The enterprise world is awash in AI optimism. Boardrooms buzz with talk of transformation, and budgets swell to accommodate the latest platforms and AI assistants.
Today, nearly three-quarters of companies report using generative AI regularly in a core business activity, according to recent McKinsey & Company research. But if you look past the headlines, a familiar pattern has emerged that reminds me of the dot-com era, when companies rebranded overnight, and investors chased the next big thing, often with little regard for a tangible way forward.
Today, talk of an AI bubble isn’t just a matter of market speculation or start-up hype. It’s unfolding inside organizations, where the pressure to “do something with AI” drives rapid procurement and internal investment. The fear of missing out (FOMO) is palpable. Leaders, wary of being left behind, greenlight projects that promise automation, insight, and competitive advantage.
But here’s the sobering reality: Most AI initiatives fail to deliver meaningful results, and the graveyard of underutilized tools and proofs of concept or pilots that don’t pan out is growing.
THE ANATOMY OF THE AI BUBBLE
The pitfalls are easy to spot, especially in hindsight. Companies rush to build bespoke AI solutions, convinced their needs are unique, only to watch those features become commoditized by vendors months later.
Others buy off-the-shelf platforms, expecting plug-and-play magic, but end up mired in costly customization and integration. Some grant AI agents sweeping permission, only to recoil at the risks when their chief information security officers (CISOs) push back.
While it’s completely speculative, if the bubble bursts, it will be driven by a flood of AI projects without clear use cases that fail to generate revenue, productivity gains, or measurable cost savings. That doesn’t mean every AI project is doomed, or that companies should stop investing in AI. It just means you need to tweak your approach.
THE PRAGMATIC PLAYBOOK
What separates the AI survivors from the casualties is discipline. The most successful organizations approach AI with a pragmatic, four-step framework.
1. Assess with brutal honesty
Is the problem unique, or is it a feature waiting to be bundled into next year’s subscription? The build versus buy decision is not just technical; it’s existential. Too much building leads to wasted effort; too much buying without adaptation leads to disappointment.
Leaders need to ask themselves if their processes are so distinctive that custom development is warranted, or if a trusted third party is likely to deliver the needed capability as a feature in one of their offerings.
2. Pilot before you leap
Small-scale experiments, tightly scoped, reveal both the functional and business value of a solution. Pilots aren’t proof of concept. They’re proof of value.
I’ve seen instances where the value looked to be there in the pilot but disappeared in a larger scope. So, resist the urge to roll out enterprise-wide until the groundwork is solid. Use pilots to continue to verify the value and readjust as needed as you scale the project.
3. Verify real impact
Does the AI do what was promised? More importantly, does it make or save money?
Functional success is meaningless without business impact. Continual verification means demanding evidence that the solution delivers measurable value. Ensure it remains a part of the process even after you roll out your AI, because bolting on the need to continually verify won’t end well.
4. Scale with caution
Prioritize quick wins and expand with measured steps. Only after value is proven does scaling make sense. Organizations that scale AI use prematurely often find themselves burdened with tools that fail to deliver on their promise.
AVOID THE COMMON PITFALLS
All too often, I see my peers fall into the same traps. Here’s my advice:
- Don’t overestimate AI’s ability to automate complex workflows. Many projects fail because organizations expect too much, too soon.
- Beware of internal FOMO because rash decisions driven by fear of being left behind often lead to wasted investments.
- Recognize the value of “traditional” AI. Not all innovations are generative or agentic. Mature, proven AI solutions can deliver immediate value.
- Guard against giving AI agents more autonomy than is reasonable. The risks are real, and your CISO is justified in being overly cautious.
THE DISCIPLINE TO THRIVE
AI is not a panacea, nor is it a guaranteed path to transformation. Its power lies in the hands of organizations that approach it with rigor, clarity, and strategic intent. The bubble will burst for those who chase hype over substance, mistake activity for progress, or fail to align investments with business strategy.
The leaders who will thrive are those who move deliberately and rationalize every decision, demand real value at every turn, and recognize both the limitations and the possibilities of AI. Success comes from understanding where AI can deliver immediate, tangible benefits and where it remains a work in progress. It means balancing ambition with pragmatism, and performance with transformation.
Much like the dot-com boom, opportunity will outlive the initial hype and early failures of AI. Every company needs to lean into the potential, but every opportunity also needs the right framework to thrive and grow. That means setting the right foundation: modernizing your data estate, determining where your data should live—whether in the cloud, on premises, or a hybrid of both—and then ensuring your networking, compute, storage, and security infrastructure can support it all.
Equally important is recognizing the growing role of the edge. Whether it’s a retail floor, remote offices, a factory, or the device in an end user’s hand, every edge has the potential to drive AI use cases from cloud-connected services to local inference.
The future belongs not to those who move fastest, but to those who move smartest. Step back, challenge assumptions, and invest in AI with discipline and foundational purpose. In doing so, your next AI initiative will be the one that outlasts the hype to deliver lasting value for your business.
Juan Orlandini is the CTO, North America for Insight Enterprises