In an era where enterprise AI projects are more likely to stall than scale, DXC Technology has quietly been rewiring how a Fortune 500 company experiments, learns and ships AI into production. At the center of that shift is Holly Grant, Senior Vice President of Strategy and Innovation, whose remit spans defining how a global workforce adopts AI, developing new products and platforms, and shaping DXC’s strategic direction alongside the CEO and leadership team. With more than 15 years at the intersection of technology, strategy and growth, she is now applying startup‑grade discipline to one of the world’s most complex IT estates.
LabX, DXC’s AI‑native product incubation lab, is the most visible expression of that ambition. Built on the company’s Xponential AI orchestration blueprint and Human+ philosophy, LabX promises to turn validated business challenges into production‑ready AI solutions in 90 days or less — first stress‑tested on DXC’s own 115,000 employees in 70 countries as “Customer Zero,” then taken to market. The model is deliberately stage‑gated: nothing advances without proof, governance and security are embedded from day one, and every initiative must have a committed sponsor, real production data and measurable outcomes before it enters the lab.
What’s striking in Grant’s approach is that the 90 days are almost incidental. For her, the real innovation is the system that makes that speed reasonable: governance designed for velocity rather than as a brake; a composable architecture that avoids hard‑wiring the business to any single model in a market that shifts monthly; and distributed decision rights that trust small, cross‑functional triads to navigate trade‑offs without constant escalation.
Just as importantly, she is explicit about the traps she wants to avoid: AI “theatre” that boosts individual productivity without moving enterprise‑level outcomes, pilots that never escape their sandbox, and success metrics that over‑index on efficiency instead of the new products, services and capabilities AI can unlock.
ALSO READ: From Siemens Energy to Bank of America: What “Quietly Advanced” Enterprises are Doing Differently
Being Customer Zero, in her telling, is less about testing models and more about exposing solutions to the messy reality of human behaviour, legacy processes and shifting compliance constraints long before they reach a client. Those internal deployments have already reshaped how LabX measures value, how it designs for trust and accountability in AI‑driven workflows, and how quickly teams are willing to shut down ideas that cannot demonstrate enterprise‑scale impact. The result is a dual mandate: build AI products that are viable, valuable and proven under real‑world conditions — and, at the same time, build AI‑fluent leaders and practitioners who can carry that mindset back into the broader organisation.
In this conversation, Grant goes deep on how LabX brings sped-up innovation into an enterprise context, the operating‑model mistakes that keep AI stuck at pilot stage, and why she believes the hardest AI problems are human, not technical. She also unpacks what it really means to be an AI‑native enterprise — from composable architectures and responsible AI guardrails to talent rotations and a Human+ approach that explicitly keeps people in the loop even as intelligent systems take on more of the work.
If you hard-wire yourself to a single model, provider, or framework, you’re betting that today’s leader will still be the leader in 90 days — and that’s a bad bet.
LabX is pitched as an AI-native product incubation lab that can take a concept to a production-ready MVP in 90 days or less. What had to be true — in terms of governance, architecture, and decision rights — before you were comfortable earmarking that 90-day timeframe?
The 90 days isn’t the hard part. Designing the system that makes 90 days reasonable, is.
First, governance designed for speed rather than against it. Compliance, security, and responsible AI sign-off are built into the process from intake — not bolted on at the end. That sounds counterintuitive, but it’s the only way to move fast, safely. The teams that get stuck in long pilot cycles aren’t stuck because they’re careful; they’re stuck because their governance lives downstream of the build. Ours sits alongside it. We also bring cross-functional perspectives in from day one — design, product, and engineering shaping the problem together before anyone writes code. Most of the rework that slows enterprises down comes from getting those perspectives in too late. We get them in first.
Second, a composable architecture. The leading AI tools are changing at least monthly, not yearly. If you hard-wire yourself to a single model, provider, or framework, you’re betting that today’s leader will still be the leader in 90 days — and that’s a bad bet. We’ve watched teams build entire stacks on a single foundation model, then face a major migration when something materially better or materially cheaper ships. A composable foundation means our triads can assemble best-in-class components, swap them out as the frontier moves, and stress-test what’s actually working against real customer problems rather than provider marketing.
Third, distributed decision rights. Every concept that enters LabX has a named, validated customer problem behind it — not a hypothetical one. But the deeper shift is what happens inside the team. Each member of the triad owns the calls in their domain — the engineer on what the architecture can carry, the designer on whether the experience will land, the product lead on whether the outcome moves. They navigate trade-offs together rather than escalating every decision upward. We trust the people closest to the work, and that trust is what lets a small group move in 90 days without losing rigour.
