Essay

AI is a business model reset, not a feature roadmap

May 2026 . 8 min read

Most B2B software CEOs are still answering the wrong question about AI. They are being asked, by their boards and their markets, what AI features are shipping this quarter. They should be answering what AI is doing to the business model that ships them. The two questions sound adjacent. They are not. One belongs on a product review. The other belongs on a board agenda, because the answer rewires pricing, margin structure, cost-to-serve, go-to-market design, and the location of strategic control. Treating AI as a feature cycle is the most expensive misread of the decade for SaaS-era incumbents.

The on prem to SaaS growth curve rewarded one set of behaviours

Classic SaaS economics rewarded a very specific operating model. Build a multi-tenant product, price it per seat, expand inside the account through access and adjacency, hide product gaps with services, and underwrite a long, predictable revenue tail. Boards trained their pattern recognition on that shape: net retention above 110, gross margin around 80, payback inside 18 months, rule-of-40 in the 30s, services kept under 25 percent of revenue (ICONIQ, 2026). That model worked because each new customer was effectively free to serve at the margin. The cost to host one more user was rounding error. Incremental seats compounded into operating leverage. The first growth curve was a leverage story.

The hidden break is on the cost line, not the demand line

The AI wave does not break demand. It breaks the cost line that made the leverage story work. Inference is not free. ICONIQ data through Q1 2026 puts the average AI product gross margin at 52 percent against the SaaS benchmark of 80 percent, with inference alone consuming roughly 23 percent of revenue at scaling-stage AI B2B companies (ICONIQ, 2026, via Monetizely, 2026). Public SaaS companies disclosing AI-driven margin pressure have begun reporting gross margins 10 to 17 points lower than pre-AI baselines, with inference-cost ratios typically running 4 to 9 percent of revenue and called out separately in MD&A (SaaS Mag, 2026). The 80-point gross margin that defined SaaS is no longer the operating range. The new range looks like 60 to 70.

That changes everything downstream. Payback math, rule-of-40 targets, expansion economics, services-to-product ratios, and headcount plans were all calibrated to an 80-point margin world. They have not been recut. Most boards are still reading the old dashboard.

What AI changes in the value chain

AI changes three economic variables at once. Cost-to-serve goes up because inference is variable and meters with usage. Pricing power shifts because buyers stop paying for access and start paying for outcomes. And the unit of value moves from the seat to the resolved task. None of these is a feature decision. Each is a business-model decision.

The pricing shift is already visible in the market. Per-seat pricing collapsed from 21 percent to 15 percent of SaaS in the twelve months to early 2026, while hybrid models reached 41 percent adoption (Monetizely, 2026). Eighty-three percent of AI-native SaaS now bills on usage rather than seats (Stormy AI, 2026). Gartner forecasts that 40 percent of enterprise SaaS will move to outcome-based billing by 2030, but the operators making that move now are doing so because the cost structure compelled it, not because the forecast suggested it (Bessemer Venture Partners, 2026). Zendesk now charges for AI-resolved tickets rather than AI seats. Intercom prices its Fin AI agent the same way (Stormy AI, 2026).

The replacement of labour is being priced as labour, not as software.

Behind those numbers sits a quieter shift: the buyer’s question has changed from “how many people will use it” to “how much work will it do.” A CEO whose entire commercial model still assumes the first question is selling a product the buyer is no longer trying to buy.

The downstream consequences are operating, not technical

Once the cost line, the pricing model, and the unit of value all move at once, the operating model has to follow. Five consequences stand out.

First, services stop hiding product weakness. SaaS-era companies have long absorbed product gaps through services revenue, often pushing the services line above 30 percent. AI products expose those gaps faster because customers measure outcomes directly. The services-to-product ratio must compress, and what survives in services is implementation and integration, not feature compensation.

Second, the renewal economy is reset. In 2025, most AI deployments operated under “adoption at all costs” pricing with minimal price sensitivity. In 2026, those contracts enter their first renewal cycle, and customers are pricing what they actually use, not what they bought (Stormy AI, 2026). Net retention will compress for any vendor whose pricing did not move with usage. That compression will look like a sales execution problem and will actually be a pricing-model problem.

Third, the go-to-market motion compresses. Outcome-priced products are easier to start and harder to expand inside, because expansion now follows resolved work rather than headcount creep. Land-and-expand becomes land-and-prove. The companies that win the second curve are rebuilding sales compensation, success motion, and forecasting around outcomes per account, not seats per account.

Fourth, capital allocation has to recognise inference as a margin call on every dollar of revenue. McKinsey reports 5.8x average ROI on AI investment within 14 months of production deployment, but only 29 percent of executives see significant ROI from generative AI, and 79 percent of organisations report challenges adopting it (McKinsey, 2025; Writer, 2026). The gap between the average and the median is the cost of letting AI investment run as a feature programme rather than an operating-model reset.

Fifth, the board agenda changes. The right board questions are no longer “what is our AI roadmap” and “are we shipping fast enough.” They are “what is our AI-adjusted gross margin,” “what is our inference cost as a percent of AI revenue,” “what is our pricing exposure at next renewal,” and “what is our outcome-priced revenue share.” A board that is not asking those four questions is still on the first curve.

Where the lazy reading goes wrong

The comfortable reading inside many SaaS C-suites is that the margin pressure is temporary, that inference costs will fall fast enough to restore the 80-point gross margin, and that the pricing reset is a niche issue for agent vendors. None of that is quite right. Inference cost per token has fallen, but customer expectations of capability have absorbed the savings. Most operators report margin compression as the structural consequence of making AI features core to the product, not a temporary glide path (SaaS Mag, 2026). Pricing models inside the agent layer are setting the expectation for the rest of the stack. Treating either as transient leaves the company exposed at the moment its renewals reprice and its margin assumptions stop matching its disclosures.

There is a sharper version of the wedge here. A roadmap is not a strategy. An AI feature list, no matter how complete, is activity that signals motion to boards and markets while leaving pricing, margin, and the location of value capture untouched. That is AI roadmap theatre. A board that asks for the AI roadmap and stops there is asking the wrong question.

The board agenda for the second growth curve

Five strategic choices are now on the desk. Each is a business-model decision, not a product decision.

One. Decide where the unit of value sits, and reprice to it. Seat, usage, outcome, or hybrid. Pick deliberately, and underwrite the gross margin implication.

Two. Recut the operating model to a 60 to 70 point gross margin world. Recalibrate rule-of-40 targets, payback assumptions, services share, and headcount plans to the new range.

Three. Sequence the reset: pricing, then go-to-market, then product boundaries. Most companies do this in reverse and discover their pricing was the bottleneck after the product change has already shipped.

Four. Replace the AI roadmap review with an AI economics review. The standing board item should track AI-adjusted gross margin, inference cost as a percent of AI revenue, pricing exposure at next renewal, and the outcome-priced share of new ARR.

Five. Identify the single AI bet that resets the curve. One, not five. The companies that win the second curve are concentrating capital behind one model-level repricing or workflow-collapse bet, not spreading it across a feature backlog.

The question that should replace “what is our AI roadmap” is simpler and harder: what is our AI-native business model, and which line of the P&L does it move first. A CEO who can answer that question on one page can run the second growth curve. A CEO who cannot is still running the first.

Sources

AI-native, SaaS, pricing, margin, boards

Back to writing