Microsoft didn't tell its engineers to stop using AI — the real story is more important

You may have seen a surprising headline recently: Microsoft told its engineers to stop using AI. Considering Microsoft has invested billions of dollars in artificial intelligence and has built AI into Windows, Microsoft 365, Copilot and many of its other products, that sounds extraordinary. There is just one problem: that's not really what happened. And I think the real story teaches us something much more important about artificial intelligence.

Microsoft Didn't Stop Using AI

Earlier this year, Microsoft began moving some of its engineers away from one particular AI programming tool called Claude Code and toward Microsoft's own GitHub Copilot tools. Then, in August, Microsoft executives began putting more controls around how much money employees were spending on AI. One executive used an interesting term for the pattern they were seeing: "tokenmaxxing."

A token is essentially a small piece of information processed by an AI system. The more complicated the work you ask AI to perform, and the more AI agents work independently, the more tokens get consumed. At a large company, those costs add up quickly.

Microsoft's message wasn't "stop using AI." It was "use AI where it produces a worthwhile result." That difference matters.

More AI Does Not Automatically Mean Better

For the last few years, much of the conversation surrounding artificial intelligence has been about using more of it — more AI in our computers, our phones, our businesses, our automation. But Microsoft discovered something interesting inside its own company: its engineers could accomplish individual tasks faster using AI, yet the company reported in September that making each employee faster did not automatically make the whole team more productive.

That is an important lesson, because technology has never been valuable simply because we use more of it. It is valuable when it improves our lives.

Think About the Calculator

A calculator can perform arithmetic considerably faster than most of us. But owning a calculator doesn't automatically make someone good with money. You still need to understand what numbers to enter. You need to recognize when an answer doesn't make sense. And you need to decide what to do with the answer.

Artificial intelligence works much the same way. AI can help you write a letter, explain a complicated subject, organize information, brainstorm ideas, or compare options. It can help you prepare questions for your doctor, attorney, accountant or financial advisor. It can even teach you how to use something on your computer. But someone still has to ask the right question. Someone still has to think about the answer. And someone still has to make the decision. That someone is you.

This Is Why Learning About AI Matters

I believe one of the mistakes we could make right now is dividing people into two groups: people who use AI and people who don't. I think the better distinction will eventually be between people who understand how to work with AI and people who simply accept whatever AI gives them. Those are very different things.

You do not need to become a computer programmer. You don't need to understand how artificial intelligence is built. And you certainly don't need to use AI for everything. What is becoming increasingly valuable is knowing when AI can help, how to ask it for help, and when not to trust it without checking.

Microsoft itself talks openly about this problem: AI can make mistakes, misunderstand instructions, and occasionally state something confidently that isn't true. That means human judgment isn't becoming less important. It's becoming more important.

Especially for Those of Us Who Didn't Grow Up With It

Many of our clients at Megabite are retired or approaching retirement. Some have told me that artificial intelligence feels like another enormous technological change they are expected to somehow understand overnight. It doesn't have to be that way.

You didn't need to understand telephone switching equipment to use a telephone. You didn't need to understand how the internet worked before sending your first email. And you don't need to understand a large language model before asking AI your first useful question.

Start small. Ask it to explain something you don't understand. Ask it to help you organize an idea. Ask it to simplify instructions. Ask it to help draft a letter. Then read the answer, question it, correct it, and ask another question. That is where the real value begins.

The Most Important Skill May Be Judgment

Microsoft's experience is a reminder that the future isn't simply about getting humans out of the way and letting artificial intelligence take over. Even one of the largest technology companies in the world is learning that AI works best when people provide the direction, judgment and accountability. I think that is encouraging, because those are human skills. Experience matters. Knowing what you are trying to accomplish matters. And asking good questions matters.

Artificial intelligence may become one of the most powerful tools we have ever placed on an ordinary person's desk. But the goal shouldn't be to use as much AI as possible. The goal should be much simpler: use it when it makes your life better.

That is the lesson Microsoft appears to be learning inside its own company. And it's a lesson the rest of us would be wise to learn as well.

Deryck Burnett is the founder of Megabite, a technology, creative and multimedia company on Amelia Island. Reach the team at 904-430-0350 or Megabite.