Nitin Rakesh is the CEO and Managing Director of Mphasis and coauthor of the award-winning book "Transformation in Times of Crisis."gettyIn 1937, Malcolm McLean sat at a New Jersey dock watching workers unload his truck by hand. Before he arrived, those same workers had been loading ships at 1.3 tons an hour for centuries. McLean didn't ask how to build a better vessel. He asked a more dangerous question: What if the cargo never had to be touched at all? The answer was simple: a standard container, one box that could fit any ship, port or truck, so the entire system could reorganize itself around a single unit. Loading speed jumped to 30 tons an hour. Costs dropped 97%. Within 17 years, container ports went from serving 1% of countries to 90%. Not because the ships got faster, but because everything around them finally could.Enterprises investing in AI are making the opposite bet. They are paying for intelligence to compensate for rigidity. But it doesn't work that way. AI accelerates whatever structure it sits inside. Fast organizations get faster. Rigid ones get more trapped. When the operating model cannot evolve, AI doesn't overcome inertia. It funds it. The constraint has always been the speed at which the system around intelligence can change. McLean understood that at a New Jersey dock in 1937. Most enterprises are still paying consultants to explain why they can't.Intelligence Is There. You Just Can't See ItEvery technology wave since the internet has been absorbed the same way: new capability layered on top of unchanged cores. Banks got a portal. Then a mobile app. Then somewhere more cost-effective to host the same logic. Each time the surface changed. The core didn't. The business logic written in COBOL in 1978 continued to govern decisions made in 2025. The vessel stayed the same. Only the packaging changed.What this created is an intelligence problem. The accumulated logic of how an enterprise works, how it prices risk, clears a payment, processes a claim, underwrites a policy, lives in source code modified by hundreds of engineers across decades, never fully documented, so deeply intertwined that nobody alive can say with certainty what is live logic and what is dead code. A Thoughtworks and IDC study of 500 large enterprises confirmed nearly 90% have adopted AI tools, yet 90% remain stuck in reactive modernization cycles. Only 12% have made the structural shift, and those that have are not just modernizing faster. They are competing differently.The typical reflex is to start a program. A multiyear, episodic transformation initiative with a current state, a target state and a timeline nobody in the room genuinely believes. The problem is not execution, but architecture. A seven-year program freezes the organization in its current assumptions for seven years, and in a market moving at breakneck speed, seven years is not a road map. It is a surrender document.The Knowledge Buried In The CodeThe instinct when confronted with a legacy system is to translate it. “Take the COBOL, convert it to Java. Move off the mainframe. Re-host in a container.” You get to declare victory without understanding the change.Three years. A Java application that does precisely what COBOL did. A monolith still at a different address. Forty years of decisions, regulatory adaptations and market responses have never been extracted. Carried across and buried again, accumulating the same debt under a different name.While working through some of the largest legacy estates in banking and payments, we have repeatedly found the same missing step: a failure to translate the code while extracting the knowledge within it. When you can read what a system actually does, map it to business language and build forward from that, you stop guessing at your own logic and start owning it. That knowledge needs a permanent home: a continuously updated intelligence layer where every modification feeds back so the organization never has to reverse-engineer itself again.That layer does not stop at the code. It trains entire systems that run the enterprise once built. When an agent has access to a system's full incident history, it can predict failures before they surface and resolve them before the business feels anything. This shift from reactive to predictive is made possible when enterprise intelligence is continuous.One engagement we ran recently involved 50 million lines of legacy code that a traditional program had estimated would take seven years to modernize. It took 18 months not by moving faster through the same process, but by changing what the process was built on. The client came out with something it had never possessed: a complete, living record of its operational logic.Research published this year confirms the pattern holds across industries: Top-performing organizations have already shifted their focus from efficiency to velocity, and those developing strategies iteratively are growing measurably faster than those still planning in annual cycles. The gap between organizations that have made that shift and those that haven't is widening every quarter.Gartner estimates that fewer than 5% of enterprise applications are currently integrated with AI agents, and projects this will rise to 40% by the end of 2026. For organizations that can change fast enough, that is an opening. For those still debating where to start, it is a deadline.The Only Question That CompoundsBoards keep asking whether their enterprise is AI-ready. It is the wrong question. The right question is whether the enterprise can change: its systems, its decisions, its operating model and the teams operating all three. An organization whose systems are modern, but whose teams are still structured for a waterfall world, has not solved the problem. It has moved it. Speed to change is an organizational condition, not a technical one.Intelligence that cannot be seen, acted on or built upon is not a capability. It is a liability with good marketing. McLean didn't win because he had a better ship. He won because he changed what shipping could become. The organizations that will define the next decade aren't the ones with the most AI. They're the ones who have built the right structure to leverage it.Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?