In the tech world, we’re used to buzzwords. Each new technology comes with a buzzword attached. The crypto world had tokenomics, the cybersecurity world is rapidly shifting left, and in the AI world, it’s all about tokenmaxxing.
If you’re new to the term, it means using as much token capacity as possible to get the most information, reasoning, or performance out of an AI model. But, like many trends, tokenmaxxing has had major economic consequences, and companies are beginning to realize it. One of those companies is Microsoft.
What’s the News?
According to an internal email seen by 404 Media, Microsoft is telling its employees to slow down on token consumption and introducing tighter controls around AI usage within the company. The company is asking employees to use the right model for the right task instead of automatically opting for the most powerful option.
Here’s an excerpt from the reported email sent by Jay Parikh, Executive Vice President of CoreAI at Microsoft:
“As we accelerate our use of GitHub Copilot to deliver on our goals, we all need to be aware of how we consume tokens.”
“Tokenmaxxing is not what we are optimizing for. I want all of us focused on maximizing outcomes that move the needle for our customers and our business.”
“As such, we are updating our internal guidance and managing token spend with the same discipline we apply to every other critical resource.”
“We are not optimizing for fewer tokens. We are optimizing for more impact per token.”
To help employees achieve these goals, Microsoft is reportedly making the budget-friendly OpenAI GPT-5.6 model the default option available for use.
How Should You Regulate Token Consumption?
The news from Microsoft around tokenmaxxing isn’t an isolated story. Companies globally are struggling to balance AI usage with AI output. Here are some of the ways you can do that:
Use the right model for the task: There are so many models out there that have been designed for particular use cases, so there’s no need to opt for a more powerful alternative by default.
Set token budgets: Setting budgets may sound simple, but many organizations haven’t yet established usage limits. Within this, you can organize a structure for raising limits in special circumstances.
Keep context tight: Giving models irrelevant information increases the number of tokens you use. Make sure the information is direct and to the point.
Use caching: Reuse repeated prompts and context instead of processing them from scratch each time.
Measure outcomes, not tokens: Remember that the ultimate aim isn’t to use fewer tokens; it’s about getting the most out of the tokens you use. That’s how you should measure success.
Final Thought
The AI industry is quickly moving away from a “more is better” to a “quality over quantity” approach, and that’s really been highlighted by this glimpse into how Microsoft is addressing the issue of tokenmaxxing.
While the first phase of AI was about proving what it could do, the next phase will be about proving it’s worth the cost.
Community Summit North America is the largest independent innovation, education, and training event for Microsoft business applications delivered by Expert Users, Microsoft Leaders, MVPs, and Partners. Register now to attend Community Summit in Nashville, TN from October 11-15.
