Atlassian, a software firm, has taken steps to monitor and control its employees’ AI spending by introducing ‘wallets’ with monthly caps ranging from $500 to $2,000 per employee. This move comes as the company recently cited AI as a factor in its decision to cut 1,600 staff, and it diverges from the trend of other tech companies that have been encouraging employees to use as much AI as possible, a practice known as ‘tokenmaxxing.’
Tokenmaxxing refers to the use of AI to the fullest extent, often resulting in significant costs. Tokens are the unit of measurement for AI responses, with OpenAI’s flagship model, GPT-5.6 Sol, charging $5 for every 1 million tokens. Other models, such as Anthropic’s Claude Fable and Mythos, charge $10 for every 1 million tokens. This has led to substantial expenses for some companies, with Uber reportedly exhausting its AI budget in just four months and Amazon advising employees to use AI judiciously.
Atlassian’s AI wallet system allows employees to use their allocated funds across four AI products, including Claude Code. Employees receive notifications when they approach their limit, and their usage is paused when the funds are depleted. However, they can request additional funds, and it is understood that the company has not denied any such requests so far.
A company spokesperson stated that Atlassian is transforming into an ‘AI-first company’ by supporting employees in building and experimenting with AI technology. The spokesperson added that the AI wallet represents a significant budget increase for employees to leverage multiple AI tools.
Industry experts have praised Atlassian’s approach, with Elastic’s ANZ manager, Jeremy Pell, calling the monthly cap on AI spend ‘smart’ and advocating for more organizations to adopt similar practices. Arun Chandrasekaran, a distinguished vice-president analyst at Gartner, noted that wallets are a simple way to ‘incentivize the right behavior’ and prevent ineffective AI use.
The cost explosion in AI is becoming a pressing issue for businesses, driven by AI agents that autonomously undertake tasks on behalf of users. Companies are exploring ways to reduce costs, including using less powerful models for simpler tasks and considering open-source models or open weight, where models can be downloaded and run on in-house systems.
