Consulting giant Accenture has introduced a new funding policy restricting employee access to Anthropic’s Claude AI tools, capping central funding at $29 per worker per month to curb soaring enterprise artificial intelligence (AI) costs.
The policy shift, effective Oct. 1, highlights a growing industry-wide challenge of managing operational expenses of consumption-based AI services, commonly known as tokenomics.
Some Accenture employees, particularly developers using coding assistants like Claude Code, were burning through $500 to $1,000 worth of AI usage monthly, according to a report from 404 Media.
The new framework limits central IT funding to $29 per user per month, requiring business units and specific project budgets to cover any additional use through internal project billing codes.
Accenture disputed characterizations of the policy as a hard usage ban. It emphasized the $29 figure reflects an internal funding structure rather than a firm ceiling on AI access.
“Reporting that suggests that we have put a firm limit on our people’s access to AI tools each month is incorrect. The figure was used to demonstrate one part of the funding approach,” an Accenture spokesperson said in a statement. “Claude is a specialized AI tool and critical to how we deliver work for our clients and within our organization. We invest in Claude and deploy to our teams where it can have the greatest impact.”
Timing of the policy change is notable given Accenture’s high-profile strategic relationship with Anthropic. In December 2025, the companies announced a major partnership establishing the Accenture Anthropic Business Group, aimed at training 30,000 Accenture professionals on Claude and driving enterprise adoption globally. Anthropic CEO Dario Amodei previously characterized the deployment as the company’s largest to date.
Because Accenture actively advises corporate clients on adopting AI at scale, its internal effort to restrain token spending underscores the friction between aggressive technology deployment and financial discipline.
Under traditional enterprise software contracts, companies typically pay fixed per-seat subscription fees regardless of usage frequency.
Modern AI tools, however, frequently rely on consumption-based pricing models tied to API token usage. Advanced agentic workflows, where an AI assistant autonomously executes multi-step tasks, tests code, and fixes errors, can generate exponential token consumption from a single prompt, leading to unpredictable monthly bills.
Accenture is not alone in re-evaluating its software expenditure. Meta Platforms Inc., Microsoft Corp., and others have instituted similar usage parameters, encouraging workers to utilize cheaper internal models or reserve high-cost reasoning models for critical tasks.
The shift presents strategic questions for enterprise service providers. Under fixed-price client contracts, AI-driven efficiency gains can boost profit margins. However, under traditional hourly billing models, rapid task completion without clear outcome-based pricing structures risks eroding billable revenue while incurring higher underlying compute costs.
Industry analysts note that basic spending caps do not distinguish between high-value usage — such as a developer spending $500 in AI credits to complete $5,000 worth of billable labor — and inefficient usage. As a result, corporate finance and IT departments are increasingly tasking teams with measuring total ROI rather than simply limiting raw token consumption.
Whether Accenture’s measure signals a temporary operational adjustment or a permanent shift toward decentralized project-level AI budgeting, it provides a clear indication that major enterprise adopters are prioritizing cost management alongside AI expansion.
