Agent Democratization: The Next Leap After Data Democratization and API-First Thinking
How Agentic AI Is Reshaping Enterprise Architecture
All the business leaders, sales heads, and technology executives I’ve met recently are exploring ways to leverage the cutting-edge tech in the AI space to transform customer experience, increase operational efficiency, increase revenue, and reduce risk.
The IT team of every line of business is trying to identify use cases that can truly benefit from AI Agents and Agentic AI. But how would their agents access the functionality and data of their core backend systems? For example, in the insurance industry, core systems for policy admin systems (PAS), claims, billing, and underwriting.
Today’s applications, soon to be legacy😉, achieve that through APIs. The future applications will be Agentic, i.e., powered by agents that can figure out what to do in which scenario, when to do it, and in which order without being concerned about the how.
Today’s core systems publish an API catalog that LoB IT teams use to identify the APIs that’ll meet their applications’ needs.
I have seen several PoCs of agentic applications where the teams have directly coded API calls in the agent, which makes them tightly coupled with the APIs and fixed. Agentic applications are supposed to be autonomous, adaptive (able to pick up new capabilities as they become available without making changes in the code), and capable of dynamically identifying which agents to use for which goals.
Now, how do we enable that? This is where an agent registry comes into play. Agents are registered here, and the LLM in the agentic application autonomously determines the user’s goal and queries this registry to find agent(s) that can achieve that goal. The application then invokes the identified agent.
It is now the responsibility of enterprises to create agents that make their business capabilities available, like policy issuance and claim adjudication, to the external world.
Eventually, we’ll have an agent marketplace where enterprises can leverage each other’s capabilities to develop innovative solutions that help businesses achieve their goals.
Recent technological developments, like MCP and A2A that standardize agent integration and collaboration, and agent frameworks like LangGraph, CrewAI, and AutoGen have established a strong foundation to make this a reality.
As data democratization made data accessible across the enterprise, agent democratization will make enterprise capabilities discoverable and actionable through agents.