Gartner has put a hard number on what agentic AI could do to enterprise software budgets: $234 billion in SaaS spending is at risk, according to the analyst firm's assessment reported by CIO.com.

The figure matters because it targets the foundation of the modern SaaS business model. Enterprise software vendors have spent two decades pricing on human seats — licenses, subscriptions, and per-user fees that scale with headcount. Agentic AI attacks that logic directly. If software increasingly executes work rather than assists people doing it, the unit of value shifts from the employee account to the task, the workflow, or the outcome. Seat counts stop being the denominator.

For product managers at SaaS companies, the threat is not abstract. An agent that orchestrates work across a stack can flatten the middle layers of that stack — dashboards, data-entry tools, point solutions whose entire job was to help a human operate a process. When the human leaves the loop, tools built for human-in-the-loop usage become candidates for removal, not renewal.

Where the pressure lands hardest

The exposure is uneven across categories. SaaS products that primarily package a workflow — form-driven systems, reporting layers, manual handoff tools — sit closest to the risk line, because an agent can replicate or bypass the workflow without the product. Platforms holding proprietary data, deep integrations, or regulatory moats face less immediate displacement, though Gartner's framing suggests no category is fully insulated.

The conditions matter. This scenario assumes agentic systems mature enough to run reliably in production, integrate without heavy human supervision, and win enterprise trust on security and auditability. Where agents remain brittle — ambiguous tasks, high-stakes decisions, compliance-heavy processes — humans stay in the loop, and seat-based SaaS retains its grip. The $234 billion is at-risk spend, not a forecast of certain cancellation.

What breaks on the vendor side

The failure modes are worth naming precisely. A vendor that reprices from seats to usage without re-architected metering will burn margin or customer goodwill, often both. One that waits risks a competitor or a customer-built agent commoditizing its category first. And vendors that bolt an "AI assistant" onto a seat-priced product have defended the interface, not the value metric — customers eventually notice they are paying per human for software whose output no longer depends on human count.

On the buy side, procurement teams now have a credible analytical basis to interrogate renewal pricing. If Gartner says nearly a quarter-trillion dollars of SaaS spend is structurally exposed, expect CFOs to ask product teams to justify seat counts agent by agent.

The practical move for PMs is metric honesty: identify which parts of your product assume a human operator, estimate revenue attached to those parts, and pressure-test an outcome- or usage-based price against your top decile of accounts. That exercise converts an analyst headline into a roadmap decision.

Gartner's number signals where product practice is heading: pricing anchored to work delivered, not people served — and vendors that wait for the renewal cliff to test that model will test it under the worst possible conditions.