When It Comes to AI, Think Waves, Not Curves

Every week, a new AI model or tool makes headlines, sometimes for jaw-dropping capabilities, other times for glaring limitations. And like clockwork, the reactions split down familiar lines:

AI fans declare, “This changes everything!”

Skeptics reply, “It’s just another iteration of the same old, hallucinating, untrustworthy tech. It’s not ready for primetime.”

Both camps are missing the bigger picture.

The exuberant voices often blur the line between true innovation and surface-level updates. And the naysayers? They’re stuck applying outdated frameworks to something that doesn’t fit their mold. Many business professionals, especially at the leadership level, have tried out a popular tool like ChatGPT, hit its limits, and walked away with a generalized takeaway: “AI just isn’t there yet.”

But that’s a misdiagnosis. The problem isn’t AI. The problem is treating it like it’s one thing, moving at one pace, on one linear path.

It’s not. It never was.

AI Isn’t One Thing

Let’s start with the obvious but often overlooked reality: there is no single “AI.”

What began with a few general-purpose models has since exploded into a multi-dimensional ecosystem. Today, “AI” spans:

• Foundation models like GPT, Claude, Gemini, and Mistral
• Image generation tools like Midjourney and DALL·E
• Predictive ML models fine-tuned for specific industries
• Agent orchestration frameworks, MCPs, custom agents, and copilot ecosystems

Trying to evaluate all of this based on a single tool or use case is like reviewing the entire transportation industry after riding a scooter.

To be fair, Gartner doesn’t treat AI monolithically. Their research breaks it into meaningful categories. But most professionals don’t live in Gartner’s reports. and when their first experience with a model includes hallucinations or shallow performance, it’s easy to dismiss the entire field.

Why the Hype Cycle Doesn’t Fit

The Gartner Hype Cycle is designed for technologies that are stable, self-contained, and evolve on a predictable curve: early excitement, followed by disappointment, followed by maturity.

That framework works for things like Private 5G, augmented reality, or a new cloud platform.

But AI doesn’t move like that.

AI moves in waves, fast, overlapping, and often unpredictable. One week, it’s a model with a 1-million-token context window. The next, it’s a lower-cost open-source challenger outperforming GPT-4. Each model or tool might have its own mini-hype curve, but trying to apply the same arc to the entire category leads to bad conclusions.

Some waves fade. Some crash. Others combine and amplify. And under the surface, there’s deeper momentum building that many miss because they’re still watching the splash from the last wave.

We Haven’t Even Digested Last Year’s Breakthroughs

There’s another dynamic at play that’s especially important for leaders to understand: technology digestion rates.

In any fast-moving space, there’s a delay between the release of a new tool and the industry’s ability to truly leverage it. A great metaphor for this comes from video games, when a new console is released, early games barely use its full potential. It takes time, experimentation, and experience before developers unlock what’s really possible.

The same applies to AI. It takes time to:

• Learn how to prompt well
• Understand how to integrate models with existing systems
• Tune for edge cases and operational reliability
• Move from experiments to real, value-generating deployments

But while we’re still learning how to harness last year’s breakthroughs, new tools keep dropping faster than we can absorb them. And we’re not magically getting better at digesting faster.

This is especially true with GenAI, where emergent behaviors often appear months after release, once people push the boundaries and begin to connect the dots across different tools and workflows.

Strategy for Navigating Fast-Moving AI

So how do you make smart decisions in the middle of the chaos? Here are two practical approaches that can help you stay focused:

1. Architect for Swapability

Don’t let the pace of change paralyze you. Instead, build systems with modularity and interchangeability in mind. Use abstraction layers. Keep your orchestration flexible. Make it easy to replace one model or tool with another.

That way, when something meaningfully better comes along, you’re ready to experiment without re-architecting everything.

2. Give the Giants Room

When evaluating what to build, ask yourself:

• Is this something OpenAI, Google, or Anthropic will likely offer in the next six months?
• Or is this something too niche, too domain-specific, or too tightly tied to your proprietary data for them to bother with?

Avoid building in spaces that will be commoditized. Focus instead on areas where the big players can’t or won’t go, either because the market’s too small for them or the problem space is too specific to your business.

That’s where your differentiation lives. That’s why Keryk focuses on the spaces that cannot be commoditized, because “Big AI” can’t see inside SME’s brains (yet?).

How to Stay Grounded

Here’s a better mindset for leaders trying to cut through the noise:

• Don’t generalize. One tool’s failure doesn’t reflect the whole field.
• Evaluate each wave individually. Know the strengths and limits of each model or tool.
• Assume rapid improvement. Today’s blockers may not exist next quarter.
• Focus on business value. Let use cases, not hype, guide your choices.
• Budget for exploration. True understanding takes time and iteration.

Closing Thought

AI doesn’t follow a hype curve. It moves in waves. Rolling, crashing, combining, and reforming, at a rate that is much faster than we can fully grasp.

Don’t get stuck on the last wave. Don’t wait for the surf to settle. It won’t.

Instead, learn to read the patterns. Build flexibility into your approach. And keep your eyes not just on the wave in front of you, but on the horizon, where the next opportunity is already forming.