G2 finds only 9% of buyers call seat pricing fair for AI agents, and 74% are ready to shop a rival at renewal

G2's new 2026 AI Agents Report finds that just 9% of B2B software buyers consider per-seat pricing fair for AI agents, a signal that the licensing model most SaaS companies still default to is losing legitimacy with the people who sign the renewal.

Usage-based pricing is winning the argument

The report, titled "Crossing the Agentic Chasm," is built on an online survey of 1,019 B2B decision-makers who influence software purchases, fielded in July 2026 across North America, EMEA and APAC, spanning companies from under 250 employees to over 5,000. G2 published it on October 14 alongside an analysis from G2 chief innovation officer Tim Sanders, and SaaStr founder Jason Lemkin wrote his own breakdown of the pricing chapter the same week.

The shift in buyer preference is sharp. Pay-as-you-go pricing, covering consumption and per-completed-task billing, rose from 38% of buyers' stated preference in 2025 to 57% in 2026. Pay-per-outcome pricing rose from 11% to 19%. Fixed pricing fell from 35% to 21% over the same period. Adoption itself climbed too: AI agent purchases were up 25% year over year, and 66% of buyers now have agents in limited or scaled production rather than a pilot.

That adoption is outpacing cost control. 46% of buyers report higher-than-expected operating costs for the agents they've deployed. The report lists the controls buyers want before they'll accept variable billing, in order of priority: real-time usage dashboards, monthly spend caps, cost-per-outcome reporting, the ability to route work to cheaper models, and cost forecasting before a deployment goes live.

Why seats stopped making sense for agents

Per-seat pricing was built for software a human logs into. Lemkin's framing of the problem is the simplest explanation for why that logic breaks on an agent: "A seat counts the humans who log in." An agent isn't a human who logs in once and works inside a session; it runs continuously, in the background, sometimes overnight, and the bill is tied to what it did rather than who opened the app. Buyers' own stated autonomy preferences back this up: only 7% want agents running with total autonomy, 48% want them operating inside guardrails, and 29% want them limited to specific, named tasks. A seat count doesn't capture any of that, which is likely why the fairness score for seat pricing sits at 9% even as overall agent adoption keeps climbing.

Renewal is turning into a competitive pitch

The report's renewal numbers are where the pricing complaints turn into churn risk. 30% of buyers say they would consider switching to a vendor with stronger agent capabilities, and another 44% say they would pilot an alternative and reconsider at renewal time, a combined figure SaaStr's Jason Lemkin rounds to 74% in his write-up. Separately, 84% of buyers expect agent capabilities to influence which vendor they choose in 2027, and 34% already call it a primary deciding factor rather than a tiebreaker.

"Before buyers will accept rising costs, they expect the controls that make spend visible and manageable."

That line, from G2's report, is really a description of the sales conversation most AI-agent vendors are now having at renewal: cost justification before expansion. 72% of buyers rate a strong ecosystem of connectors, actions, skills or marketplace integrations as very important or a must-have, which raises the bar further for vendors selling a standalone agent rather than a platform.

Lemkin's own read of the data, published on SaaStr, is blunter about what buyers will tolerate: "They will not accept a bill that tripled because an agent looped all weekend."

What this means for the pricing page

Marketing teams that still lead with "per seat" on the pricing page are selling against the buyer's own stated preference, not just a competitor. The fix isn't necessarily abandoning seats, since many software products still have a genuine per-user value story, but it does mean the agent line items need their own logic.

• Publish usage or outcome-based pricing for agent features as a named, visible option next to seats, not a custom-quote conversation
• Build spend visibility into the product marketing, not just the product: dashboards and spend caps are now buying criteria G2 says buyers rank above price itself
• Brief renewal-facing marketing and sales content on agent capability specifically, since 84% of buyers say it will shape 2027 vendor choice
• Audit competitor agent ecosystems before the next renewal cycle, since 72% of buyers weigh connector and integration depth as a must-have

The underlying pattern is one software marketers have seen before with any metered feature: buyers will pay for usage they can predict and refuse to pay for usage they can't see coming. G2's numbers suggest most software companies haven't rebuilt their pricing communication around that yet, even as their own buyers say they're ready to shop elsewhere.