On April 14, 2026, HubSpot stopped charging for its Customer Agent by the conversation and started charging by the result. The old model billed $1.00 every time the agent picked up a chat. The new one bills $0.50, but only when the agent actually resolves the issue. Conversations that go nowhere now cost nothing. HubSpot says the agent resolves 65% of conversations and cuts resolution time 39% across 8,000 customers, and its pitch to buyers was blunt: you were tired of paying for AI potential instead of AI performance.
That single change is a preview of a problem landing on every SaaS renewal this year. The unit you buy software in is shifting from the seat to the outcome, and the budgeting model that finance and IT have used for a decade was built entirely around seats.
The repricing is already industry-wide
HubSpot is not an outlier. Across customer service and CRM, the vendors that dominate your SaaS stack have all moved, or are moving, off flat per-user pricing for their AI agents:
- Intercom Fin charges $0.99 per resolved conversation and refunds up to $1M a year if it misses agreed resolution targets. Fin now closes roughly a million conversations a week and carried Intercom from $1M to $100M in ARR on that model alone.
- Salesforce Agentforce launched at $2.00 per conversation, then rolled out Flex Credits in late 2025 to bill per action instead: about $0.10 for a standard action, sold in blocks of 100,000 credits for $500. Salesforce now recommends Flex Credits for most new deployments.
- Zendesk priced its autonomous resolutions at $1.50 per committed resolution, or $2.00 pay-as-you-go, when it announced the model at its Relate 2026 conference.
- ServiceNow Now Assist takes a 50% to 60% uplift on the existing base tier plus per-token consumption of roughly $0.015 to $0.04. It generated more than $600M in net new annual contract value in FY25 and is tracking toward $1B in FY26.
- Microsoft is running both models at once. Microsoft 365 Copilot still bills $18 to $42.50 per user per month in pure seat pricing, while Copilot Studio, where customers build their own agents, runs on consumption credits at $0.01 each or $200 per 25,000. Microsoft is collecting seat revenue and consumption revenue from the same accounts, and plenty of buyers do not yet realize the second meter is running.
The logic behind all of this is simple. When one AI agent does the work of five support reps, the vendor cannot keep selling you five seats. The value it delivers is no longer tied to how many people log in, so the price is being unbundled from headcount and re-attached to work performed. Gartner projects that 40% of enterprise SaaS spend will shift toward usage or outcome-based pricing by 2030. Today the transition is early: one 2026 survey found only 9% of companies have fully implemented outcome-based models, with another 47% actively piloting them, and hybrid structures (a base subscription plus a variable meter) are the most common landing spot, used as the primary model by around 37% of software vendors.
Why this breaks the budget, not just the invoice
I spent 20 years in IT operations before moving into fractional COO work, and for most of that run the SaaS budget was the easy part of the plan. You knew your headcount, you knew the per-seat rate, you multiplied, you added a contingency for growth, and you were within a few percent by year end. Software was a fixed cost that behaved like rent.
Outcome-based pricing turns that fixed cost into a variable one that tracks a number nobody in finance controls: how much work the tool does. Your Intercom bill next quarter depends on how many support tickets your customers open. Your Agentforce bill depends on how many actions your agents fire, which depends on how aggressively your revenue team leans on automation. None of that lives in a headcount plan. It lives in demand, and demand is exactly the thing annual budgets are worst at predicting.
The FinOps Foundation’s 2026 State of FinOps report found that 90% of FinOps teams now manage SaaS spend, up sharply as software costs started behaving like cloud costs. That is the real story here. Seat-based SaaS was a procurement problem: negotiate the rate, count the seats, renew once a year. Consumption-based SaaS is a FinOps problem: it needs the same visibility, allocation, and forecasting discipline that cloud bills demanded a decade ago, because it is now variable, usage-driven, and capable of surprising you mid-quarter.
Three specific things break:
Forecasting. A per-seat contract is a line you can draw a year out. A per-resolution contract is a curve that bends with usage. If your support volume spikes 30% during a product incident, so does your Fin bill, in the same month you can least afford a surprise.
Allocation and chargeback. Seat licenses map cleanly to teams and cost centers. Consumption does not, unless you instrument it. When the marketing team’s prospecting agent and the support team’s resolution agent both draw from the same Salesforce credit pool, splitting that bill back to the right budget owner requires tagging and metering you probably have not built yet.
Procurement leverage. The annual seat negotiation was your one moment of leverage. In a consumption model, the meter runs continuously and the “negotiation” is really about credit-block discounts, rate floors, and what counts as a billable outcome. That last point matters more than it sounds.
Outcome-based is not automatically cheaper
The marketing frames outcome pricing as pure buyer protection: you only pay when it works. Sometimes that is true. Intercom’s $1M refund guarantee is real money on the table. But do not confuse aligned incentives with lower cost.
Two traps are already visible. First, the definition of “outcome” belongs to the vendor. A “resolved conversation” is whatever the contract says it is, and a system optimized to maximize billable resolutions is not the same as one optimized to help your customers. Watch for agents that close tickets fast to book the resolution, then generate a second contact you also pay for. Read the definition clause the way you would read an SLA, because that is what it is.
Second, the total cost of ownership often rises before it falls. Analysis of Now Assist and similar deployments points to year-one TCO increases north of 30% before optimization, driven by setup, integration, and the simple fact that a working agent gets used more than the tool it replaced. The seat you retired was capped. The meter you turned on is not.
What to actually do before your next renewal
You do not need a new platform to get ahead of this. You need to treat variable SaaS the way a mature FinOps team already treats cloud.
Start by finding the meters you have already switched on. Anywhere you have an AI agent from HubSpot, Salesforce, Intercom, Zendesk, ServiceNow, or Microsoft, confirm whether it bills by seat, by consumption, or both, and pull three months of actual usage. The Copilot Studio credit meter is the one people miss most often.
Then forecast the variable line the way you forecast cloud: off a usage driver, not a headcount. Tie the Fin estimate to projected ticket volume, tie the Agentforce estimate to projected actions, and put a range on it, not a point number. Set an alert threshold on each meter so a mid-quarter spike reaches finance before the invoice does.
At the negotiation, spend your energy on three clauses: the exact definition of a billable outcome, the rate floor and credit-block discounts at your expected volume, and a cap or true-up mechanism so a demand surge cannot run unbounded. If a vendor will not define the outcome precisely, that is your answer about how the meter will behave.
The seat-based SaaS budget had a good run. It survived the shift to subscriptions and the shift to the cloud. It will not survive software that charges for work instead of access, and the vendors have already made the call. The teams that adapt first will be the ones who stopped treating SaaS as rent and started treating it as consumption, before the meter taught them the hard way.
Sources: HubSpot company news, MarTech on HubSpot’s outcome-based pricing, FinOps Foundation State of FinOps 2026, Gartner outcome-based pricing projection via industry analysis, Intercom Fin resolution pricing overview.