If you’ve been paying attention to tech news over the past two years, you’d be forgiven for thinking that AI is simultaneously the greatest invention in human history and an existential threat to civilisation. The truth is considerably less dramatic than either extreme, and understanding where we actually are in the hype cycle is useful for making good decisions — whether you’re a business leader, a worker, or just someone trying to figure out what’s real.

A Brief History of AI Hype

AI has been through this before. In the 1960s, early AI researchers predicted that machines would match human intelligence within a generation. That didn’t happen. In the 1980s, expert systems were going to revolutionise business. They didn’t, really. Each wave of enthusiasm was followed by an “AI winter” of disillusionment and reduced funding.

What’s different this time is that the technology actually works. Large language models can genuinely write, translate, summarise, and reason in ways that previous AI systems couldn’t. Image generation models produce results that would have seemed impossible five years ago. The underlying capability is real.

But “real capability” and “the way it’s being marketed” are two very different things.

What AI Actually Does Well Right Now

Let’s be specific. AI excels at text generation and editing — producing first drafts, summarising documents, and translating between languages. It’s a useful productivity tool for writers, not a replacement. AI coding tools measurably improve developer productivity for boilerplate code and debugging. Data analysis benefits from AI’s ability to identify patterns and surface anomalies in large datasets, though human interpretation is still needed. And well-implemented chatbots handle routine customer service queries effectively.

What AI Doesn’t Do Well (Despite the Marketing)

AI models still produce confident-sounding errors — fabricating citations, confusing details, presenting nonsense as fact. They can simulate reasoning but don’t truly understand concepts, meaning they fail in unpredictable ways on novel problems. They can write about medicine or law convincingly but lack the judgment and accountability of actual experts. And despite being creative within existing patterns, genuinely novel ideas still come from humans.

Where the Hype Exceeds Reality

Several claims being made about AI right now are either premature or overstated:

“AI will replace most jobs within five years.” Some jobs will be significantly changed by AI. A few will be eliminated. But the “mass unemployment” narrative ignores how slowly organisations adopt new technology, how much work involves physical presence and human interaction, and how new technologies historically create new job categories.

“AGI is just around the corner.” Artificial General Intelligence — AI that can do everything a human can do — remains theoretical. Current AI systems are impressively capable in narrow domains but lack general understanding. The gap between “really good at specific tasks” and “generally intelligent” is vast and may not be closeable with current approaches.

“Every business needs an AI strategy right now.” Some businesses will benefit enormously from AI adoption today. Others won’t. If your business processes don’t generate or consume large amounts of data or text, the current crop of AI tools may offer marginal benefits at best. It’s okay to wait.

I’ve been reading what the team at Team400 has been saying about this, and they take a refreshingly measured approach: start with a specific problem, assess whether AI is genuinely the best solution, and implement pragmatically rather than because everyone else is doing it.

The Trough of Disillusionment Is Coming

If you follow Gartner’s hype cycle model (and it’s imperfect but useful), we’re probably somewhere near the peak of inflated expectations for generative AI. The next phase — the trough of disillusionment — is when reality catches up with marketing, failed projects accumulate, and the narrative shifts from “AI can do everything” to “AI doesn’t work.”

Both extremes are wrong. AI is genuinely capable and genuinely limited. The companies that navigate the hype cycle best are the ones that maintain realistic expectations throughout, investing steadily rather than swinging between frantic adoption and bitter abandonment.

How to Think About AI Practically

Here’s a framework that works: separate capability from readiness — just because AI can do something in a demo doesn’t mean it’s production-ready. Look for augmentation rather than replacement, since AI plus humans consistently outperforms either alone. Start small and measure results rigorously. Stay sceptical of vendor claims — every software company is now an “AI company,” and many have just bolted a language model onto an existing product. And keep learning, but don’t feel pressure to adopt everything immediately.

The Bottom Line

AI is a transformative technology. It’s just not transforming everything, all at once, right now. The hype cycle is doing what hype cycles do — amplifying expectations beyond what the technology currently delivers. The reality is impressive enough without the exaggeration.

In five years, we’ll look back and see that AI changed many things about how we work and live. We’ll also see that many of the most breathless predictions didn’t come true. The people who did best will be the ones who stayed curious, stayed practical, and didn’t let either enthusiasm or fear drive their decisions.