Some Accenture employees spent $1,000 a month on Anthropic's Claude AI, the company will now pay only $29; memo warns: All costs are...
Accenture , one of the world's largest consulting firms and a vocal champion of AI adoption among its clients, has put a tight ceiling on how much it will pay for its own employees' use of Anthropic's Claude . According to an internal announcement seen by 404 Media, the company will now centrally fund just $29 a month per person towards Claude Enterprise usage, effective October 1.
That figure is small next to what some staff have been spending. Several employees have told colleagues internally that their monthly bills run between $500 and $1,000, largely from heavy use of Claude Code. Under the new rules, anyone on the standard plan who goes past the $29 allowance needs a project or team to cover the rest.
Anything above $29 now gets charged to a project budgetThe memo states that all Claude costs are now consumption-based. Any usage beyond the central allowance has to be funded through a Work Breakdown Structure, or WBS, which is Accenture's internal billing code for a specific project or team.
Employees on the Premium plan are better placed, since it includes $250 worth of token usage every week. Most of the workforce, however, sits on the General plan.
The memo also makes clear that a licence is not permanent. Access is tied to role and business need, allocations are reviewed periodically, and licences can be moved to wherever the company feels they deliver the most value. Even supplying a WBS code does not guarantee continued access.
Heavy Claude Code users are scrambling for Premium upgradesInternal messages viewed by 404 Media show how the change has landed. One employee said their usage had crossed $500 for several months in a row and asked how to fast-track a move to Premium, adding that the upgrade option appeared blocked on the support page. Their concern was simple: the extra spend would now hit project costs.
Another employee uses Claude Code extensively to build and test an AI accelerator meant for several Accenture clients. Their monthly usage typically sits between $500 and $1,000. "Honestly, this is disappointing," they wrote. A third employee said their access to the tool had been revoked altogether.
An Accenture spokesperson told 404 Media that the company makes AI available to all its people and aims to match the right tool to the right work, while continuing to invest in AI capabilities and skills.
Accenture joins Microsoft and Meta in pushing back on tokenmaxxingAccenture is not the first big company to rethink open-ended AI budgets. Microsoft introduced division-level token budgets in July and switched its internal default to a cheaper model, with CoreAI EVP Jay Parikh telling staff that "tokenmaxxing is not what we are optimizing for." Months earlier, its Experiences and Devices group cancelled most Claude Code licences and moved engineers to GitHub Copilot CLI.
Tokenmaxxing describes employees burning through as many AI tokens as possible to look productive on usage dashboards. At Meta, an internal leaderboard called "Claudeonomics" ranked staff by token consumption before being pulled within days. Amazon, Uber, Walmart and AT&T have also capped or throttled employee AI spending this year.
For a firm that has spent years urging clients to pour money into AI, Accenture's $29 cap shows where the conversation inside companies is heading. Heavy usage alone no longer justifies the bill, and teams are now expected to show what every dollar of tokens actually delivers.
That figure is small next to what some staff have been spending. Several employees have told colleagues internally that their monthly bills run between $500 and $1,000, largely from heavy use of Claude Code. Under the new rules, anyone on the standard plan who goes past the $29 allowance needs a project or team to cover the rest.
Anything above $29 now gets charged to a project budgetThe memo states that all Claude costs are now consumption-based. Any usage beyond the central allowance has to be funded through a Work Breakdown Structure, or WBS, which is Accenture's internal billing code for a specific project or team.
Employees on the Premium plan are better placed, since it includes $250 worth of token usage every week. Most of the workforce, however, sits on the General plan.
The memo also makes clear that a licence is not permanent. Access is tied to role and business need, allocations are reviewed periodically, and licences can be moved to wherever the company feels they deliver the most value. Even supplying a WBS code does not guarantee continued access.
Heavy Claude Code users are scrambling for Premium upgradesInternal messages viewed by 404 Media show how the change has landed. One employee said their usage had crossed $500 for several months in a row and asked how to fast-track a move to Premium, adding that the upgrade option appeared blocked on the support page. Their concern was simple: the extra spend would now hit project costs.
Another employee uses Claude Code extensively to build and test an AI accelerator meant for several Accenture clients. Their monthly usage typically sits between $500 and $1,000. "Honestly, this is disappointing," they wrote. A third employee said their access to the tool had been revoked altogether.
An Accenture spokesperson told 404 Media that the company makes AI available to all its people and aims to match the right tool to the right work, while continuing to invest in AI capabilities and skills.
Accenture joins Microsoft and Meta in pushing back on tokenmaxxingAccenture is not the first big company to rethink open-ended AI budgets. Microsoft introduced division-level token budgets in July and switched its internal default to a cheaper model, with CoreAI EVP Jay Parikh telling staff that "tokenmaxxing is not what we are optimizing for." Months earlier, its Experiences and Devices group cancelled most Claude Code licences and moved engineers to GitHub Copilot CLI.
Tokenmaxxing describes employees burning through as many AI tokens as possible to look productive on usage dashboards. At Meta, an internal leaderboard called "Claudeonomics" ranked staff by token consumption before being pulled within days. Amazon, Uber, Walmart and AT&T have also capped or throttled employee AI spending this year.
For a firm that has spent years urging clients to pour money into AI, Accenture's $29 cap shows where the conversation inside companies is heading. Heavy usage alone no longer justifies the bill, and teams are now expected to show what every dollar of tokens actually delivers.
Next Story