What do you learn when you place an agent in charge of HR for 30,000 people? How is the concept of minimum viable agent changing the way agentic AI gets deployed? How do you control and measure performance in agents - and clamp down on proliferation? Hitachi Digital Services CTO Premkumar Balasubramanian opens up to HFS Research Executive Research Leader David Cushman.

HDS is learning from its internal deployments of agentic AI at scale. Grounding on an internal knowledge base, applying rigorous evaluations, and keeping humans in the loop are all mandatory.

Among the key discussion points:

• The need for open telemetry and a single-pane view to control proliferation and control costs.
• The risk of a new tech debt delivered through agent sprawl.
• Coming soon: HARK Agents focused on reliability, responsibility, observability, and optimality - to jumpstart the enterprise journey.

Embrace agentic AI but do so with strong governance, measurement, and observability to maintain control at scale.