How IT leaders are hedging against tech debt and 'agent sprawl' to get the most from AI

The hype around agentic AI often feels disconnected from the day-to-day reality of enterprise IT, where legacy systems and rigid processes move at a glacial pace. But every so often, a single, concrete moment cuts through the noise, making the abstract future feel shockingly present. For one seasoned enterprise leader, that moment arrived not in a keynote speech, but during a hands-on experiment that collapsed a months-long development cycle into less time than it takes to make coffee.

Amitabh Ranjan Sharan is the IT Solutions Consulting Director at DHL IT Services with over 18 years of experience leading global engineering, DevOps, and IT operations. Sharan operates at the complex intersection of strategy and execution. A recent breakthrough didn't just change his perspective; it revealed both the revolutionary potential of AI-driven development and the profound new risks that come with it.

• Ready, set… Go: "I once asked the AI to use Go for the backend and React for the front end, and the whole thing was done with a GitHub code check-in and documentation in less than three minutes. The pace really mesmerized me, and that became the gateway to unlocking more use cases."

• Blurred lines: This new reality, Sharan explains, is beginning to fundamentally reshape the enterprise, altering traditional roles and workflows. "It's blurring the line between development, operations, and the product team," he says. "Imagine that somebody who has business knowledge can just put in their requirement and things will be done."

This incredible power, however, immediately introduces a new and dangerous conflict. The primary risk isn't failure, but a chaotic, uncontrolled success. It's a phenomenon Sharan calls "agent sprawl". This new form of technical debt doesn't just accumulate passively, but actively burns cash in real-time if not managed properly.

• Managing agent sprawl: "With the speed of AI, we now have a risk of having agent work sprawling," Sharan warned. "If you have one problem and can solve it in a hundred different ways, how do you decide which is the correct and optimal way to solve it?" This isn't just a technical problem; it's a direct threat to the bottom line. "These are all very high CPU- and GPU-consuming processes. If you do not use them wisely, you may spend more money than you were spending before. Your budgets will be damaged."

The answer, according to Sharan, isn't to slow down but to build smarter oversight. He argues for the urgent development of two missing architectural layers: "AI Observability" to see how agents are performing and an "AI Registry" to manage the governance, intelligence, and effort of every agentic system. While he admits a security gap currently exists, he is confident that AI itself is best suited to solve it, citing how structured security knowledge can be fed into systems using tools like Open Policy Agent to enforce compliance.

• An eye for governance: With proper governance, the ultimate payoff is the creation of "self-healing, self-fixing environments" that fundamentally raise the bar for quality across the entire industry. "This will definitely raise the bar," Sharan explains, drawing two powerful parallels. "Imagine we never had touch screen phones, and when the touch screen phone came, it became the new normal. The same thing happened in IT with containerized applications—suddenly the '4 Cs of security' promoted by the CNCF raised the bar for everyone."

This future inevitably forces the human question: is this the end for developers? Sharan’s answer is a firm "no," but it comes with a critical caveat. This power must be grounded in architecture and process, or organizations risk creating not quality output, but trash work at an unprecedented scale.

• Means to an end: "This is a new beginning, not an end. Imagine that one developer can now do the work of 20 developers. If each development organization is working with this multi-fold power, you will be able to create more outcomes which were not even thought possible."

This new paradigm presents every business leader with a fundamental choice. One path is to see the productivity gains as a simple opportunity to cut costs and "axe the workforce." But the other, Sharan argues, is the path of a wise leader who sees a far greater opportunity. "The companies which are wise will definitely continue to invest in their human capital and leverage their strength. It's a multi-fold calculation to improve the quality, reduce the bugs in the system, and increase the resiliency."