Accenture caps staff Claude spending at $29 a month. Some consultants were burning $1,000

Accenture sells AI transformation work to roughly 9,000 clients. Inside its own walls, it just put a ceiling on how much of that AI it will pay for.

Starting October 1, the firm centrally funds $29 per employee per month for Claude, the model built by Anthropic, according to an internal memo reported on October 15. Any usage above that has to be billed to a project.

The gap between the allowance and real usage is wide. One Accenture consultant building an AI solution for several clients said their Claude Code usage ran between $500 and $1,000 a month. Another employee said spending had topped $500 a month for months.

What the memo actually says

The memo, titled "Claude Enterprise Users: Important Updates to Licensing and Usage Costs," states plainly that "all costs are consumption-based." Anything over the allowance goes to a Work Breakdown Structure, or WBS, the billing code Accenture uses for a project or team.

Heavy use is still allowed. It just needs someone willing to own the budget. Premium plans include $250 of token usage per week, but many employees aren't on Premium.

Access isn't guaranteed either. "Providing a WBS and obtaining a license does not guarantee indefinite or ongoing access to Claude," the memo says. At least one employee lost access entirely.

Accenture hasn't published the memo or said how many people it affects. A spokesperson said only: "Our approach today is to provide the right tool for the right work."

Why $29 doesn't cover agents

Twenty-nine dollars is a consumer chat subscription price. Agentic coding tools, which carry out multi-step tasks on their own, chew through far more tokens, the units AI models bill by. A flat per-seat budget works for someone asking a chatbot questions. It breaks the moment that person starts running agents.

CEO Julie Sweet said on the October 1 earnings call that falling token costs will "enable companies to use AI in more places and at scale." The memo suggests prices have further to fall before agentic work can be funded seat by seat.

The question is no longer whether employees use AI. It's which work is worth paying tokens for.

The margin math behind the cap

Accenture earns its margin on labor across roughly 814,000 employees. Unlimited token spend on that base is a cost line that grows without a matching revenue line.

Routing overages to a project code ties AI spending to client work. That matters because fixed-price contracts make up more than 65% of Accenture's bookings. On those deals, AI costs that weren't priced in come straight out of Accenture's margin, not the client's budget. Management has guided to a 15.9% to 16.1% operating margin for fiscal 2027, which leaves little room for surprise expenses.

The lesson for everyone else

A flat allowance is a blunt instrument. It treats every employee's work as equally worth automating, which is rarely true. Some workflows justify $1,000 a month in agent time several times over. Others don't justify $29.

Accenture's answer is to make each overage find a project that will pay for it. Companies outside consulting face the same question in a different form: which processes deserve an AI budget at all? Answering it means knowing where time really goes, including the parts of the job that never show up in a system.

What we still don't know

This is one policy, reported secondhand. Total spend, the number of employees affected, and whether other AI tools face similar limits are all unknown.

Accenture is still expanding in AI, with nearly 110,000 AI and data professionals, and Sweet has described the Anthropic relationship as non-exclusive. For AI vendors, tighter limits at the largest IT services firms could slow enterprise spending, though targeted use on proven work may prove more durable.

Investors didn't react to the memo itself. Accenture shares rose 4.63% to $224.92 on October 15 during a broad software rally, but remain well below their 52-week high of $291.09.

This story draws on reporting by 24/7 Wall St., syndicated on Yahoo Finance.