DXC partners with LangChain to bring governed AI agents to critical IT operations

DXC Technology is pairing AI agents that investigate IT problems with controls over what they can do next. Its October 13, 2026 partnership with LangChain builds on an existing production platform, OASIS, using LangGraph to coordinate agents and licensing LangSmith to develop, evaluate and track their work in mission-critical enterprise IT systems.

The agreement is not the launch of OASIS. DXC says the platform is already operating in production and scaling toward 125 customer environments across regulated industries and multiple geographies. The partnership builds on its use of LangGraph, an open-source framework that coordinates agent work. LangSmith supplies the development, evaluation and observability tools—the means to inspect what agents did and how their work turned out.

LangGraph handles delegation, keeps track of a workflow’s state, and supports steps that involve a person. Within OASIS, DXC says agents can respond to operational events, investigate problems in parallel and draw on specialized capabilities across technology domains.

The agents also receive information about the systems they are working on. DXC says OASIS draws on customer-specific and system-specific information, its operational expertise and vendor documentation, rather than relying only on a model’s general knowledge. OASIS is designed to use lessons from agent executions to improve subsequent work.

DXC says OASIS assesses proposed actions using four factors: whether they can be reversed, their impact, whether their results can be verified, and the agent’s track record. Sensitive actions can be paused for human approval before execution. Permission depends on the action and the agent proposing it.

Agent decisions, tool calls and outcomes are traced so operators can inspect how the system acted. DXC presents these records as another part of the controls needed to operate agents within defined authority in critical technology environments.

Those controls sit within a platform designed to work with different AI models and third-party or customer-owned systems. DXC says customers can select models by workload, weighing quality, cost, latency and privacy. Deployments can span public and private clouds, dedicated infrastructure and sovereign environments without redesigning the underlying architecture. Customers retain control of their data, workflows and the operational capabilities built on the platform.

DXC’s headline performance claim is an average 95% reduction in resolution time across more than 500 measured use cases. It also reports diagnostic accuracy of 92% or better, validated against senior engineer judgment. These are company-reported results: one measures speed, while the other measures agreement with an experienced engineer’s assessment.