Inc. Magazine
Rushing into AI can backfire. Microsoft urges businesses to slow down and plan carefully, warning that AI tools often behave unpredictably—like smart but untrained new hires. Errors like hallucinations, data omissions, and prompt manipulation are common. Safe deployment means knowing what might go wrong, preparing for it, and always keeping humans in the loop.
Why Microsoft Is Cautioning Businesses About Launching AI Tools
Author: Kit Eaton
Deploying AI may feel like something your company has to do to avoid being left behind the competition. Businesses of every type are trying the new technology in hopes of achieving all sorts of efficiency boosts, cost cutting, or even new idea generation. But AI has proven to be a risky, complicated prospect—potentially bringing with it legal problems, insider threat risks, and even complex copyright violation issues. That’s why you should pay attention to a new blog post from Microsoft offering guidance on how to deploy AI solutions safely inside your business. Some of it is common sense, but other suggestions will make you think twice, and might mean changes to your company AI guidelines.
The blog post by Yonatan Zunger, corporate vice president and deputy chief information security officer for AI, explains that “deploying safely” doesn’t mean that “nothing can go wrong; things can always go wrong.” In a “safe deployment, you understand as many of the things that can go wrong as possible and have a plan for them that gives you confidence that a failure won’t turn into a major incident,” he notes.
Zunger’s advice starts with some planning tips, including having your tech team work to “understand the things that might go wrong in your system, and for each of those things, have a plan.” It also underlines that when you’re analyzing your business system for AI risks, it “means the entire business process that uses it, including the people, and ‘things that might go wrong’ includes anything that could end up with you having to respond to it, whether it’s a security breach or your system ending up on the front page of the paper for all the wrong reasons.”
Zunger also points out something that’s really easy to forget about how AI systems function—they’re not like typical computer software, which you can roll out and, typically, it “just works.” Instead Zunger notes that “the most important thing we’ve learned is that error is an intrinsic part of how AI works.” There are built-in issues like hallucination, in which AI just makes up information that seems real but isn’t, and prompt injection, in which AIs can be made to misbehave or give out incorrect responses by users manipulating the requests they make.
He warns that when companies use AI tools, either made by their own engineers or bought from third-party providers, they need to watch for misinterpreted data, hallucinated output, output omissions—where the AI just ignores some of the data you showed it before asking for analysis—and other typical AI errors.
Microsoft’s advice boils down to a single idea, which may be a useful thought exercise for any company planning on deploying AI: You should plan on double-checking your AI output by “imagining ‘what would happen if I replaced the AI with a room full of well-intentioned but inexperienced new hires?’ ” Zunger says.
“Don’t think of the AI like a senior person—think of it like a new hire fresh out of school, enthusiastic, intelligent, ready to help, and occasionally dumb as a rock,” he suggests. Then when it comes to writing guidelines for your staff to follow when they’re using AI, you should “build safety into your process by considering what you’d do for humans in the same place—for example, having multiple sets of (AI or human) eyes on key decisions.”
Zunger’s advice resonates with plenty of other experts’ views of best practices when deploying AI. The core principles include educating the workforce that will be using new AI tools—to both ensure you’re getting maximum return on the investment, and also to be certain that your employees understand the risks. Also, you should keep a human in the loop to check that the output of an AI tool really is what it says it is and isn’t a hallucination or a partial solution.
And if all of this gives you pause or makes you reconsider some of the overblown assertions that AI can replace some of your workers—that’s part of the process of doing AI introduction well.
A Reddit discussion among IT experts about leadership decisions to deploy AI reminds business leaders that everything about this process is new. “Had several discussions with management about use of AI and what controls may be needed moving forward. These generally end up being pushed at IT to solve when IT is the one asking all the questions of the business as to what use cases are we trying to solve,” one commenter noted. In reply, another commenter suggested IT teams “need to be upfront with the organization about needing the budget to do it, you also need to demand help where needed like legal issues.”
Lastly, one Reddit user was emphatic that AI deployment is a top-down issue: “It should be 100% up to the business to own the policy. At the very least, they need to have a FULL understanding of what they are asking AI to accomplish, and what it could potentially cover (and replace?)”
Credits: TCA, LLC.