not transition by simply swapping seats for outcomes. They will transition by adding two things at the same time: an outcome contract on the front and an auditability discipline on the back. The vendor sells the result. The buyer measures the result. The contract reconciles the two with enough evidence to defend in front of a regulator.

This is harder than per-seat pricing. It is also the only architecture I have seen work in financial services. The SaaS vendors that survive the next cycle will be the ones that can ship not just an outcome but the receipts: the trace of which sub-system contributed, the policy that approved it, the version of the model that produced it, the human reviewer who signed off where one was required. Without the receipts, outcome pricing collapses into a procurement argument the buyer always loses.

What this means for vendors is awkward. The capability that was once a back-office concern — provenance, audit trail, explainability — has just become the part of the product that justifies the price. And most vendors have not built it as a product feature. They have built it as a compliance afterthought.

What this does to regulated buyers

For CIOs in financial services, the practical implication is sharper than the headlines. Three things follow.

First, the seat-line on the budget is going to be defended less, not more. Every renewal becomes a real conversation about what the licence actually produced. Vendors that cannot answer it in the language of outcomes will be repriced on the way down, or replaced. I do not think this is universally bad. A lot of the seat-line is genuinely wasted, and the conversation is overdue.

Second, the in-house engineering team's job changes shape. The work is no longer to integrate the vendor's tool into the bank's system. The work is to build the orchestration layer that treats the vendor as a utility. That is a different engineering brief, and a different talent profile, and most banks have not yet adjusted hiring to match.

Third — and this is the one that matters most — the boundary of accountability moves. When the vendor is selling tool access, the bank is accountable for the decision. When the vendor is selling outcomes, the accountability is shared, and the contractual shape of "what failure means" has to be renegotiated. The legal teams I have been talking to are starting to realise that the boilerplate from the SaaS era does not cover this case at all.

The transition will not be fast in the sense of "complete in eighteen months." The plumbing of regulated procurement does not move that fast. It will be fast in the sense that the margin compression on the affected vendor categories will outpace their ability to reprice. The first cohort of outcome-priced competitors does not need to take the whole market. It only needs to take enough of the high-value flow to make the per-seat economics impossible to defend at renewal. The next twenty-four months will be the cleanest test we will get, and my read is that most of the SaaS leadership I speak with has not metabolised it.

The CIO with the AML platform sent me a follow-up message a week after that first conversation. He had taken the seat-line into his next vendor meeting and asked the supplier how many of the seats had produced a decision in Q1. The supplier said they did not measure that. He cancelled the renewal the same week.

The era of selling tools to people who hire teams to operate them is not going to end with a public announcement. It is going to end with a thousand quiet renewals that do not happen, and a market that — by the time it notices — is already on the other side of a $700 billion displacement.

enterprise-ai, ai-strategy, agentic-ai, saas-displacement, financial-services