--- title: Agentic AI Is Moving Enterprise Software to Pay-Per-Action Pricing, and Buyers Lack Cost Controls description: Agentic AI is shifting enterprise software to pay-per-action pricing that vendors meter. Why buyers lose control of the bill, and how to keep it. author: Darie Nani (Editor-in-Chief) updated: 2026-08-27T07:15:22.921Z canonical: https://www.sovereignmagazine.com/article/agentic-ai-pay-per-action-pricing-bill-shock image: https://cdn.nanimediahouse.com/agentic-ai-pricing-bill-shock-211985.webp categories: Artificial Intelligence content_type: Analysis region: Global publication: Sovereign Magazine schema_type: Article --- Enterprise software is changing how it charges, and the meter is moving out of the buyer’s hands. IT research and advisory firm Info-Tech Research Group warns that as companies deploy agentic AI, the predictable per-seat license is giving way to consumption-based AI pricing that most organizations are not equipped to control. In a new blueprint aimed at CIOs and their procurement, legal and finance counterparts, "Negotiate Safe AI Contracts to Prevent Bill Shock," the firm argues that the financial controls, contractual safeguards and forecasting mechanisms needed to manage variable pricing are missing from the agreements companies are already signing. ## A Single Prompt Can Trigger Dozens of Billable Events The reason agentic AI resists old budgeting habits is structural. Under per-seat licensing, a company knew its bill because it knew its headcount. Under usage-based AI contracts, one user prompt can set off dozens of separately billable events: tool calls, recursive reasoning loops, background API calls, model-tier escalation, retries and multi-agent orchestration. The work happens out of sight, and it accrues charges the whole time. What makes that hard to govern is where the rules live. Info-Tech notes that billing definitions often sit in vendor-controlled documentation rather than in the signed contract, which means their meaning can shift after deployment. Pricing logic can change with notice, and audit rights and hard caps are frequently absent altogether. The firm groups the exposure into unclear billing definitions, hidden cost drivers such as retries and recursion, weak or advisory-only financial caps, vendor-controlled pricing changes, and lock-in. "Organizations are entering agentic AI agreements without the financial controls, contractual safeguards, and forecasting mechanisms required to manage variable, intelligence-driven pricing models," said John Donovan, principal research director at Info-Tech Research Group. "Once AI workflows are embedded, the switching cost is high, and the leverage to renegotiate is gone." ## Salesforce Has Shipped Three Agentforce Pricing Models in 18 Months The market is not settling on one meter, and that itself is the problem for anyone trying to forecast. Salesforce has now run three different pricing models for its Agentforce agents in roughly 18 months: $2 per conversation at launch, then Flex Credits at $0.10 per action in May 2025, and per-user licenses starting at $125 per user per month. All three operate at the same time, on the same product, [as SaaStr has documented](https://www.saastr.com/salesforce-now-has-3-pricing-models-for-agentforce-and-maybe-right-now-thats-the-way-to-do-it/). The churn is industry-wide. The PricingSaaS 500 Index tracked more than 1,800 pricing changes across the top 500 B2B and AI companies in 2025, about 3.6 changes per company in a single year. Credit-based models grew 126 percent year over year, from 35 companies to 79, with HubSpot, Figma, Adobe, Salesforce and Cursor all adding credits. The shift is not simply predatory. Salesforce’s per-conversation and per-action options are forms of outcome-based and hybrid pricing that tie the charge to work actually done, which can limit a buyer’s exposure rather than widen it. Some vendors now build spend caps directly into their products. For a buyer, the pricing model is a variable to negotiate, not a fixed fact to accept. ## Some Buyers Are Capping the Spend Themselves Where contracts do not cap the risk, operators are learning to cap it internally. Uber imposed a monthly limit of $1,500 per employee for each agentic coding tool, including Anthropic’s Claude Code and Cursor, tracked on an internal dashboard. The cap followed a disclosure by Uber’s CTO in April 2026 that the company had burned through its entire annual AI budget in four months, after urging staff to use AI "as much as possible" and ranking usage on internal leaderboards, [TechCrunch reported](https://techcrunch.com/2026/06/02/uber-caps-employee-ai-spending-after-blowing-through-budget-in-four-months/). For CIOs, the discipline is closer to cloud cost management than to software procurement. Michael Corrigan, CIO of World Insurance Associates, a top-25 US insurance broker with about 3,000 employees across roughly 300 locations, describes AI as "a fundamentally different cost model, one that’s usage driven, non-linear, and tightly coupled to business activity," with spend shifting "from a license seat cost to a token consumption or token burn cost, or even a hybrid." Success, he [told CIO.com](https://www.cio.com/article/4190605/5-ways-for-cios-to-avoid-ai-bill-shock.html), "requires shifting from traditional IT budgeting to FinOps-style discipline where consumption, value, and governance are actively managed in real time." ## Info-Tech Tells Buyers to Negotiate Caps and Audit Rights Before Signing Info-Tech frames AI cost governance as work to finish before signature, not after the first surprising invoice. Its blueprint sets out four phases: decode how the vendor actually bills before negotiating; assess contractual and architectural exposure, from billing definitions and recursion treatment to audit rights, dispute paths and lock-in; design financial guardrails such as forecast simulations, thresholds, throttles, kill switches and clear ownership; and keep tracking vendor pricing changes after the deal closes. "Clear definitions, consumption caps, audit rights, pricing-change protections, and dispute mechanisms must be negotiated before signing," Donovan said. "When billing rules are vague, the vendor decides what they mean, and the invoice reflects that." Info-Tech’s wider research on agentic AI contracts is at its [Agentic IT Research Center](https://www.infotech.com/research-centers/agentic-it-research-center). ## FAQ **Q: What is consumption-based (usage-based) AI pricing?** It is a model where a customer pays for what an AI system does rather than for a fixed number of user seats. Charges accrue against measured activity such as actions, API calls, conversations or tokens processed, so the bill rises and falls with how heavily the software is used. **Q: What is model-tier escalation?** Many AI products route requests to different models depending on difficulty, and more capable models cost more per use. Model-tier escalation is when a task is automatically pushed to a higher, pricier tier, which raises the cost of an individual action without any change on the customer’s side. **Q: What is a token in AI billing?** A token is a small unit of text, roughly a word fragment, that language models use to read a prompt and generate a response. Usage-based AI contracts often meter both the tokens sent in and the tokens produced, so longer prompts, larger context and more verbose outputs all increase the charge. **Q: What is FinOps and how does it apply to AI spend?** FinOps is a practice, originally developed for cloud computing, that brings finance, engineering and operations together to manage variable, consumption-driven costs in close to real time. Applied to AI, it means monitoring token and action consumption continuously, tying that spend to business value, and enforcing budgets and controls rather than reconciling costs only after the invoice arrives.