Who else does this help, and how does the value compound? If the only honest answer is “the person who built it,” we’ve got theatre.
Many enterprises are stuck in “AI theatre”: pilots that look impressive but never reach scale. When you designed LabX, what specific anti-patterns were you trying to avoid, and how do you detect early that a use case is veering toward theatre rather than value?
The pattern I see most often is personal productivity without enterprise productivity. A team deploys an AI tool, individuals get faster, dashboards light up, the demo is impressive — and none of it shows up in the business. The intelligence is bolted onto a system or a workflow where it helps one user, but it can’t move horizontally across a team, a function, or a process. That’s when you know you’re in theatre.
AI is a horizontal intelligence. It creates compounding value when it can flow across functions, data, and workflows — and it traps value locally when it can’t. Most failed pilots aren’t failures of technology. They’re failures of the operating model. The tool works exactly as advertised. The enterprise just isn’t designed to let the value scale.
LabX was built to interrupt that pattern early. We designed against it in three ways. Every product starts with a real customer and an outcome we can measure — not an idea looking for a use case. Every product is stress-tested as Customer Zero inside DXC before it goes to a client, so we see whether it actually moves enterprise productivity, not just an individual’s. And every product faces a graduation gate: if it can’t demonstrate enterprise-scale value, we stop. We don’t reward effort; we reward outcomes.
Detecting drift early is mostly about the question we keep asking: who else does this help, and how does the value compound? If the only honest answer is “the person who built it,” we’ve got theatre. That’s our cue to redesign or shut it down.
ALSO READ: Inside the Orchestration Crisis: Why AI‑Driven Enterprises Need a Control Plane, Not More Tools
A distinctive choice is DXC’s willingness to be “Customer Zero,” stress-testing new AI solutions across 115,000 employees in 70 countries before going to market. What are the hardest problems you will only discover by being your own first customer?
The hardest problems are the human ones, not the technical ones. The technical issues you can find in a lab. The ones you only find at scale are the ones tied to trust, behaviour, and how work actually gets done.
When you deploy AI at scale, you discover very quickly that adoption is the real bottleneck — not capability. People don’t resist AI because the technology is bad. They resist it because they don’t know which decisions they’re still accountable for, where the system’s judgement ends and theirs begins, or how the workflow around them is supposed to change. That’s a design problem, and it’s invisible until real people meet real systems in real conditions.
Being Customer Zero gives us that exposure. We’ve learnt where AI creates friction we didn’t expect, where it shifts accountability in ways that need new guardrails, and where the gains we projected on paper compound very differently in practice. By the time a LabX product reaches a customer, it has been pressure-tested against the messiness that defines an enterprise — not the cleanliness of a demo. That’s a different starting position than most providers can offer, and it’s very hard to replicate from the outside.
The hardest problems are the human ones, not the technical ones.
Being Customer Zero also exposes you to the chaos that comes with real-world data, processes, and compliance. How have internal learnings from Customer Zero deployments materially changed the roadmap or design of LabX solutions before they reach customers?
The most material shift has been to how we measure value in the first place.
When we first deployed AI internally, the instinct — ours and the market’s — was to measure the win in efficiency: hours saved, processes compressed. Customer Zero showed us that’s the smaller part of the story. The bigger value, consistently, has been in what AI lets people do that they couldn’t do before — new products, new customer experiences, capacity for work that simply didn’t exist. That’s our Human+ philosophy: people working alongside AI to make their contribution exponential.
ALSO READ: 6 Enterprise Tests to Expose Hidden AI Compliance Risks Across Borders
That reframing has changed how LabX evaluates every product. We still measure efficiency, but we don’t graduate ideas on efficiency alone. We ask whether a product opens something genuinely new — for the people using it and for the customers they serve. If the answer is no, we’re probably building a tool, not a transformation.
The other shift is operational. Customer Zero confronts you with imperfect data, fragmented processes, and compliance requirements that move under your feet. So, we’ve learned to build for adaptability from day one — short MVP cycles, continuous validation, and a willingness to reposition a use case when the evidence says we should. The teams rotating through LabX carry that mindset back into the wider organisation, which means the feedback loop runs both ways: better products for our customers, and a more outcome-focused culture inside DXC. That, ultimately, is what shows up in the work we do for clients.
ALSO READ: Deloitte, Informatica on Why Data Reliability Now Governs Scaling GenAI
