The question everyone’s asking isn’t whether we’re in an AI bubble – it’s what happens when it bursts. But before we get caught up in the latest panic-inducing headlines, we might be asking the wrong question entirely.

Markets moved sharply on news of a MIT study claiming 95% of AI pilots are failing. The methodology was questionable at best – essentially 52 interviews and some public announcement reviews – yet it wiped billions off market caps. That tells us something important: when investors are this jumpy, we’re definitely in bubble territory.

But here’s where it gets interesting – and where we need to resist the urge to panic.

The bubble question: Missing the point

Yes, there’s almost certainly a bubble in AI stocks. When US$44 billion flows into AI startups in just the first half of 2025 – more than all of 2024 combined – and when questionable studies can trigger massive sell-offs across tech companies, you know market sentiment has gotten ahead of reality.

But the overreaction reveals something more important than a stock bubble: we’re making investment decisions based on headlines rather than substance, and missing the real story about what’s happening with AI in the enterprise.

Learn more: How AI is changing work and boosting economic productivity

But here’s what the doomsayers are getting wrong: the technology itself is genuinely transformative. The failure isn’t in the AI models – as MIT researcher Aditya Challapally notes, failures were “less about the quality of the underlying models and more about how organisations attempt to use them”.

This isn’t the dot-com bubble, when companies with no viable business model were getting billion-dollar valuations. It is something more nuanced and ultimately more correctable.

Why AI pilots failing is actually good news

Here’s what multiple sources are consistently telling us: most corporate AI initiatives aren’t delivering the expected returns. But that’s not because AI doesn’t work. It’s because organisations are approaching it all wrong.

The core problem isn’t the technology – it’s what we might call a “learning gap.” While tools like ChatGPT work brilliantly for individuals because of their flexibility, they often stall in enterprise environments where they can’t adapt to established workflows and processes.

We saw this play out dramatically in Australia just last month with Commonwealth Bank’s AI chatbot debacle. The bank laid off 45 customer service workers, claiming its new AI system had reduced call volumes by 2000 per week. Within weeks, they were forced to rehire all 45 staff with apologies and back pay. The reality was that call volumes had actually increased, with remaining staff working overtime and management being drafted to answer phones just to keep up.

The Finance Sector Union accused CBA of “an outright lie” about productivity gains, and the bank eventually admitted the roles were not redundant after all. What makes this story particularly instructive isn’t just that the AI failed – it’s that CBA rushed to scale without proper testing or change management, treating AI implementation like a traditional IT deployment rather than the iterative learning process it actually requires.

The CBA experience perfectly illustrates the ownership problem we discussed earlier. This was most likely an IT-led initiative that bypassed the people who actually understood customer service workflows and the nuanced nature of customer interactions.

Think about this for a moment: we’re watching organisations learn how to use electricity all over again. The early factories that simply replaced water wheels with electric motors didn’t see dramatic productivity gains. The transformation came when entire workflows were redesigned around what electricity made possible.

A healthy dose of scepticism (but not cynicism)

The scepticism surrounding AI adoption is not only warranted – it’s essential. If you’re familiar with Gartner’s Hype Cycle, you’ll recognise exactly where we are with AI: somewhere between the peak of inflated expectations and the trough of disillusionment. This is precisely when the most valuable work happens – when the hype dies down and organisations start focusing on real problems rather than technology for its own sake.

Every transformative technology goes through this cycle, and the scrutiny ultimately strengthens the implementations. The companies that are succeeding with AI right now are those that started with specific problems, not generic “AI initiatives” – and crucially, they’re not waiting for perfect studies to validate their approach.

Learn more: How should Australia capitalise on AI while reducing its risks?

There’s wisdom in Simon Sinek’s recent conversation with Steven Bartlett, in which he warned against our obsession with destinations over journeys. “People keep telling us life is not about the destination. Life is about the journey. But when we think about AI, we only think about the destination” – the magical output, the perfect solution. We forget that the real value comes from the process of learning, adapting, and growing alongside these tools.

The pattern among successful implementations is clear: focus, execution, and partnership – not “let’s wait for more comprehensive research” or “let’s build the perfect enterprise solution from scratch.”

The long game: Why change takes time

Here’s where history offers some perspective. In 2000, breathless predictions claimed newspapers would be dead by 2005, email would kill postal services within years, and physical retail would vanish by 2010. Well, it’s 2025, and while newspapers are struggling and physical mail has declined significantly, they’re still here. Amazon revolutionised retail, but it coexists with physical stores rather than replacing them entirely.

This pattern – where change happens more slowly than promised but more completely than sceptics expect – is exactly what we should anticipate with AI. The transformation will be profound, but it will unfold over years and decades, not quarters.

The real challenge: Building the learning organisation

The companies that will thrive in the AI era aren’t necessarily those with the biggest AI budgets or the fanciest models. They’re the organisations that develop what we might call “AI learning capability” – the ability to experiment thoughtfully, fail fast, adapt quickly, and scale what works.

This means establishing proper guardrails that enable experimentation rather than prevent it. It means creating feedback loops that help AI systems adapt to organisational workflows. Most importantly, it means empowering “line managers, not just central AI teams, to take a lead in integrating AI into daily operations”.

The MIT report suggests something profound: we’re not in an AI capability crisis, we’re in an organisational learning crisis. The technology is ready; the question is whether our institutions can adapt fast enough to harness it effectively.

What this means for leaders

If you’re leading an organisation, the findings should be reassuring rather than alarming. They’re telling us that success isn’t about having the best AI models or the biggest budgets – it’s about having the right approach.

Start small. Pick one specific problem that AI can solve better than your current approach. Partner with specialised providers rather than trying to build everything in-house. But here’s the crucial bit: don’t hand this off to IT and hope for the best. Put the business unit that owns the problem in charge of finding the solution.

Give your line managers the authority and tools to experiment. Measure impact ruthlessly, but be patient with the timeline. Create an environment where a department head can try something for three months, learn what works, and either scale it up or shut it down – without waiting for enterprise architecture approval.

Subscribe to BusinessThink for the latest research, analysis and insights from UNSW Business School

Most importantly, resist the urge to chase AI for AI’s sake. The companies succeeding with AI aren’t doing so because they have better technology – they’re succeeding because they have better strategy and clearer ownership.

The bottom line

Is there an AI bubble? Almost certainly. Will it burst? Probably. Will that matter in the long run? Not really.

The real story isn’t about market valuations or quarterly earnings – it’s about a fundamental shift in how work gets done. Yes, it’s taking longer than the hype cycle predicted. Yes, most early implementations are failing. But the underlying transformation is real, and the organisations that learn how to harness it thoughtfully will find themselves with sustainable competitive advantages that compound over the years.

The question isn’t whether you should bet on AI. The question is whether you’re willing to do the hard work of becoming the kind of organisation that can learn, adapt, and improve alongside these powerful new tools.

In the end, that’s not just the path to AI success – it’s the path to thriving in an age of continuous technological change.

Dr Jeffrey Tobias is an accomplished and prominent innovation thought leader and strategist, drawing expertise from the academic, government, entrepreneurial and corporate worlds. He serves as an Adjunct Professor and Fellow at AGSM @ UNSW Business School, and holds a B.Sc (Hons), University Medal and PhD from UNSW Sydney. In 2003, he founded The Strategy Group, where he currently serves as Managing Director.