Microsoft CEO Satya Nadella is urging employees to be more deliberate about how they use artificial intelligence, warning that not every task requires the company's most advanced models.
In a live taping of The New York Times' Hard Fork podcast, Nadella was asked how much "tokenmaxxing" is happening inside Microsoft, a term used to describe maximizing AI usage and token consumption.
"A lot," he replied. "I'm a tokenmaxxer too, it's addictive."
Still, he suggested that the excitement around AI should not overshadow practical considerations.
"But you have to step back when the novelty wears off to say, 'What is it that I'm trying to create?'" he said.
Nadella argued that employees should focus on using the right AI tool for the right task rather than defaulting to the most powerful models available.
"Don't use frontier models for non-frontier problems," he said, pointing to Microsoft Copilot's Auto mode, which automatically selects the most appropriate model for a given task.
David Adubiina
According to Nadella, matching workloads with the right model helps organizations achieve better outcomes while keeping costs under control.
"Let's kind of match these things such that you get the outputs, you get the economics. It can't be a race to doing things that just don't add value," he said.
His comments come as technology companies begin paying closer attention to the cost of AI adoption. Over the past year, many Silicon Valley firms like Meta and Amazon have encouraged employees to use AI tools extensively, sometimes tracking usage through internal metrics and leaderboards, but as spending on AI infrastructure, computing power, and model training continues to climb, companies are increasingly looking for ways to improve efficiency rather than simply maximize usage.
Nadella stopped short of suggesting Microsoft is restricting employee access to AI tools. Instead, his message was that workers should think carefully about which model best fits the problem they are trying to solve.
According to Business Insider, Nadella has been working to reshape Microsoft for the AI era, seeking to turn the roughly 220,000-person company into one that can operate more like smaller, faster rivals.
This also comes as concerns surface around how newer AI systems handle data. According to a report by The Verge, the software company is limiting internal employee access to Anthropic's newly launched Claude Fable 5 model while its legal teams review changes to the AI model’s data retention policies.
Unlike previous Claude models that operated under Zero Data Retention rules, Claude Fable 5 retains prompts and outputs for up to 30 days to support Anthropic's safety systems, with some flagged data potentially stored for longer periods.
The reported review highlights Nadella's broader point that not every model is appropriate for every use case. As organizations deploy multiple AI systems across their workflows, choosing the right model is becoming as much a business and risk-management decision as a technical one.
Damilare Odedina
