The Three Horizons of AI: How to Tell Hype From Value

The clearest way I’ve found to tell AI hype from AI value didn’t come wrapped in either fear or worship. It sorts any new technology into three horizons, and once you see the sorting, you can’t unsee it. I picked it up from a conversation with a strategy executive who has spent twenty-five years in financial services, and it’s reshaped how I read every new tool since.

Think of the three horizons as three doors in a hallway. Behind the first door, people are rebuilding furniture that already exists. Behind the second, they’re sketching new designs. Behind the third, they’re inventing things no one has named yet. Most of the crowd is jammed against the first door, congratulating each other, while the room with the actual value sits nearly empty at the end of the hall.

The hype cycle is a racket, and the people who profit from it are worse. AI is a tool, and a useful one, but most of what’s being sold as a revolution right now is old software with a chat box on the front and a bigger price tag. The people who pay for that mislabel are business owners who can’t afford to guess wrong.

What are the three horizons of technology?

The first horizon is replication. You take a thing people already do and make a tool that does it. Pull text out of a PDF. Draft a document from a prompt. Summarize a meeting. These are useful, and none of them are new. People built software for every one of these tasks long before the current wave, and the wave mostly repackaged them with a chat box in front.

The second horizon is where new models appear: a different way of doing the task, or a capability that didn’t exist in usable form before. This is harder, rarer, and slower. It’s where real progress lives, and it takes patience and money that horizon-one repackaging doesn’t.

The third horizon is the cutting edge, and it’s where the value is. Things that don’t exist today: new business models, new inventions, capabilities that change what’s possible instead of speeding up what was already easy.

Most people right now are in the first horizon. Everybody is recreating everything and calling the recreation a breakthrough, and it drives me a little crazy to watch smart people applaud it.

Watch: AI Hype Vs. Reality: What the Channel Is Really Seeing, CRN

Why is most of the AI boom stuck in the first horizon?

Because the first horizon is where the easy money and the easy applause are. You ship a tool that writes emails, and a million people say the future has arrived, when the accurate description is that email got a new front end. There’s nothing wrong with a good front end. The problem is calling it horizon three when it’s horizon one, because that mislabeling drives billions of dollars of decisions.

The vendors come out of that mislabeling fine. The owner who thought he was buying the future and got a faster email writer is the one left explaining the invoice to his partners, and I doubt the vendor loses a minute of sleep over it.

The force behind the crowd is group-thinking. Everybody jumps into the same square at the same time, chasing the same infrastructure, the same chips, the same data centers, because everyone else is. I’d question the loudest voices in the field hardest, the ones who partner with people, absorb their information, and reappear with something that looks a lot like what already existed. Hype spreads socially, and greed keeps the game running long after the smart money can see the ceiling.

How do you use the three horizons on your own decisions?

The structure does more than judge the industry.

It’s a filter for your own choices. When a vendor pitches you an AI tool, ask which door it stands behind. If it does something people already did and just makes it faster, that’s horizon one. Fine, buy it if the speed is worth the price, but don’t overpay as though you’re buying the future. If it does something new to your work, that’s worth a harder look and a bigger bet.

I spent 33 years in enterprise IT, and a demo that makes an old task look new doesn’t impress me. Ask the salesperson what the tool does that your current software couldn’t do at all. If they can’t answer in one plain sentence, you’re looking at horizon one priced like horizon three.

I’d stay on the cutting edge without getting onto the bleeding edge. Years ago I was the lead designer on a product meant to improve DECnet, Digital Equipment Corporation’s networking suite. It looked like it was going to be the next big thing. It died anyway. It’s fine to be early, but make sure what you’re betting on is at least tried and true. The right technology is the one that’s accepted and works.

I use AI every day, and I use it as a horizon-one tool on purpose.

It outlines, it summarizes, it turns up angles I hadn’t considered. I’ve written about exactly where it’s worth using and where it wrecks the work in Six Places AI Will Break Your Work and its companion piece on where it breaks. Horizon one has its uses. What matters is knowing which horizon you’re standing in, so you neither overpay for repackaging nor miss the rare thing that’s new.

Is AI hype, then, or is it real?

It’s both, and the three horizons let you hold both at once without lying to yourself. The current wave is mostly hype, because most of it is horizon-one replication sold as revolution. The underlying technology is also real and will matter, because horizon two and horizon three are coming, on a slower clock than the marketing claims. The people who lose the most can’t tell the difference. They either dismiss all of it as a fad or buy all of it as destiny, and both mistakes cost real money.

I’ve argued against both extremes in The Doomers and the Hypers Are Both Wrong.

The three-horizon structure is the antidote to both errors. Sort the tool into a horizon. Pay what that horizon is worth. Watch the far door for the rare thing that’s new, and ignore the crowd cheering at the near one. There’s a sharp read on where the AI infrastructure race is heading in this conversation on seeing the whole chessboard.

I think the hype merchants are doing real damage, and this sort is the cheapest defense I know against them. Put every tool through it before you sign anything. Owners who skip that step pay revolution prices for repackaged software, get burned, and then swear off the whole technology right when the useful part finally shows up.

If you’re a leader trying to think clearly about AI without the fear or the worship, the whole cluster is built for that. Start with the AI and Writing Hub, and if you’re weighing whether your own hard-won expertise belongs in a book before the field shifts again, my writing services are where that conversation starts.

Frequently Asked Questions