Most of what I hear about AI in business and education falls into two camps. One wants to go all in. Adopt everything, automate everything, let the model drive. The other wants to ban it. Block it, police it, treat it like a threat to learning, work, and trust.

I understand both instincts. But both are incomplete. The biggest risk isn’t solved by pretending AI doesn’t exist, and it isn’t solved by letting AI become the default operating system for creativity. The real danger is quieter and doesn’t require advanced AI use to feel it.

We’ve been here before. Social media promised connection and creativity. For a while, it delivered. What it quietly took in return (depth of attention, tolerance for uncertainty, ability to sit with an idea before seeking validation) didn’t disappear in a single moment.

It eroded in a million rational decisions until dependency on external validation became walls we stopped seeing.

We watched it happen to our teenagers first (likes, follower counts, compulsive checking, anxiety when it didn’t come). A generation learned to measure the worth of a thought, a face, a body, a life by how many strangers approved of it. Not just one generation. How much am I thinking if people will “like” this when I post it on LinkedIn?

Some of what was lost with social media (certain habits of mind, expectations of privacy, unmediated experience) feel like they’re not recoverable. How about confidence, patience, trust in instinct, feeling comfortable in your own skin? We didn’t notice them leaving. A pattern worth considering as AI moves into the center of how we work, create, and decide.

The real danger isn’t takeover. It’s permission.

Innovation doesn’t die because AI becomes “too smart.” It dies when AI becomes the permission structure.

When AI is the first draft and the final judge, teams begin to internalize a new rule. “If the model didn’t suggest it, it’s probably not worth trying.” That sounds dramatic, but in practice it shows up as premature convergence. Ideas get smoothed down into something plausible, defensible, and safe. AI makes the walls feel like the edge of possible.

Creative leaps disappear, not because people lack imagination, but because they stop trusting imagination that can’t be justified by the machine.

There’s a second, less visible danger that compounds this one. It’s about what happens to the ideas that do get conceived, once they enter a world where AI drives gatekeeping. Not just on publishing platforms, and funding algorithms, but inside organizations.

It’s a manager who runs a proposal through an AI tool before greenlighting it. A client who asks for validation before approving an original direction. A colleague who reports back that an idea “scored low on feasibility.” A hiring process that screens for a pattern of past success rather than shape of future potential.

At every layer, the filter is the same: does this match what has worked before? Radical ideas may still get born. The problem is they have a harder time surviving the rooms they have to pass through due to the invisible walls created by AI. Thus, the threat to innovation is psychological and infrastructural.

The third way: humans in the loop, with a method

This is why I’m convinced the answer isn’t all-out AI or zero AI. The answer is AI + human judgment, guided by a method that protects what AI can’t replace.

Design thinking can be that method. “Design” is most associated with how things look. But “design thinking” is a human-centered problem-solving process. It’s a disciplined way of framing questions, testing assumptions, and learning from reality that applies to anything: a product, a strategy, a curriculum, a client relationship, a hiring decision. This includes problems where no data exists. Where the market hasn’t formed, the behavior hasn’t been measured, the future hasn’t happened.

AI requires data to recommend. Humans can imagine a world that doesn’t exist and work backwards to it.

That capacity, to conceive of something genuinely new and then build toward it, is what the permission structure quietly erodes. Here is why designing thinking works. It forces a loop that AI can’t complete on its own:

Frame → Prototype →Test → Learn

A prototype can be a napkin sketch – just enough to get a reaction to a concept or idea.

Crucially, this process keeps validation anchored in human reality rather than model confidence. It doesn’t just keep humans in the loop. It gives humans a reason to trust their own judgment over the machine’s. It helps you see the walls which is a prerequisite to getting past them.

Where AI flattens innovation

AI systems trained on existing data are fundamentally prediction engines built to find the most plausible path between what already exists. That makes them ill suited for the leap to the adjacent possible: the space beyond the walls of the known, close enough to reach but too original for any model to have predicted.

Anyone who’s studied business recognizes the pattern. Organizations built on early innovations by visionary founders stagnate with small improvements and eked out efficiencies to be upset by a startup that sees past the corporate walls.

AI, deployed the wrong way, can accelerate that stagnation. LLM output is substituted for human contact: personas of people never met, best practices relieve lived tensions, staid concepts fit in instead of stand out, strategies sound smart yet feel interchangeable.

