It's kind of humorous now that AI is everywhere and on everything. The toothbrush I used this morning has AI. The toaster has AI. The dashboard has AI. There's an AI button slapped onto the same old software we've used for years. And here's the thing nobody wants to say out loud. Calling something AI does not make it useful. The real test is simple in my view. Does it save time? Does it reduce friction? Does it make the work cleaner? If not, it's not transformation. It's a label.
Let me sit on this, because the labeling has gotten out of hand and it's worth understanding why.
Does calling a product "AI" make it useful?
We've hit the stage of a technology cycle where the word becomes a marketing sticker. It happened with "smart." Remember when everything got the word smart in front of it? Smart fridge. Smart toothbrush. Smart toaster. Most of it was a normal product with a chip and an app that nobody opened twice. AI is at that exact stage now. The word has detached from the substance. It's become a thing you put on the box to move units, not a description of what the product actually does.
What is "AI washing"?
And I get why companies do it. AI sells. Put it in the headline, put it on the feature list, watch the attention come. There's enormous pressure to have an AI story, so everyone manufactures one. The fastest way to manufacture one is to take what you already had, add a chatbot or a button, and call it AI-powered. Nothing underneath changed. The label changed.
Why do so many AI products fail to deliver value?
Here's the problem with the label. It doesn't survive contact with reality. You can call a toaster AI, but the bread comes out the same. You can put an AI button on twenty-year-old software, but if the workflow underneath is the same broken workflow it always was, the button is decoration. The label promises transformation and delivers a sticker, and the gap between the two is where all the disappointment lives.
What did the MIT GenAI Divide study find?
There's hard data on this gap, and it's striking. MIT ran a study this year called The GenAI Divide, looking at the state of AI in business. They examined hundreds of deployments. What they found is that 95 percent of generative AI pilots delivered no measurable impact on the business. None. Only 5 percent produced real value. And the thing that separated the 5 percent from the 95 was not the model or the technology. It was whether the company actually rebuilt the work around the AI, or whether they slapped it on top of what they already had.
Read that again, because it's the whole point. 95 percent got a label. 5 percent got a transformation. And the difference wasn't the AI. It was what the company did with it.
How do you tell real AI from an AI label?
So this is the test I run on everything now, and I'd encourage anyone reading to run it too. Three questions. They're simple, and they're brutal, and almost nothing passes all three.
Does it save time? Not in theory. In practice. When you actually use the thing, do you get time back, or do you spend the same amount of time in a slightly different way? Most AI features fail this immediately, because they add a step to look impressive instead of removing a step to be useful.
Does it reduce friction? Does the work get easier, smoother, less annoying? Or did the AI add a new thing to manage, a new output to check, a new place where something can go wrong? Real AI takes friction out. A label adds it and calls the addition innovation.
Does it make the work cleaner? When you're done, is the process in better shape than before, or did the AI leave a mess behind that someone has to clean up? This is the one almost nobody thinks about, and it's the one that matters most over time.
If a tool passes those three, it's real. It's transformation. It earned the name. If it fails them, it doesn't matter what's on the box. It's a label, and labels don't change your life, your work, or your business. They just change the marketing.
I think about this constantly from where I sit, because it would be easy to be one of the labelers. It's the path of least resistance. Take the thing, add the word, ride the wave. But a label gets exposed the moment someone actually uses the product, and in a category with real consequences, getting exposed is fatal. So the only version worth building is the one that passes the three questions. Does it save the time. Does it reduce the friction. Does it make the work cleaner. If the answer isn't yes to all three, it isn't ready, and putting AI on the box doesn't make it ready.
The labeling phase will pass, the same way the smart phase passed. The toothbrushes and toasters will quietly drop the word when it stops selling. And what'll be left standing is the small percentage of AI that actually did the work, that rebuilt the process instead of decorating it, that passed the test when the marketing wore off. That's the AI worth building and the AI worth paying for. Everything else is a sticker.
It's not transformation if it doesn't save time, reduce friction, and make the work cleaner. If it doesn't do those things, it's a label. And the market is about to get very good at telling the difference.
*Judd Hoffman is CEO and Co-Founder of Ethica AI, building AI-powered tools for real estate transaction workflows.*