The end of the predictable software bill
For two decades, enterprise software had a simple price, a number of users times a monthly fee. AI is ending that model, and the new AI pricing models make software spend far harder to forecast.
Look at almost any renewal in 2026 and you will find a new line next to the seats. It may be called credits, consumption, actions, or outcomes. Bain analyzed more than 30 SaaS vendors adding generative AI and found 65% layered a usage-based AI meter on top of their existing seat pricing, while the rest raised seat prices by bundling AI into higher tiers (UC Today).
The change is fast. ICONIQ’s State of AI 2026 report shows consumption pricing rising from 35% to 42% of AI products in just six months (ICONIQ). Gartner predicts at least 40% of enterprise SaaS spend will shift to usage, agent, or outcome-based models by 2030 (SoftwareSeni).
For CFOs, this turns a fixed cost into a variable one. For CIOs, it turns license management into consumption governance. Neither can afford to treat the next renewal as routine.
Why are SaaS vendors abandoning per-seat pricing?
Vendors are moving away from seats because AI breaks the link between headcount and value, and because AI features carry real compute costs that a flat seat cannot cover. Three forces are pushing at once.
AI shrinks the seat base. If an agent does the work of several people, a vendor selling per person earns less as its product succeeds. Bain’s analysis captures the risk that seat pricing makes AI vendors cannibalize their own revenue (UC Today).
AI changes vendor margins. Classic SaaS runs at roughly 70% to 80% gross margin, but every agent action consumes compute. The Growth Unhinged 2026 survey found the median vendor targets only about 50% margin on AI features, and 29% of vendors already sell AI credits with another 33% planning to (SimpleKPI).
Buyers are tired of paying for empty seats. Zylo’s 2026 SaaS Management Index puts unused paid licenses at 36% (Zylo). Customers want pricing tied to usage, and AI features add consumption in tokens, credits, and compute minutes that cannot be priced sensibly per user (Flexera).
The result is rarely a clean swap. Most large vendors keep the seat and add a meter beside it. Microsoft charges $30 per user for Copilot and meters its task-running agent on top, and HubSpot now combines seats, agent outcomes, and credits in one contract.
AI pricing models compared
Four AI pricing models now coexist, and they trade budget predictability for alignment with usage in very different ways.
| Model | How you pay | Example | Budget predictability | Main risk for the buyer |
|---|---|---|---|---|
| Seat with AI bundled | Higher per-user price that includes AI | Premium tiers that fold AI into the base plan | High | Paying for AI features many users never touch |
| Hybrid seat plus credits | Per-user fee plus a pool of AI credits | Microsoft Copilot seats with metered agent tasks; monday.com seats plus credits | Medium | Credit overages and unclear value per credit |
| Consumption | Per action, token, or API call | Salesforce Agentforce Flex Credits at $0.10 per agent action | Low | Bills scale with adoption and agent loops |
| Outcome | Per resolved ticket, lead, or task | Intercom Fin from $0.99 per resolution | Low to medium | The vendor defines what counts as an outcome |
Examples from SimpleKPI and UC Today.
The models are also unstable. Salesforce priced Agentforce per conversation in late 2024, per agent action in mid-2025, and then offered a per-user AI license a few months later after customers pushed back on open-ended bills. GitHub moved Copilot from a flat fee to usage billing in June 2026. Any multi-year plan built on today’s pricing unit should assume the unit will change.
What do AI pricing changes mean for your budget?
AI pricing models raise software spend and make it harder to forecast at the same time. Analysts call it the AI tax, and it shows up in three ways.
Higher prices on the same products. Zylo found annual SaaS spend rose 8% while the number of applications stayed essentially flat (Zylo). Vertice’s SaaS Inflation Index puts software price increases at 13.2% a year, close to five times G7 consumer inflation (Vertice). Some vendors retire legacy tiers, pushing customers into premium packages that include AI whether they want it or not (Baytech).
Consumption on top of licenses. NPI Financial modeled a 150-user Copilot rollout and found consumption added about 75% to the license cost at moderate use, overtaking the license entirely if usage doubled (NPI Financial).
Bill shock. Zylo reports 78% of IT leaders have been hit with unexpected AI or consumption charges, and 61% have cut projects because of unplanned SaaS costs (Zylo).
The deeper problem is structural. Procurement teams have budgeted software by headcount for twenty years, and moving budget lines from labor to consumption is hard (UC Today). A seat-plus-credit contract can also mean your bill does not fall when headcount does, because the meter replaces the revenue the seats used to carry.
How CFOs and CIOs should negotiate AI-era SaaS renewals
The winning posture is to treat every AI meter as a variable cost that needs a defined unit, a cap, and an owner before the contract is signed. These eight moves cover both the negotiation and the governance that follows.
- Get the unit in writing. Ask exactly what one credit, action, or outcome buys. If the vendor cannot define it, you cannot budget it.
- Know what happens at the cap. Clarify whether usage throttles, stops, or bills overage, and whether unused credits expire or roll over.
- Model the bill at twice expected usage. Consumption often overtakes the license once adoption grows, so stress-test the contract before you commit.
- Ask the headcount question. If you reduce staff by a third next year, what happens to the bill? A seat-plus-credit contract that grows while your team shrinks is an upsell, not a saving.
- Negotiate price protection and caps. Seek spending ceilings, rate locks on credits, and the right to move between seat and consumption models at renewal.
- Favor shorter terms while pricing is unstable. With pricing units changing within a single year, a 12-month term can be worth more than a multi-year discount.
- Monitor consumption monthly, not at renewal. Assign every metered contract an owner, set budget alerts, and bring credit burn into the same FinOps review as cloud and AI API spend.
- Govern the portfolio, not just the contract. Consolidate overlapping AI tools, route shadow AI purchases through procurement, and allocate consumption costs to the business units that drive them.
The division of work is clear. Procurement and the CFO own the commercial terms and budget thresholds. The CIO and FinOps own usage data, license and credit optimization, and the early warning when a meter starts running fast. As Flexera puts it, the industry is moving from reporting consumption to governing it (Flexera).
Frequently asked questions about AI pricing models
Is per-seat SaaS pricing dead? Not yet, but it is declining. Most vendors now add an AI usage meter on top of seats rather than replacing them, and Gartner expects at least 40% of enterprise SaaS spend to move to usage, agent, or outcome-based models by 2030.
What are AI credits? AI credits are prepaid or metered units that vendors use to charge for AI features such as agent actions, generated content, or automated tasks. What one credit buys varies by vendor, so the definition should be confirmed in the contract.
What is consumption-based pricing? Consumption-based pricing charges for actual usage, such as API calls, tokens, actions, or compute time, instead of a fixed fee per user.
What is outcome-based pricing? Outcome-based pricing charges per result, such as a resolved support ticket or a qualified lead. It aligns cost with value but lets the vendor define what counts as a result.
How can companies control AI credit costs? Define the unit in writing, set spending caps and alerts, monitor usage monthly, allocate costs to business units, and model renewals at double the expected usage.
Why is my SaaS bill rising even with fewer users? Many vendors now charge for AI usage alongside seats and raise prices on tiers that include AI. As agents take on more work, metered charges can grow even when headcount falls.
The bottom line for CFOs and CIOs
The seat charged you for headcount. The meter charges you for work. Many vendors now do both, which means your software bill can grow whether your team grows or shrinks.
That is not a reason to avoid AI features. It is a reason to buy them the way you buy cloud, with defined units, caps, owners, and monthly visibility. The companies that adapt their procurement and FinOps practices now will negotiate from strength. Those that treat AI renewals as routine will find out what a credit costs only after the invoice arrives.