AI itself is novel and appears to be fast and cheap. But by definition it converges toward the middle, toward the expected, toward what has already been validated by the aggregate of human output. Genuine originality, grounded in human reality and shaped by real judgment, will become rarer than ever.

Why design thinking helps (in practical terms)

Design thinking mitigates the loss of innovation risk because it changes what you optimize for and where you look for proof. And it works whether you’re new to AI or already deep in it. It’s less about correcting a habit than building the right one from the start.

It prioritizes problem framing before solution picking

AI will happily answer any question you ask. Design thinking asks whether you’re asking the right question. Reframing is where originality often lives. If you only optimize inside an inherited frame, you’ll get better and better at solving the wrong problem. Reframing is also how you become aware of walls. The assumptions so embedded in how you’re thinking that you stopped noticing they were assumptions at all.

It replaces “AI says” with “reality says”

If AI becomes the validator, teams outsource judgment. Design thinking moves validation back into the world of prototypes and user behavior. You don’t need permission from a model when you can say, “We built something small and tested it. Here’s what people actually did.” This is also the antidote to the infrastructure problem. When you can demonstrate real-world proof, you don’t need the algorithm’s blessing to proceed.

It trains the habit AI can quietly erode: human agency

One of the subtler effects of AI is the habit of letting the tool do the thinking you used to do yourself. You reach for it before you’ve sat with the problem, before you’ve let the uncomfortable question breathe.

Design thinking builds the opposite habit of mind: observe, synthesize, imagine, test, learn, repeat. These aren’t soft skills. They’re capabilities that keep unexpected ideas alive long enough to find an audience. Design thinking keeps you oriented toward the adjacent possible rather than the algorithmic average.

Two paths (a quick mental model)

If you want a simple operating principle: AI is a powerful teammate. It shouldn’t be the judge. Whether you’re just beginning to use AI or have been using it for a while, it helps to see the two patterns it tends to create and to decide deliberately which one you want.

Click on the graphic to download a PDF.

The first path is faster and cleaner. It’s the one that forms without noticing. Ask, get an answer, justify it, move on. It’s not wrong exactly, but it keeps you inside the walls, optimizing the known and mistaking the plausible for the possible. The second path feels slower at the start. But it protects originality because it stays anchored in human reality where the proof of concept is the world, not just the algorithm.

Innovation isn’t dead. But it’s no longer default.

In an AI world, “good enough” will be cheap. Polished content, plausible strategies, and decent ideas generated fast. All of it living comfortably within the walls of what’s already been tried because no one pointed it elsewhere.

The advantage shifts to teams who can see past those walls and make leaps toward the adjacent possible. But that capacity doesn’t happen automatically. It has to be trained and the humanities are where that training has always lived.

Shakespeare doesn’t just teach you plot. He teaches you to sit with ambiguity, hold contradictions, and understand humans who don’t behave rationally. Storytelling doesn’t just teach you to communicate. It teaches you to imagine a future that doesn’t exist yet and make others believe in it before evidence arrives. Those aren’t decorative skills. They’re exactly what AI can’t replicate.

That’s why I believe the Freeman College of Management’s Markets, Innovation & Design (MiDE) approach matters now more than ever. Business education rooted in the liberal arts. The skills often dismissed by data-driven business curriculums turn out to be the ones that matter most in a data-driven AI world.

MiDE trains students to work at the intersection of markets, innovation, and human-centered design, so AI becomes a tool inside a disciplined thinking process rather than the thinking process itself.

There is one final irony worth considering. AI itself was born from the human capacity this piece is arguing to protect. Alan Turing imagined a world that didn’t exist yet, with no data to confirm it was possible, and leaped. The question now is whether what human AI visionaries built will illuminate that capacity in us or quietly dim it.

Your turn

As AI moves further into your work: what role do you want it playing in how you think and create? The walls we stopped seeing with social media took years to become invisible. With AI, most of us are still early enough to decide what we let form and what we don’t.

A confession.

I wrote this post with AI assistance for research, drafts, and pushing my thinking further than I could alone. That’s not the issue. At one point, I asked Claude whether a thought was good enough and then caught myself wondering whether it would perform well on LinkedIn. Two different tools. The same instinct. Both optimizing against past data. Neither capable of telling me whether an idea is genuinely new or worth pursuing. I caught myself doing exactly what this post argues against. I’m not immune to this. Neither, I suspect, are you.