As Fast as you can.
Thirty years ago, analysts at Gartner defined a typical path of technology adoption in the marketplace they called the Hype Cycle. According to this model, as any disruptive new technologies is introduced, markets react in five predictable stages:
- The Innovation Trigger: some new technology gains media attention and sparks excitement.
- The Peak of Inflated Expectations: exaggerated claims about its capability climb to a fevered pitch.
- The Trough of Disillusionment: expectations aren’t met, failures become apparent, leading to disappointment.
- The Slope of Enlightenment: the technology’s realistic potential and use cases emerge.
- The Plateau of Productivity: the technology becomes mainstream and delivers sustainable value.
While this model was developed to describe how markets react to disruptive new technologies, it also reflects the way businesses and individuals react to disruptive technologies, like AI. For the most part, businesses and workers prefer the status quo. Disruption is a challenge to effective workflows and requires energy to absorb, so it takes a spark of real excitement and media hype to get most people to pay attention to something truly innovative.
Once the excitement takes off, because we see real potential, belief in transformative capabilities spreads. We move faster toward adoption and begin changing our workflows. It isn’t until the reality of technological limitations hits us head on that we pull back and reassess, often with a sense of disappointment when expectations meet practical challenges and costs.
When we push through the disappointment, we recalibrate our expectations and find the practical applications of the technology. Only then can we figure out the best practices for our own workflows. Only then does the technology become integrated into our new status quo.
Inflated Expectations Hide the Real Path to Success
We’ve all seen this cycle before, and while we can quibble over competing models or the degree to which all technologies may or may not follow the same curve, the general truth is self-evident: When a new technology drives viral market interest, expectations driven by hype inevitably spike before reality sets in with some disillusionment, before any true and sustainable value can emerge.
It’s hard to argue that the entire market for AI isn’t currently riding near the Peak of Inflated Expectations. That isn’t to say there isn’t real transformative value to come from AI. It’s to say that our expectations for access to that value are ignoring the hard path it’s going to take to unlock it. Demos look great, but things tend to break down in production because we haven’t finished the hard work of adapting both technology and workflows; we’ve been led to believe the technology will automagically define new workflows that deliver value.
If you look at the history of CRM, it took roughly 10 years after the technology was logging 50-80% failure rates according to analysts likes of Gartner and Forester for it to become an essential tool for most businesses. It took 10 years to get fully planted on the Slope of Enlightenment.
That wasn’t because the technology was fundamentally flawed. It was because businesses didn’t have workflows built around CRM, much less professional roles with institutional knowledge of how to adapt it to existing business processes, or a culture that recognized the collective value of primary data. It took 10 years to develop the experience, the consulting practices, product support, training and culture to truly adapt both the technology and business workflows to work together seamlessly.
Find and Face the Limitations
The current hype that suggests this evolutionary process of adapting businesses and technology will be leap-frogged by LLMs is exactly the kind of peak exaggeration that will cause many to stumble into the Trough of Disillusionment as their demo agents hit real walls in production.
But this is where I’m advising my colleagues and customers to run faster. Run to the Trough of Disillusionment as fast as you can. Because it’s only when you come face to face with the limitations of AI that you’ll be able to see what it will take to find the Slope of Enlightenment.
If you’re still high on all the great things AI is going to do for your business, you haven’t pushed hard enough. Sure, maybe you’ll be one of the unicorns for whom AI magically propels you to unfettered returns, but generations of experience suggest you won’t. Like everyone else, you’re going to hit a wall before having to roll up your sleeves and do the hard work of adaptive integration. The faster you get there, the faster you’ll be able to start climbing the hill toward productivity.
Assessing the Real Value
For me, the disillusionment came from constant use of AI on engineering projects to the point where I discovered AI was lying to me. Not hallucinating, not confabulating, but lying as that term might be used before firing an insubordinate employee.
Once the exaggerated potential was shattered in my own mind, the problems and limitations with AI only seemed to accelerate until I felt completely disillusioned, even a bit despairing since a lot of my business and work is focused on AI. I dropped several of my subscriptions, I stopped using AI every day for every task, and I discovered that some of my traditional workflows–like writing–actually work better for me without AI. But that’s another post.
For now, I’m carefully reassessing where AI delivers real value in my workflows, and limiting my use to validating whether that value is real. It’s far too easy to lose traction to laziness in letting AI do everything, most critically in the thought processes it’s taken me years to sharpen to my tasks. I *need* that sharpness of thought to find the enlightenment of exactly how my workflows can be genuinely improved sustainably with AI.
As the Hype Cycle model suggests, you’re probably not going to reach enlightenment without first experiencing disillusionment. So if you’re still super excited and don’t see the pitfalls, push on with some alacrity. The faster we all get there the better.
