DXC Technology partners with LangChain, the company responsible for the LangGraph open source environment and the LangSmith agent development platform, to enable the deployment of agentic AI in mission-critical IT operations. The agreement aims to boost the use of AI agents with governance and control guarantees.
The alliance builds on DXC’s existing use of LangGraph in DXC OASIS, its intelligent orchestration platform and operating model for mission-critical IT systems. DXC will also acquire a license from LangSmith, which will support the development, evaluation and observability of OASIS agents. This platform already operates in production environments and will expand its reach to 125 customer environments in regulated sectors in various regions.
Incorporation of AI agents in critical environments
Through this collaboration, DXC uses LangGraph and LangSmith to refine the creation, orchestration, evaluation and continuous improvement processes of OASIS agents. LangGraph provides the open source framework for agent orchestration, with capabilities such as delegation, state management, and human-involved workflows, while LangSmith provides tools for agent development, testing, and observability.
OASIS integrates these capabilities into enterprise IT operations so that agents can respond to operational events, investigate problems in parallel, leverage specialized capabilities from various technology domains, and incorporate human intervention when judgment or approval is needed. This allows experts to focus on higher-value decisions, while agents provide the capacity and speed needed to address large-scale operational problems.
Reduction of management times
For specific use cases in production environments, file system-related incidents that previously required 53 to 90 minutes can now be resolved in 3 to 6 minutes. For their part, database performance incidents have gone from requiring between 2 and 3 hours to being resolved in between 5 and 12 minutes. Similarly, the time to handle complex issues has been reduced from 2-6 days to 8-10 minutes.
“Agent-based AI can fundamentally transform the way businesses manage their most critical technology, but models alone are not enough,” said Dan Gray, president of Global Infrastructure Services at DXC Technology. «In more than 500 use cases analyzed, OASIS has reduced resolution time by 95%, achieving a diagnostic accuracy of 92% or higher when compared to the judgment of expert engineers. “Through our collaboration with LangChain, we combine advanced agent technology, operational expertise and governance to deploy AI agents securely and at enterprise scale.”
Scale with governance and human oversight
Using LangChain technology, OASIS now allows agents to operate within defined authority boundaries, with the controls necessary for mission-critical environments. Actions are evaluated based on factors such as reversibility, impact, verifiability, and agent track record. Sensitive actions can be paused until human approval is obtained before execution, while agent decisions, tool calls, and results are recorded to provide visibility into their operation.
OASIS also provides agents with the context of each business environment. Agents can draw on system- and customer-specific information, DXC operational experience, and vendor documentation. This allows them to reason from the context and requirements of each client, rather than relying only on general knowledge of the model. Combining this context with defined governance policies and human oversight allows you to expand the role of agents in production environments without losing responsibility or control.
Organizations can select different models for each workload based on their quality, cost, latency, and privacy requirements.
“Agents are transforming the application layer of every company, and those at the forefront are taking control of their own agent ecosystem,” said Julia Schottenstein, COO at LangChain. «This involves adapting the execution infrastructure to the needs of the business, providing the agents in their own context and establishing governance mechanisms and continuous improvement cycles that guarantee the reliability of the agents in production environments. “We are proud to bring this model to critical business operations through our partnership with DXC.”
An open approach to AI agents in the enterprise
The collaboration also promotes an open approach that offers companies flexibility and control as AI technologies evolve. OASIS does not depend on a specific model and allows interoperability with third-party systems and those of the clients themselves. Organizations can select different models for each workload based on their quality, cost, latency, and privacy requirements. Additionally, OASIS is designed to continually improve during operation. To do this, it uses information from the execution of the agents to optimize performance and incorporate what has been learned in specific cases into the system as a whole, over time.
Deployments can span public and private clouds, as well as dedicated and sovereign infrastructures, without the need to redesign the underlying architecture. Customers retain control of their data, workflows and operational capabilities built on the platform, as well as the flexibility to choose models, infrastructure and deployment environments.
