For much of the last decade, enterprise cloud adoption was driven by urgency. Organisations migrated workloads, modernised applications, and measured progress by how quickly they could move away from legacy infrastructure. That phase delivered momentum, but it also created a misleading sense of completion. As cloud adoption matures, enterprises are discovering that transformation does not end with migration—it begins there.
According to Uwem Ukpong, Vice President of Global Services at Amazon Web Services (AWS), this shift is already reshaping how enterprises evaluate cloud platforms, as adoption gives way to expectations around operational performance and reliability.
“Only 20% of the world’s enterprise workloads are on the cloud,” Ukpong says. The figure is often cited as untapped opportunity, but it also reveals where complexity is concentrated. The workloads still off the cloud tend to be regulated, operationally critical, and tightly embedded in business processes. Moving them requires architectural intent, governance, and organisational readiness—not just infrastructure.
What has changed, Ukpong argues, is not just the scale of cloud adoption, but the way enterprises now frame their cloud strategies. As organisations move beyond early migration phases, cloud decisions are increasingly shaped by industry operating models, regulatory exposure, and the realities of running core systems at scale. This has shifted the conversation from whether workloads can move, to how organisations operate once they do.
Those expansion decisions are no longer driven solely by demand curves. They are informed by how industries operate, how governments regulate data, and where long-term digital capacity must be built. Cloud infrastructure, in this context, has become a strategic asset tied to national and sector-specific priorities.
Experimentation to execution
If cloud maturity has raised the bar, generative AI has raised expectations even further. Over the past two years, GenAI has moved rapidly from experimentation to executive mandate. Enterprises rushed to explore what was possible, often through proofs of concept designed to demonstrate potential rather than sustain operations.
In 2023, AWS launched the Generative AI Innovation Center to work directly with customers navigating that transition. The challenge, Ukpong recalls, became evident almost immediately. “When we started the Generative AI Innovation Center, there were a lot of POCs everywhere,” he says. “We said, let’s step back and find a way to assess properly which POCs would have the best chance to go into production.”
Production, he explains, introduces a different set of demands. Systems must connect deeply to enterprise data, align with how the business actually operates, and run reliably at scale. They also require transparency and traceability, so organisations can understand system behaviour when issues arise. Many early GenAI initiatives stalled not because the models failed, but because the surrounding operating model was never designed for production.
Those customer engagements reinforced a broader shift in how AWS views GenAI deployment—less as a generic capability and more as an industry-specific exercise. The insights gathered from early production efforts helped clarify where AI delivers practical value, and where industry context, operating discipline, and execution capability matter more than model sophistication.
As GenAI matures, its impact is becoming more clearly defined by industry context. Ukpong is explicit that value does not emerge uniformly. “Every industry has its own role to play,” he says.
In financial services, that role has centred on risk and fraud detection, where AI is already delivering operational impact. Ukpong also points to how development teams themselves are changing. “We’ve seen a lot of our customers, who have large development teams, really take on all of the vibe coding and the offerings we have,” he says, pointing to strong early adoption of AWS’s developer-focused tools.
Beyond vertical-specific use cases, AWS is seeing horizontal impact in areas such as supply chain operations. “One big horizontal area we’ve seen very interesting is in the supply chain space,” Ukpong says. Customers are using generative AI to analyse contracts in detail, understand where value can be driven with vendors, and ensure compliance across increasingly complex partner ecosystems.
Healthcare and life sciences illustrate how quickly AI can reshape core processes when conditions are right. Ukpong highlights “the speed at which they are doing drug discovery now using GenAI,” describing how organisations are revisiting historical research materials and proposing new formulations at a pace that would have been difficult to imagine just a few years ago.
What connects these examples is not novelty, but institutional learning. AWS is increasingly focused on capturing best practices within industries and applying them across markets. Ukpong recalls meeting recently with a major bank in Thailand and seeing an opportunity to transfer experience gained elsewhere. “As a big bank, they’re looking to take best practices we’ve had with other large banks and bring them to bear here,” he says.
That industry-led approach becomes even more critical when regulatory requirements enter the equation. Sovereign cloud, once discussed largely in policy terms, has become an architectural necessity.
“If you speak to every government out there, you’ll get a different definition of what sovereignty means,” Ukpong says. Yet when those definitions are broken down, common themes emerge: data residency, operator access control, transparency, reliability, and resilience.
In the Middle East, these priorities have shaped AWS’s partnership-led models. In the UAE, operator access control was a defining concern. “The UAE government will now put its data on the AWS cloud managed by e&, not by AWS employees,” Ukpong explains, describing how AWS addressed sovereignty by working with a local partner that manages the interface to the cloud.
A different construct underpins AWS’s recently announced $5 billion AI Zone investment with Humain. In this model, Humain will own and operate the data centre and manage regulated and sensitive workloads, while AWS provides the rack and server technology. “These are two different constructs,” Ukpong says, “but both solve for sovereignty.”
Skills, sovereignty, and scaling what works
As AI and industry workloads move into production, skills have emerged as a structural constraint. AWS is seeing strong growth in AI-focused certifications, while traditional cloud certifications begin to plateau after a decade of adoption. “We’ve put a lot of focus on the AI piece,” Ukpong says, pointing to just-in-time digital training, simulation-based learning, and hybrid delivery models in regions such as the Middle East and ASEAN, where in-person instruction remains important.
It is against this backdrop that Ukpong’s role at AWS is expanding to include responsibility for AWS Industries, while retaining oversight of sovereign cloud and regional investment decisions. For him, the shift is not about redefining strategy, but about applying what he has already learned at scale.
“The fact that for the last four years I have run the expert services gives me a much bigger insight into what I see our day-to-day customers need,” he says. That insight now shapes how AWS approaches industries—starting with customer strategy, working through how transformation happens in practice, how organisations grow, and how generative AI can be applied productively rather than experimentally.
With the announcements made at re:Invent 2025, Ukpong expects customer engagement to accelerate. “We expect that we’re going to see a flood of customer requests to engage and move forward,” he says.
Preparing for that demand is already underway. Ukpong points to the work happening across both his current and incoming departments to ensure teams are ready to respond. That includes enablement, shared learning, and clarity on how AWS brings its technical teams together to execute on customer projects.
“A lot of the work we’re doing now is all the enablement, learning, understanding how we respond to customers, and how we bring our technical teams to execute on those projects,” he says. “So that’s going to be the focus.”
Discussion about this post