The problem with the Hype Cycle.

“AI is at the peak of inflated expectations.”

Okay.

So what should we do?

Wait?

Invest?

Build?

Cancel the project?

That’s where I start having questions.

I find the Hype Cycle interesting. It’s a useful way to discuss how expectations around technology can rise, disappoint, and eventually settle into something more practical.

But expectations aren’t the same thing as engineering.

And they’re certainly not the same thing as business results.

The Hype Cycle is a Gartner framework that describes a pattern of expectations. It isn’t a direct measurement of how productive a technology has become.

That distinction matters.

Because when people talk about AI, they sometimes treat a position on the curve as if it tells us what to build, when to invest, or whether the technology is ready.

I’d rather look at the evidence.

History gives us some interesting examples.

Consider electricity.

The electric motor existed long before factories fully reorganized their work around electric power.

Early factories could replace a steam engine with an electric motor and keep much of the same layout.

But the larger productivity opportunity came from redesigning how the factory worked.

Power could be distributed differently. Equipment could be arranged around the production process rather than around a central mechanical drive.

The invention mattered.

So did the redesign.

Now think about computers.

Buying computers didn’t automatically make every business more productive.

Organizations also needed better processes, new skills, reliable data, and changes in how work was managed.

And the internet?

Connecting a business to the internet was one thing.

Redesigning sales, distribution, customer service, and entire business models around connectivity was something else.

These technologies followed different paths. Their adoption, costs, and productivity effects cannot be reduced to one universal curve.

But there’s a useful recurring lesson.

The technology alone was rarely enough. The bigger gains came when the work was redesigned around it.

This is why I’d be careful about using the Hype Cycle as an investment decision.

It tells us something about expectations.

It doesn’t tell us whether a particular AI application will work in your business.

For that, we need different evidence.

• Can the technology perform the task reliably?
• Can it work with the information and systems we already have?
• What does it cost to operate?
• What happens when it fails?
• How much work does it actually save?
• And does the benefit survive outside the demonstration?

There’s another reason history matters.

Technology can become useful long before an organization is ready to benefit from it.

And a business can sometimes improve dramatically using technology that stopped being fashionable years ago.

That’s why I don’t think the right question is whether AI has passed the peak of inflated expectations.

I’d ask whether a specific AI capability is ready for a specific business problem.

Then I’d build a small, measurable example.

If it works, expand it.

If it doesn’t, understand why.

Maybe the model isn’t reliable enough.

Maybe the data isn’t ready.

Maybe the process is badly designed.

Or maybe AI was never the right answer.

That’s not a failure of innovation.

That’s engineering.

The Hype Cycle can help us discuss expectations.

But I wouldn’t let a curve make architecture decisions for me.

Especially one that doesn’t know anything about my client’s business.

Further reading

• Paul A. David, “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox,” American Economic Review (1990). On why electrification paid off only after factories were redesigned.
• Erik Brynjolfsson, “The Productivity Paradox of Information Technology,” Communications of the ACM (1993). On why computers didn’t show up in productivity right away.
• Gartner’s own description of the Hype Cycle methodology, on gartner.com.