How the tech industry can avoid another AI winter

Shelf awareness

As technology leaders, we’re all too familiar with the problem of shelfware. You procure a new piece of software and after a short burst of enthusiasm, user adoption tails off. Soon your shiny new solution becomes just another piece of shelfware, collecting digital dust in your already crowded tech stack.

One US study found that 37% of software bought by businesses goes unused, wasting an estimated $30 billion. That shows shelfware is already a huge problem. But when it comes to AI, the bottom line is bigger than our budgets.

AI is a new technology. Many are still forming opinions about it. So every time we don’t deliver on the promise of its potential, we chip away at people’s enthusiasm for AI and diminish the potential of any future investment in it. But what if we’re not the only business failing to deliver results from AI?

The AI winters

Today’s AI hype might feel new, but it isn’t. The earliest period of AI enthusiasm was in the 1950s when AI first emerged as a concept. Researchers stunned the world with AI programs that could solve difficult mathematical problems. There was even a natural language chatbot named ELIZA with dialogue so convincing that users couldn’t tell they were talking to a machine.

This age of AI made people imagine a world where menial tasks, once considered the exclusive domain of humans, would be taken care of by computers, inspiring countless books and films on the topic. In 1965, H.A. Simon — a prominent political scientist who wrote extensively on AI — said, “machines will be capable, within twenty years, of doing any work a man can do.”

But this world failed to materialize. People’s expectations were soon confronted with the reality that delivering this imagined world was much more complicated in practice. Soon after, funding for AI dried up. This would be the world’s first AI winter — a period where AI research all but disappeared. But it would not be the world’s last AI winter. Because it would be followed by another winter that started in the late 1980s and in many respects lasted right up until the early 2010s.

How to avoid another AI winter

Each of the AI winters had one thing in common: AI could not meet people’s sky-high expectations. Today, we’re facing another period of hype, with the promise of AI yet again the topic on everyone’s lips, describing the same sort of optimistic visions of our AI future.

The difference is that this imagined AI future is now within our grasp. The technology behind AI has come a long way to the point where AI genuinely has the potential to transform the world in the coming years.

But for that to happen, we need to nail AI adoption. The problem is that while the technology has developed at lightning speed, the thinking on how we implement AI from a change management perspective has slowed to a halt.

If we want to see the future AI promises, we have to not only progress the technology but how we approach user adoption. The secrets to AI adoption can no longer be the closely guarded secrets of companies that succeed with AI.

We need to start sharing those secrets with the world. If we don’t, the AI solutions we’re all so excited about risk becoming just another piece of shelfware and we face yet another cold AI winter.

This article was originally written and published for Forbes Technology Council on 5 February 2024.