CIOs Build The Orchestration Layer That Keeps Composable Architecture From Sprawling

Composable architecture promises enterprises the freedom to assemble systems on their own terms. Modular platforms, headless CRM layers, and agentic AI are accelerating that shift. But for CIOs operating under margin pressure with legacy infrastructure still running core operations, the opportunity is not modularity for its own sake. It is designing the orchestration, governance, and internal capability needed to make flexible systems produce measurable outcomes instead of unmanaged complexity.

Sandeep Bansal is CIO at A-One Steels India Limited, where he leads technology for a major steel manufacturing enterprise. Since joining, he has reduced security incidents by 90% and achieved 100% technical audit clearance for regulatory filings. Before A-One Steels, Bansal led digital transformation at Manipal Academy of Higher Education, where he collapsed 180 systems into 15 enterprise platforms, cut enterprise OPEX by 40%, and accelerated a $10M AI-first transformation portfolio. That operating record shapes how Bansal sees the next wave of enterprise architecture: flexibility only matters if the connective tissue is built to hold.

"When you are creating an AI-centric enterprise architecture, it has to be API-first. And when you are talking about orchestration, that API layer needs to be secure as well." Bansal described composable platforms as "Lego boards" that let enterprises build structures however they want. That modularity is attractive for CRM, pipeline automation, and AI agent deployment. But without orchestration discipline, the same flexibility produces fragmentation.

• The Frankenstein risk: Bansal cautioned that composable stacks without integration governance quickly become harder to manage than the rigid systems they replace. "If you don't design for orchestration from day one, you don't get transformation. You just end up with more tools and more complexity," he said. At A-One Steels, where steel pricing is volatile and margins are thin, every solution faces CFO scrutiny. There is no room for platforms that cannot prove value quickly.

• Milestone-based proof: Bansal treats transformation as a sequence of earned decisions. In margin-sensitive environments, he starts with a small investment, attaches it to measurable outcomes, and scales only after the business can see the value. “Within the first month, leaders can see the numbers. By the third month, the data we bring to the table has already started to build trust,” he said. In a previous role, that model helped deliver 30% efficiency improvement and 70% improvement in collections in a sector where CRM adoption had never been attempted.

The conversation turned to what sits between modular applications and enterprise value. For Bansal, that layer is orchestration, and it has to be secure by design. A composable stack only works when the business understands how systems exchange data, which tools have permission to act, and where accountability sits when workflows cross platforms. That makes the integration layer more than a technical connector. It becomes the control plane for AI adoption, CRM modernization, and enterprise resilience.

• Legacy meets agents: Bansal sees the first governance challenge in the architecture many enterprises already have. AI needs data, but that data often sits across legacy systems never designed to feed modern tools. “You have to get back to the basics: what data is sitting where, and where does it need to go?” he said. Agentic AI raises the risk because once systems are not only reading data but acting on it, CIOs need clear visibility into where agents sit, what they can access, and who owns the outcome. “When an agent is added to an application, leaders need to know where it sits, what role it plays, and who is accountable for what it does,” he said. “Without governance and a human in the loop, it becomes dangerous. Ownership starts to disappear.”

The deeper structural argument Bansal made was about control shifting from vendors to internal teams. Many organizations are stuck in vendor lock-in where legacy solutions run critical operations and nobody wants to touch them. Composable architecture offers a way out, but only if the organization builds internal capability alongside it.

• Capability before scale: Bansal’s answer to vendor lock-in is to build more capability inside the organization before the next wave of tools arrives. “If you’re bringing in these solutions, you have to develop the internal capability to own them,” he said. He sees that training as both operational insurance and a retention strategy: employees who learn new platforms have a reason to grow with the business, while those who avoid AI become more exposed as the work changes. That is why he warns against agentic AI adoption driven by pressure or sales momentum. “I’ve seen organizations adopt agentic AI when their internal architecture was nowhere near ready,” he said. “That becomes a FOMO purchase, not a transformation strategy.” His advice to CIOs is to plan five years ahead, design for resilience, and stop rebuilding the enterprise architecture every three years.

For CIOs under pressure to adopt composable and AI-enabled architectures, Bansal’s final point is that the mandate has already outgrown the title. CIOs are being asked to think like CISOs, CAIOs, and CDOs at once, often while legacy systems, security risk, business scrutiny, and AI urgency collide in the same roadmap. The leaders who manage that expansion will be the ones who treat architecture as a long game, not a sequence of reactive technology decisions. “A CIO needs to play like a grandmaster and think five moves ahead of the board,” Bansal said. “Otherwise, pressure comes from every side, and it becomes impossible to manage.”