From APIs to AI Agents: Building Enterprise Agentic Systems at Exasol Chennai

On Saturday, September 5, 2026, I had a wonderful experience being part of Exasol Chennai Chapter #2 - "From Data to Agents", a community meetup focused on the practical shift from data platforms and APIs toward intelligent, agentic systems.

I had the opportunity to speak on "From APIs to AI Agents: Building Agentic Systems for the Enterprise". The session brought together perspectives on how LLMs, APIs, MCP, tools, and enterprise data can work together to move applications beyond traditional workflows and into systems that can reason, plan, call tools, and support real business outcomes.

From Integration To Orchestration

One of the central ideas I explored was the shift from simple integration to orchestration. For many years, enterprise software has been built around deterministic workflows: a user action triggers an API call, the system applies fixed rules, and the result moves through a known path.

Agentic systems change that pattern. APIs still matter deeply, but they become part of a broader tool layer. The agent interprets intent, plans a path, selects the right tools, reasons over available context, and works through a task with guardrails, observability, and human accountability.

That shift is especially important in the enterprise, where AI cannot be treated as a standalone chatbot. It has to connect with existing systems, respect approvals, handle data responsibly, and operate within clear business constraints.

The Enterprise Agent Loop

In the session, I broke down the agent loop as a practical architecture: understand the user or business goal, plan the next step, call the required tool, observe the result, and continue until the task reaches a useful outcome. This loop is simple to describe, but it becomes powerful when connected to enterprise systems in a controlled way.

The important point is that reasoning alone is not enough. An enterprise agent needs reliable access to tools and data. MCP and similar approaches are useful because they create a cleaner contract between the model and the systems it needs to interact with. APIs expose capability; tools make those capabilities usable by the agent; enterprise controls decide what is allowed, logged, reviewed, and trusted.

Control, Context, And Real-World Adoption

Enterprise AI adoption depends on more than model capability. It depends on context quality, permission boundaries, auditability, accuracy, cost awareness, and a clear understanding of where human review is required. The strongest agentic systems will not be the ones that try to automate everything blindly. They will be the ones designed with sensible control points.

For me, this is where digital architecture becomes important. Agentic AI should sit inside an architecture that is measurable, governable, and understandable. The real opportunity is not only to build impressive demos, but to build systems that can safely assist teams in sales, support, operations, analytics, engineering, finance, and other enterprise workflows.

Community Conversations

What made the event especially valuable was the opportunity to connect with fellow technology professionals and exchange ideas around the practical adoption of Agentic AI. The conversations, questions, and energy from the Chennai tech community made the meetup feel thoughtful and engaging.

Events like this are a reminder that technology becomes more meaningful when people discuss it openly: what is possible, what is risky, what is useful today, and what still needs to mature. Agentic AI is moving fast, and the best learning happens when builders, architects, data professionals, and business teams come together to compare real experiences.

Gratitude

A big thank you to the Exasol team for inviting me and organizing such an engaging community meetup. It was great sharing the platform with Felix Chandran and interacting with the speakers, organizers, and everyone who participated.

A special thank you to Ranjan Moses for the invitation, Felix Chandran for the support throughout, Keerthika Sengottian for the encouragement and motivation, and Premkumar N for being part of the experience.

Thank you, Exasol, for the opportunity and the warm hospitality. I truly enjoyed the conversations, questions, and energy from the Chennai tech community, and I look forward to continuing to learn, build, share, and contribute as AI agents increasingly become part of real-world enterprise systems.