How DXC Technology used Anthropic’s Claude to build DXC OASIS, an agentic operations platform

Fourteen months later, in April 2026, DXC OASIS went live. Already deployed to more than 60 customers, the AI-native orchestration platform for managed services represents a fundamentally different way to run enterprise IT operations: AI agents handle routine and repetitive tasks, freeing people to focus on work that helps move the business forward.

Building DXC OASIS with Claude became the foundation of a formalized partnership agreement between DXC and Anthropic. DXC is now one of a small number of professional service firms to join the Claude Partner Network as a Global Premier partner — a relationship built on DXC's open model architecture. DXC is also a member of Project Glasswing, Anthropic's cybersecurity initiative to secure the world's most critical software, which DXC applies to ensure the safety and security of products like DXC OASIS.

DXC OASIS is also validated by Anthropic's Applied AI team, reflecting how deeply the platform has been tested and vetted by the organization. What began with DXC engineers using Claude Code to build DXC OASIS has since expanded into a joint effort to train tens of thousands of Claude-certified engineers and specialists — a dedicated, engineer workforce equipped to bring Claude into production across the mission-critical systems DXC operates for the world's largest banks, airlines, insurers, manufacturers and government agencies.

Before Claude entered the picture, the DXC engineering team had already begun using AI to accelerate development, exploring how large language models could speed up software delivery across design, build, testing and product management. By the time prototyping began, engineers had already logged meaningful productivity gains with AI coding tools, in the range of 20% to 30%.

Those gains shifted significantly when DXC's engineering team stood up Dev X Labs, an internal initiative to evaluate the next generation of AI development tools. In August 2025, the team evaluated a combination of Opus and Sonnet, and within a week every engineer on the team that builds DXC OASIS had made it their default. When the evaluation period ended, they made it clear it was too valuable of a tool to give up.

"I've been in this industry long enough to recognize when something genuinely changes the way work gets done, and building DXC OASIS was that moment for our team," said Dan Gray, DXC President of Global Infrastructure Services. "Claude was embedded in how we thought, designed and built. We moved faster than we have on any platform of this complexity, and the way we worked to create it is as much a proof point as the platform itself."

That response marked a turning point. What had been incremental productivity gains became a fundamental shift in how the team worked. Claude’s Opus and Sonnet models didn't just help engineers to write code faster. They became a thinking partner for architecture decisions, a tool for diagnosing and fixing problems quickly, and a faster path from design to working code.

Here’s a closer look at three ways Claude helped DXC's engineering teams build DXC OASIS.

Building a sophisticated enterprise platform means rigorous testing, as well as fast and precise debugging when the technology doesn't behave as expected.

Before launch, the team subjected DXC OASIS to the same rigorous security and reliability testing expected of any production software. But Claude gave engineers an edge: it helped them generate and pressure-test a vast range of scenarios that DXC OASIS's AI agents would encounter, including edge cases like slow responses and agent handoffs, which the team then validated against DXC's security and reliability standards.

Once development was underway, Claude also helped engineers quickly find and remove bugs

"This was transformative," said Nick Tamburro, who heads DXC’s Research & Development team. "A platform as layered as DXC OASIS has dozens of interconnected systems, and our engineers are rigorous about tracing every issue to its root cause.”

Claude made that process dramatically faster, Nick added. An engineer would describe a failure, and within minutes get back a structured list of likely causes ranked by probability.

“One integration bug that had stumped the team for a full afternoon was diagnosed in minutes,” Nick said.

DXC OASIS is built on a unified data foundation that gives every AI agent the context it needs to act and delegate tasks autonomously. An agent managing enterprise IT infrastructure knows exactly when to call in a server administration agent, tap an IT Service Management (ITSM) agent for incident details or query an observability agent for telemetry data.

Building the coordination layer that makes this work was a core challenge. The team had to determine which agent handles which task, how agents hand off work to one another and how to make the entire experience seamless for the user. All of this had to work at production scale, with no established playbook to follow.

That required reasoning from first principles. Claude became the team's thinking partner for navigating complex architectural questions with no single right answer: How should a supervisor agent decide which sub-agent gets a task? What happens when a task is interrupted mid-execution? Where does the supervisor's control end and the sub-agent's autonomy begin?

Decisions that would have taken two to three days were resolved in one to two hours. The team would pose a question and get back a structured breakdown outlining possible approaches, the trade-offs of each, what breaks under load and what introduces security risks. They would then choose the path that best fits their requirements.

"Claude didn't decide how agents should behave," said Dan, who has spent 19 years at DXC, leading technology strategy and delivery for some of the firm’s largest customers. "The engineering team did, with Claude compressing the research-and-debate phase from days into hours."

Building DXC OASIS's chat interface presented a deceptively complex design challenge: create a screen intuitive enough for any user — much like a consumer app — while meeting strict enterprise accessibility standards.

The team used Claude to accelerate the entire design process. Rather than having a designer sketch concepts and walk stakeholders through them in a meeting, the team described what they wanted and Claude assembled realistic mockups using DXC OASIS's actual design components. Stakeholders could review something close to the finished product on the same day, rather than interpreting rough sketches and imagining the final result.

Claude also enforced consistency across the platform. By loading DXC OASIS's design guidelines upfront, Claude produced the correct colors, fonts and button styles from the first attempt, every time a new feature needed mocking up.

The biggest gain came from closing the gap between design and code. Once a design was approved, the team passed it directly to Claude Code, eliminating the back-and-forth between designer and developer where errors typically sneak in.

"What used to take one to three weeks became a same-day deliverable," said Lisa Beaudoin, Chief Product Officer at DXC Technology. "Individual components that took one to two hours now took 15 to 20 minutes. When you multiply that across hundreds of interface elements, the compounding effect was enormous."

Across the full build, DXC estimates that Claude made OASIS software delivery ten times faster, with more than 95% of code generated by Claude before human review.

The acceleration the team experienced now extends to customers searching for better ways to manage their infrastructure, cybersecurity and workplace needs. To bring that capability to enterprise customers, DXC is building a dedicated workforce of Claude-certified, forward-deployed engineers, recruited from its existing talent and trained through Anthropic's Partner Academy. Initial deployments will focus on insurance, cybersecurity, and application services, where DXC's deep domain expertise and Claude's agentic capabilities converge to empower mission-critical systems.

"In managed services, we solve one-in-a-million problems dozens of times a day," said Dan, who co-leads the build of DXC OASIS alongside Lisa. "What's changed is the complexity. The growing sophistication of AI means more systems, more interdependencies more signals than any team can process manually."

The old operating model for managed services doesn't scale to meet these evolving demands, Dan added.

DXC OASIS is built to address the gaps: an orchestrated platform where agentic AI and human expertise operate together, so people aren't buried in signal noise and can focus on areas where their judgment matters most.

Dan points to early results. Across 56 IT use cases, DXC OASIS hits 95% first-time response accuracy and cuts average resolution time by ~95%. Problems that consumed two to six days are now closed in eight to ten minutes.

"We're solving the same problems under a radically different operating model," Dan said. "If you scale that across an enterprise IT estate, managed services start to look fundamentally different."

• DXC OASIS
• DXC Engineering
• Q&A: DXC OASIS is bringing transparency to enterprise IT