- DXC Technology, a 115,000-employee global IT services company, served as “customer zero” for Amazon Quick, AWS’s agentic AI platform, deploying it internally before taking it to clients.
- DXC Global Chief Digital and Information Officer Russell Jukes led the initiative, describing the relationship with AWS as true co-innovation rather than a traditional vendor engagement.
- DXC built a two-tier agent model: personal agents for individual productivity, and professional agents built by IT for department-specific workflows with tailored security and data access.
- Employee feedback from DXC’s rollout directly shaped Amazon Quick’s product roadmap, including tiered governance controls for sharing agents across individuals, work groups, and the enterprise.
- Jukes distinguishes “adoption” (deployment metrics) from “AI fluency” (individual learning journeys), arguing enterprises must manage the latter, not just measure the former.
- DXC built an internal AI Playground using synthetic data for risk-free experimentation, an AI Advisor to field employee questions, and a centralized, tool-agnostic agent registry to track agent lifecycle.
- DXC’s “AI Standard of Business Conduct,” governing how AI agents should interact with employees, was written by DXC’s HR and people-and-culture team, not IT.
- Jukes’s advice to fellow CIOs: don’t wait for a pilot, and don’t equate deploying more agents with delivering more value; nobody is “late” to enterprise AI.
When DXC Technology decided to become “customer zero” for Amazon Quick, AWS’s agentic AI platform, the company wasn’t just running a pilot, it was rebuilding how 115,000 employees interact with data, decisions, and one another. On a recent episode of our Age of AI podcast, DXC Global Chief Digital and Information Officer Russell Jukes walked through what that journey has actually looked like, including challenges experienced along the way, and the lessons are relevant well beyond DXC’s walls.
From Vendor to Co-Innovator
As the company began thinking about its enterprise AI journey, Jukes shared that DXC evaluated multiple AI platforms before committing meaningfully to Amazon Quick. Jukes is careful to frame the relationship as something different from a typical vendor engagement. He describes a progression from buyer-supplier to partner to co-innovator, the point at which two companies effectively function as one team. Jukes shared that DXC’s feedback loop with AWS isn’t theoretical: real deployment challenges inside DXC directly shaped product capability. When DXC realized that letting 115,000 employees create and share agents without controls would produce an unmanageable “agent swarm,” the two companies put their collective heads together and built tiered governance: individual, work group, and enterprise-level sharing permissions, directly into the product.
A Deliberate Split: Two Kinds of Agents
Central to DXC’s approach is a deliberate split between personal agents, which employees build for their own productivity, and professional agents, which Jukes’s team builds to run defined workflows with appropriate security and data access. That split emerged from a debate inside DXC over whether the sheer number of deployed agents correlated with business value. The team concluded that it didn’t, and the personal/professional model was the result.
Why Adoption Isn’t Fluency
Perhaps the most useful distinction Jukes offers is between adoption and fluency. Adoption, in his telling, is the traditional deployment scorecard: you’re looking at whether people logged in, and whether they use the tool. Fluency, however, is an entirely different thing, and so incredibly much more relevant. Fluency is the individual learning journey every employee has to go through, and at their own pace. AI fluency requires sustained change management rather than a one-time rollout, and this is often a major stumbling block for companies embarking on their own enterprise AI journeys. Jukes shared that DXC’s early attempts to enforce adoption by notifying managers of low usage failed outright and were quickly abandoned in favor of a strategy that focused deeply on employees’ individual AI fluency journeys.
How They Did It: They Built an AI Playground and Invited Everyone In
To drive fluency without risk, DXC built what they call an “AI Playground” stocked with synthetic data that mirrors real DXC data without exposing anything sensitive. It was a safe place where employees could hang out, experiment freely, form teams, compete in challenges, and see winning capabilities pushed into the live environment. Jukes also personally emailed rollout announcements to the entire company from his own inbox, a move that generated roughly 20,000 replies but built visible trust. A separate, tool-agnostic agent registry now tracks every agent’s lifecycle, from creation through retirement.
Earning Trust at Scale
Jukes and I talked a lot about trust in our conversation exploring DXC’s entperprise AI journey. Success on the AI front absolutely cannot happen without employees trusting the process, the goals, and the leaders working to roll out AI initiatives. Jukes is fairly passion about the concept of trust and he shared that trust runs through nearly every part of DXC’s approach. Notably, the company’s “AI Standard of Business Conduct,” which governs how AI agents should interact with employees, was written by DXC’s HR and people-and-culture team, not by IT. To me, this speaks volumes about the company’s commitment to building trust in the team doing the rollout, as well as trust in the goals of corporate AI initiative. Jukes also described how internal champions emerged organically rather than being appointed, which was a departure from DXC’s earlier top-down model, and also played a significant role in the company’s overall success.
Governance Without Gridlock
Another smart move by Jukes and team was that rather than connecting AI directly to every data source, DXC routes access through a curated Universal Data Platform. They have moved from role-based personas (sales, HR, finance) to outcome-based groupings that reflect what employees are actually trying to accomplish, which overall, leads to better, quicker results.
Advice to Fellow CIOs: Don’t Wait
Asked whether other CIOs should wait before moving on platforms like Amazon Quick, Jukes was direct: no one is “late” to enterprise AI, organizations simply join the cycle at different points. Jukes’ guidance is to start now, at enterprise scale rather than in isolated pilots, while building the resiliency, continuity, and audit practices that production-grade AI systems require, something he admits DXC didn’t fully plan for at the outset.
For CIOs weighing their own AI journey, DXC’s experience offers a clear signal: the technology decision is often the easy part. Finding the right vendor partner can play an outsized role in overall success, and for DXC, the partnership with AWS and the co-innovation work they are doing together has been a game-changer. Beyond finding the right vendor partner, the harder, more consequential work is building the trust, governance, and fluency that determine whether AI actually changes how an organization works.
Watch the full episode of our conversation here:
Read more of my coverage:
Connectivity, Context, Control: Why the Data Layer—Not the Model—Decides Whether Enterprise AI Works
This article was originally published on LinkedIn.