Enterprise AI is entering a more demanding phase. Models and compute are advancing quickly, but business value depends on a broader equation: the data that feeds AI, the processes it changes, the infrastructure it runs on and the trust required to put it into production.

That tension framed a fireside conversation at AI Infra Summit 2026 between C Vijayakumar, CEO & Managing Director, HCLTech, and John Roese, Global Chief Technology Officer and Chief AI Officer, Dell Technologies. Their discussion focused on a central leadership challenge: how to move faster without losing control of economics, data, governance or sovereignty.

Hybrid AI is an economic and sovereignty decision

Infrastructure becomes part of the business equation once AI moves into production at scale.

Different workloads carry different requirements. Some are high-volume and cost-sensitive. Others justify more expensive models because the decisions they support carry greater strategic value. Sensitive data, regulatory obligations and latency requirements can also change where a workload should run.

“Eventually, the world is going to be hybrid,” said Vijayakumar.

Enterprise architectures will combine private AI infrastructure with access to frontier models and external services where they make sense.

Roese made the same point from an infrastructure perspective. Enterprises need enough choice to align each workload with the right economics, performance and control. “Software does not run on magic; it runs on infrastructure,” he said.

This is where sovereignty becomes practical. Enterprises need to decide where data, prompts, models and inference should reside, how much control they need and which workloads can safely use external services.

HCLTech AI Factory is designed around that hybrid reality, supporting AI across on-premises, cloud, edge and sovereign environments. Our offering brings together infrastructure, data, platforms and AI models in a governed environment, with capabilities including FinOps for AI to optimize GPU utilization and control costs.

Enterprise AI starts with the process

Roese drew a distinction between building frontier models and applying AI inside an enterprise. Enterprise AI must work across existing systems, data, people and business processes, making production deployment a broader transformation challenge.

“The journey is much longer than just simply building a model or standing up an AI technology or an infrastructure,” said Roese. He described Dell’s own experience as involving governance, skills, data architecture and process change as much as technology. “You cannot implement AI in the enterprise unless you know what process you’re trying to change, and it turns out processes are not necessarily documented.”

Vijayakumar described a similar pattern across HCLTech’s work with global enterprises. Data can remain locked in legacy systems or lack the context needed to support reliable AI. Existing processes can also contain layers of approval and handoffs designed for a different operating model.

“Trying to do Agentic AI adoption in an enterprise using the same processes is not going to yield the right outcomes,” he said.

Organizations need to decide which processes should be redesigned, what decisions can be automated and where human judgment remains essential.

Vijayakumar summarized the challenge in three words: “Data, process and trust. And, a very important theme is ROI on AI.”

Our research illustrates the scale of that challenge. The Value Edge: Powering the Next Era of TMT, research we carried out with Economist Enterprise, found that 91% of 202 C-suite executives surveyed across the US and Europe are satisfied with their AI investments, yet only one-third have a framework to measure AI's commercial impact. Integration challenges and technical debt are actively stalling AI progress for 62%, while 76% say unresolved Responsible AI questions and compliance gaps are actively stalling their ability to scale.

That emphasis also connects with our broader AI direction. In a recent Harvard Business Review article, Vijayakumar described how advances in AI are altering the economics of IT services, breaking the traditional link between revenue growth and workforce expansion.

Standardization creates velocity

Greater velocity comes from reducing the engineering friction between an idea and a production outcome while maintaining the controls required for enterprise deployment.

That pressure is reflected in our The AI Impact Imperatives, 2026 research. Respondents expect, on average, 43% of major AI projects initiated over the next 24 months to fail, while the median expected payback period for major AI investments is roughly 18 months.

Roese said the AI journey has accelerated as platforms, protocols and interfaces have matured. A year ago moving an AI idea into production took several months. Today, he said the cycle can be less than four weeks. He attributed that improvement to a technology stack becoming more standardized and repeatable.

That is also the logic behind our AI Factory model: validated reference architectures, standardized design patterns and industrialized operations are intended to make AI deployment more repeatable and scalable.

The broader demand for Advanced AI capabilities is growing. HCLTech reported Advanced AI revenue of $171 million in Q1, FY27, up 62.1% year over year in constant currency. We have also announced plans to invest up to ₹3,500 crore in AI data centers, complementing its existing AI infrastructure and operations capabilities.

Turning velocity into enterprise advantage

The next stage of enterprise AI will be defined by how effectively organizations orchestrate the system around them.

That means modernizing the right data, redesigning processes, matching workloads to the right infrastructure and embedding governance into the architecture. Speed and sovereignty do not have to be opposing goals. The right hybrid architecture can give enterprises greater flexibility and greater control.

The enterprises that move fastest will not necessarily be those deploying the most AI. They will be those that reduce the friction between investment and measurable value, and between an idea and a trusted production outcome.

That is the AI velocity equation: speed created by better foundations, sovereignty built into architecture and enterprise advantage measured in business value.