Accenture Put a $29 Meter on Claude

Every company that rolled out AI coding agents this year eventually got the same surprise: the bill. Accenture, one of the largest buyers of AI seats on the planet, has now put a number on how much of that surprise it is willing to absorb centrally. It is $29 per person per month.

What the memo says

According to an internal memo reported on October 15, titled "Claude Enterprise Users: Important Updates to Licensing and Usage Costs," Accenture began centrally funding $29 per user per month toward Anthropic's Claude from October 1, 2026. Anything above that has to be charged to a WBS code, Accenture's billing code for a project or team.

The memo makes the economics explicit: "all costs are consumption-based." Employees on Premium plans get $250 of token usage a week, but many staff are not on Premium. The memo also warns that "providing a WBS and obtaining a license does not guarantee indefinite or ongoing access to Claude." At least one employee lost access entirely, and The Telegraph reported that some staff lost licences because their roles did not require one.

Accenture did not publish the memo or say how many of its roughly 814,000 employees are affected. A spokesperson said, "Our approach today is to provide the right tool for the right work," and the company has described the $29 figure as one part of its funding approach rather than a hard limit on access.

The number that explains the cap

The useful context is what people were actually spending. A consultant building an AI solution for several clients said their Claude Code usage ran $500 to $1,000 a month. Another employee said they had been above $500 a month for months.

Set those against $29 and the gap is the story. Twenty-nine dollars is roughly what a consumer pays for a chat subscription. Agentic coding tools are not chat. They read whole repositories, plan, call tools, run tests, retry and loop. Each of those steps consumes tokens, and a productive afternoon can burn through more than a casual user spends in a year.

The Telegraph also reported that Uber capped engineering spending on Claude at the same monthly amount, which suggests $29 is becoming a reference point rather than an Accenture quirk.

Why a consultancy cares more than most

Accenture's business model sharpens the problem. More than 65% of its bookings are fixed-price work, according to coverage of its latest results. On a fixed-price contract, every extra dollar of tokens a consultant burns comes straight out of margin unless it can be billed through to the client.

That is exactly what the WBS rule does. It moves the cost from a central IT line, where nobody owns it, onto the project that benefits, where someone has to justify it. Seen that way, the memo is less a cap on AI than a demand that AI usage show up in the same ledger as everything else.

Accenture's own leadership has been bullish on where this goes. On its October 1 earnings call, chief executive Julie Sweet said falling token costs will "enable companies to use AI in more places and at scale," and described the Anthropic relationship as non-exclusive. The two firms formed the Accenture Anthropic Business Group in a multi-year partnership announced in December 2025, with formal Claude training planned for about 30,000 employees.

None of that is contradicted by a funding cap. Expecting costs to fall and refusing to pay unlimited bills while they are still high are the same position.

The agent-era budget problem

Every large company is going to have this conversation, and most have not started. Per-seat licensing made AI easy to budget: count the people, multiply by the price. Consumption pricing makes it easy to use and hard to forecast. The heaviest users are often the most valuable ones, which means a blunt cap risks throttling exactly the work that justifies the spend.

The likely end state is something like cloud computing went through a decade ago: central allowances for light use, chargeback for heavy use, and a growing finance function dedicated to asking whether each project's token bill earned its keep. Accenture's memo reads like the first page of that playbook.

For Anthropic, the signal is mixed but not bad. A cap on central funding is not a cap on demand; work above the line can still be billed. But it does mean the next phase of enterprise growth depends less on getting seats deployed and more on proving, project by project, that an agent's output is worth its token count.

The takeaway

The $29 figure is not a verdict on Claude. It is a verdict on the gap between how AI is priced and how enterprises budget. Agents that do real work consume real compute, and someone has to own that cost. Accenture has decided it should be the project, not the IT department.

Expect a lot more memos like this one. The interesting ones will be the companies that can show the meter was worth running.