Enterprise software companies like Salesforce, SAP, and Pegasystems are actively experimenting with AI tokenomics pricing structures to balance customer cost concerns and measurable value from AI usage. Meanwhile, escalating legal tensions between the US government and AI startup Anthropic spotlight critical discussions over the future framework governing artificial intelligence.
- Vendors seek balanced AI token pricing that reflects customer value and usage.
- Tokenomics models vary: fixed licenses, consumption tokens, outcome metrics.
- US legal case against Anthropic escalates AI regulatory scrutiny.
What happened
Several leading enterprise software vendors have publicly detailed their evolving approaches to AI tokenomics, as customers grow concerned about unpredictable costs tied to token consumption. Salesforce’s Chief Marketing Officer Patrick Stokes highlighted their 'Agentic Enterprise Licensing Agreement' which offers an all-you-can-eat model but acknowledged its expense. His team also introduced Agentic Work Units (AWUs) designed to measure actual work performed by AI, rather than mere token usage.
SAP’s Chief AI Officer Philipp Herzig expressed customer resistance to token models, emphasizing the need to differentiate between base AI functionalities included at no extra cost and premium AI capabilities charged via AI unit SKUs. Pegasystems CEO Alan Trefler criticized the high token prices set by large language model providers, advocating for strategic AI runtime token use focused on task necessity to control costs. These vendor perspectives reflect broader industry challenges in delivering transparent, value-driven AI billing models.
Why it matters
The complexity of AI pricing and tokenomics models directly impacts enterprise IT budgets and AI adoption decisions. Customers want clarity on the relationship between AI consumption and tangible business outcomes, pushing vendors to innovate pricing that aligns with real value rather than raw token metrics. Without effective models, enterprises risk sticker shock and underutilization of AI tools.
Simultaneously, the ongoing US government versus Anthropic legal dispute brings regulatory considerations to the forefront, potentially shaping rules around AI safety, transparency, and operational governance. This conflict illustrates growing governmental focus on AI ethics and control, signaling that vendors and buyers alike must anticipate tighter regulatory frameworks influencing product development and deployment.
What to watch next
Enterprises and vendors will closely monitor how tokenomics models evolve, especially the adoption and impact of combined consumption and outcome-based metrics like Salesforce’s AWU and SAP’s AI unit approach. These could become best practices if they deliver clearer ROI while simplifying billing. Vendors may also experiment further with runtime controls and hybrid LLM usage, as advocated by Pegasystems, to optimize cost-effectiveness.
On the regulatory front, the US-Anthropic case will be pivotal in defining the scope of AI oversight in the United States. Its outcomes could influence international debates and create new compliance requirements for enterprise AI software providers. Stakeholders should watch for emerging regulation signals that might affect AI business models, token pricing transparency, and vendor liability.