DXC and LangChain Integrate Agentic AI into Critical IT Operations

The collaboration embeds LangChain’s orchestration framework directly into DXC OASIS, a platform currently scaling toward 125 customer environments. By utilizing LangGraph for state management and LangSmith for observability, DXC enables its agents to handle complex diagnostic tasks that previously stalled human engineering teams. For instance, filesystem incidents that once consumed up to 90 minutes are now resolved in as little as three minutes, while multi-day problem management cycles have compressed into ten-minute windows.

Central to this deployment is a governance-first architecture. Rather than allowing autonomous agents full autonomy, OASIS subjects AI-driven actions to strict verification protocols, including impact assessment and reversibility checks. When an action involves sensitive infrastructure, the system enforces a human-in-the-loop requirement, pausing execution until a qualified engineer provides oversight. This "harness" approach ensures that agents remain grounded in specific enterprise contexts, vendor documentation, and proprietary operational data rather than relying on generalized model knowledge.

Dan Gray, President of Global Infrastructure Services at DXC, noted that the platform maintains a diagnostic accuracy rate of 92% or better when measured against the judgment of senior engineers. Because the OASIS architecture is model-agnostic, enterprises retain the flexibility to swap underlying AI models based on shifting requirements for latency, privacy, or cost. This interoperability extends across public, private, and sovereign cloud environments, allowing companies to modernize their technology stacks without being locked into a singular vendor’s infrastructure.