Use the Calm to Build: Why AI’s Pause Is the Best Moment to Accelerate

We’ve all seen the headlines: Gartner flags generative AI sliding toward the “Trough of Disillusionment,” and high-profile industry voices have even begun warning of an AI bubble burst. 

Let’s keep our heads. This isn’t a bubble popping—it’s the normal adjustment that follows every technology surge. We saw it with cloud, with CRM, even with enterprise email. The pattern is familiar; the opportunity is real.

The eye of the hurricane—and what to do with it

Think of the last two years like the leading edge of a storm. ChatGPT hit like a squall. Many leaders lacked time (and frankly, viable enterprise options) to prepare, so teams did the rational thing: they built. 

Now we’re in a period of relative calm—the eye—before the back side of the storm arrives. In the Northern Hemisphere, that trailing right-front quadrant is often the most intense. Translation for the enterprise: the real force of AI-driven change is still ahead. Use this calm to board up properly: governance, compliance, safety, testing, auditability, integration.

What the early builders got right—and wrong

Two years ago, there was little to buy. If you wanted enterprise-grade AI, you built prototypes on top of public LLMs (which you rightly couldn’t expose to sensitive data) or you stood up DIY infrastructure. That work wasn’t wasted. It taught you the contours of your data, where value lives, which risks matter, and where internal limits—model runs, scaling, ops—start stifling progress.

But here’s the hard truth from decades in the chair: when you keep custom-building past the learning phase, you spend money you don’t need to spend. I’ve watched teams sink $3–6M into bespoke scaffolding before admitting that the market has caught up.

We’ve seen this movie before. Cloud began with “build your own private cloud,” racking hardware and rolling your own. Over time, that gave way to SaaS, then to fully managed cloud platforms. Email followed a similar arc—from on-prem stacks to Microsoft 365 and beyond. In every case, platforms won because they industrialized the hard parts and let you focus on outcomes.

AI is following the same playbook, only faster.

Buy vs. build, reframed

This is not a call to stop investing in AI. It’s a call to stop burning cash on undifferentiated plumbing. Only a handful of organizations can justify the ongoing infrastructure spend required to operate and evolve multi-model AI at scale. 

For everyone else, the smart pivot is to stand on the shoulders of platforms that have already solved the hard stuff. Even MIT’s recent research, while grim on internal build rates, found that initiatives leveraging external platforms were moving faster toward production and impact. In my advisory work, I’ve seen how platforms such as Vertesia can shorten time to value—helping organizations build, deploy, and scale AI apps and agents without the overhead of reinventing the wheel.

I’m not a salesperson; I’ve spent 36 years buying and running enterprise technology (CIO at AXIS Capital, Chubb, Travelers, and AIG) and advising a slate of high-tech companies. From that vantage point, here’s what matters now: an end-to-end approach, not scattered point tools; freedom to swap models as they evolve; low-code so business ideas can become governed agents in days, not quarters; accelerators that move projects into production faster; and enterprise-class operations—role-based access, auditability, and cost controls.

The moment is positive – if you make the pivot

Call it a lull if you want; I see a gift. If you’ve invested heavily, great—accelerate on top of what you’ve learned. If you stayed in the background, that’s fine too—adopt with the benefit of hindsight. Either way, use this calm to get your house in order and put your foot on the gas. The stronger side of the storm is still ahead—and the organizations that prepare now will come out of it stronger, faster, and measurably better.