DXC Technology 與 LangChain 合作，擴大具治理機制的 AI 智能體部署

DXC Technology said on October 13, 2026, that it is partnering with LangChain to advance governed agentic AI across mission-critical enterprise IT environments, with its OASIS orchestration platform already in production and scaling toward 125 customer environments across regulated industries and multiple geographies. LangChain is the company behind the open-source LangGraph agent framework and the LangSmith agent engineering platform.

LangGraph Foundation and New LangSmith License

DXC was already using LangGraph inside OASIS, which the company describes as its intelligent orchestration platform and operating model for mission-critical IT operations. Under the partnership, DXC is also licensing LangSmith to support the development, evaluation, and observability of OASIS agents. LangGraph serves as the open-source orchestration framework, handling delegation, state management, and human-in-the-loop workflows, while LangSmith supplies the agent development, evaluation, and observability tooling. DXC said the combined capabilities let it build, orchestrate, evaluate, and continuously improve AI agents while applying the enterprise context, governance, human oversight, and continuous assurance required in critical technology environments.

DXC first introduced OASIS on April 28, 2026, describing it as an intelligent orchestration platform that pairs human expertise with agentic AI in a new managed services operating model, moving service delivery from reactive support to real-time operations. That launch announcement described OASIS as an open agentic orchestration layer sitting across an organization’s existing tools rather than replacing them, and DXC’s platform documentation names ServiceNow and Dynatrace as examples of systems the platform connects without displacing.

Production Results Reported by DXC

DXC said OASIS applies these capabilities to enterprise IT operations: agents respond to operational events, investigate issues in parallel, pull in specialized capabilities across technology domains, and bring people into the process when judgment or approval is required, leaving human experts to concentrate on higher-value decisions while agents provide capacity and speed at scale.

Across specific production use cases, the company reported, filesystem incidents that previously took 53–90 minutes can be resolved in 3–6 minutes, database performance incidents have moved from 2–3 hours to 5–12 minutes, and complex problem management that once required 2–6 days now takes 8–10 minutes.

“Agentic AI has the potential to fundamentally change how enterprises operate their most critical technology, but models alone are not enough,” said Dan Gray, President of Global Infrastructure Services at DXC Technology. Gray said that across more than 500 measured use cases, OASIS has reduced resolution time by 95% on average, with diagnostic accuracy of 92% or better when validated against senior engineer judgment.

Governance and Human Oversight

A central focus of the partnership is expanding the use of AI agents while preserving governance and human oversight, according to the announcement. Using LangChain technology, OASIS lets agents operate within defined authority while maintaining the controls required for mission-critical environments: actions are assessed on factors including reversibility, impact, verifiability, and an agent’s track record, sensitive actions can be paused for human approval before execution, and agent decisions, tool calls, and outcomes are traced to provide visibility into how agents operate. DXC’s platform documentation states that every action in OASIS is traceable, with role-aware access controls and progressive autonomy that can be expanded or constrained at any level.

DXC said OASIS also grounds agents in the specific enterprise environments where they operate. Agents can draw on system-specific and customer-specific information, DXC’s operational expertise, and vendor documentation, so they reason from the context and requirements of each customer environment rather than relying solely on general model knowledge. Combining that enterprise context with defined governance policies and human oversight lets organizations expand the role of AI agents in production while maintaining accountability and control, the company said.

Julia Schottenstein, LangChain’s chief operating officer, said the enterprises pulling ahead are taking ownership of their agent stack, which she described as adapting the harness to their business, grounding agents in proprietary context, and building the governance and continuous improvement loops that make agents reliable in production.

Open Deployment Model and Technical White Paper

DXC describes OASIS as model agnostic and interoperable with third-party and customer-owned systems, so organizations can select different models for different workloads based on quality, cost, latency, and privacy requirements. The platform is also built to improve continuously as it runs, feeding insights from agent executions back into performance and folding lessons from individual scenarios into the broader system over time.

According to DXC, deployments can span public cloud, private cloud, dedicated, and sovereign environments without any redesign of the underlying architecture, and customers keep control of their data, workflows, and the operational capabilities they build on the platform.

The two companies detail their approach in a new technical white paper, “DXC OASIS: The Harness for Mission-Critical IT Operations,” which examines how organizations can operationalize AI agents while maintaining governance, accountability, and human oversight across mission-critical environments.