Over the past few years, artificial intelligence (AI) has been one of the most popular topics of discussion as its capabilities grow and we get closer to seeing the technology actually being deployed. According to McKinsey, there have been twice as many articles referencing AI in 2016 as in 2015 and four times more than in 2014 — a trend that appears set to continue. But are we nearing the end of all this hype?

I was recently in a meeting with some of my colleagues where we were discussing the ongoing exposure around AI. A few attendees mentioned that they felt that AI hype reached its peak last year. I had to disagree; I think we’re still a couple of years from approaching the peak and another couple of years still from seeing widespread adoption of AI technology. Gartner shares this view and has made AI one of its top trends of the current hype cycle, claiming it has another two to five years before it reaches its peak.

The obsession with whether we’ve reached this peak reminds me of the ongoing discussions surrounding peak oil. Since 1956, people have been making predictions about when we will see peak oil demand — the point at which demand for oil stops increasing and slowly begins a terminal decline. While AI hype, like oil demand, has still not yet reached its peak, AI is in an interesting position since this is not its first hype cycle. Although AI interest is still growing, AI has not yet left the second stage, the trough of disillusionment, from its previous hype cycle 30 years ago.

From 1973 through the 1990s, no one wanted to fund anything related to AI, and interest in researching the technology waned. Frustrated by a lack of progress, pessimism among those in the AI community spread to others, and — much like the dot-com bubble — that interest level burst and led to what is now known as AI winter. Much of what we’re seeing now (a tidal wave of AI companies, continuous coverage of the latest news in AI, etc.) is eerily similar to the AI landscape of the 70s and 80s before the winter.

So what’s going to happen next? Will AI be able to overcome the burden of unmet expectations from 40 years ago, or are expectations so inflated regarding AI’s capabilities that we’ll see, to some degree, the same cycle of disappointment and limited investment that we’ve seen before?

In my opinion, this time the industry will be able to learn from the mistakes of the last hype cycle, and we’ll see real progress in understanding how AI can be used to automate and solve critical business issues, with businesses investing in the foundation for its introduction.

What will ensure AI’s success and help it emerge unscathed from the hype cycle will be plugging AI into the very fabric of every enterprise and making it part of the decision making process? This will be a particular challenge because it will require companies to make a fundamental change to how they think about business and how they build strategies. However, it will also require companies to think about how they will overcome employee and customer skepticism before they can reach the ultimate end goal of completely automating processes, whatever those may be.