Enterprise artificial intelligence, by pricing AI in granular units such as tokens, compute cycles and API calls, is breaking that model open. Adobe last week (April 21) announced plans to use  outcome-based pricing for its new suite of AI products called Adobe CX Enterprise, OpenAI has reportedly begun offering ChatGPT advertising campaigns based on how many users click ads, Anthropic has also begun charging enterprise customers based on their levels of AI use and a Monday (April 27) report from The Information claims that SaaS firms like Salesforce and HubSpot are getting ready to join the outcome-based AI pricing movement, too.

Meanwhile, AI providers are increasingly doubling down on the enterprise market. Google has touted its Model Context Protocol servers as a way to standardize the way AI systems retrieve verified data across environments, while Anthropic launched its own platform for agentic AI enterprise applications. At the same time, OpenAI is reportedly working with consulting firms to integrate its enterprise solutions into business workflows.

And from outcome-based models to engineering teams “token-maxxing,” the disconnect between engineering velocity and financial visibility is becoming harder for CFOs to ignore. A surge in internal experimentation, a new product feature or even a poorly optimized prompt can cause costs to spike in ways that are difficult to anticipate. The unit economics are precise, but the aggregate behavior is not.

Decisions about which model to use, then are no longer purely technical. They carry financial implications that must be understood and managed.

See also: B2B’s New Battlefield Is Everything Before the Button

Enterprise AI Needs to Conquer Invoicing Before It Can Scale

In theory, tokenized access to AI through metered compute, data rights, usage credits, etc. can help turn opaque engineering and compute spend into something auditable and tradable in real-time for finance. In practice, most of the billing models and interfaces being offered to enterprise users can look in some cases like accounting theater, with volatility where CFOs want predictability and governance models that can lag the balance-sheet implications.

As a result, scaling AI across the enterprise may depend less on technical capability than on financial infrastructure. Invoicing, billing and cost allocation, which have traditionally been viewed as back-office functions, are now becoming strategic enablers in the age of enterprise AI.

The rapid pace of AI adoption, after all, has fast outstripped the development of financial governance frameworks. Engineering teams are moving quickly, integrating models into workflows, products and internal tools. New capabilities are announced frequently, each promising greater efficiency or competitive advantage.

Finance, by contrast, operates within systems designed for a different era. Approval processes, cost allocation methods and reporting structures are not always equipped to handle the fluidity of token-based consumption. The lag is not due to inertia but to structural constraints.

Achieving alignment may demand new capabilities. Finance teams could need to move closer to the operational layer, working alongside engineering to define usage policies, optimize prompts and evaluate trade-offs. Engineering teams, in turn, may need to incorporate cost awareness into their workflows, treating tokens not just as technical inputs but as financial resources.

The April 2026 edition of the “Payments Innovation Tracker^®,” a collaboration with Paymentology, examines how agentic commerce is redefining payments infrastructure requirements and how intelligent, API-enabled platforms are enabling secure, scalable autonomous transactions.

See also: Agentic B2B Is Here. Are Your Contracts and Invoices Ready? 

CFOs Navigate Finance’s New Procurement Equilibrium

Traditional software investments could be capitalized or at least forecasted with a high degree of certainty. Token-based AI spend, however, is typically expensed as incurred. As usage grows, it can materially impact operating margins in ways that are difficult to smooth over time, at least with existing financial infrastructure and organizational know-how.

Finance teams can see what is happening, but they often lack the mechanisms to shape it in real time. The result is a reactive posture, where spend is analyzed after the fact rather than governed proactively.

Research from PYMNTS Intelligence’s “The Enterprise AI Benchmark Report,” which shows that 71% of executives at companies with at least $1 billion in annual revenue believe that organizational readiness is the chief limitation on AI performance. Only 11% said they think AI technology itself is the primary barrier.

Writing about AI adoption last week, PYMNTS CEO Karen Webster argued that tools like Claude are able to gain ground as consumers encounter new AI models on the job.

“ChatGPT expands outward from the consumer, earning trust in low-stakes, high-frequency tasks and carrying that trust into the workplace. The habit comes first; the enterprise follows,” Webster wrote.