Why AI pilots struggle to scale inside enterprises: Glory Nelson

Enterprise AI has moved well beyond experimentation. Organisations are investing heavily in models, platforms and infrastructure. Yet many continue to struggle to turn pilots into business-wide transformation.

In an interview with People Matters, Glory Nelson, Country Head, India Delivery Centre at Xebia, says the problem is rarely a lack of ambition. Instead, enterprises are finding it difficult to prepare their people, redesign work and create the organisational conditions needed for AI to deliver measurable value.

Workforce transformation has become the real AI challenge

According to Nelson, organisations often approach AI as a technology programme when the larger challenge sits within the workforce itself.

"As organisations move toward AI-first operating models, the biggest workforce challenges are no longer just about technology adoption. They are about whether the organisation can equip its people, redesign the way work gets done, and build the confidence needed to operate in a more intelligent, data-driven environment."

She identifies four workforce challenges slowing enterprise AI adoption:

• Capability gaps, where employees must combine domain expertise with AI literacy, data understanding and stronger decision-making skills.
• Workflow redesign, as traditional process-led structures struggle to support cross-functional, adaptive ways of working.
• Organisational alignment, with AI initiatives often developing independently across business units without shared governance or standardised frameworks.
• Cultural readiness, where uncertainty around changing roles and limited confidence in new ways of working continue to slow adoption.

Nelson believes organisations making the transition successfully are treating AI adoption as a workforce transformation agenda rather than simply deploying new technology.

Why AI investments fail to deliver enterprise-wide impact

Despite growing investment in AI infrastructure, Nelson says many initiatives stall because organisations are unable to operationalise the technology.

"Many enterprise AI initiatives are struggling not because organisations lack ambition or investment, but because technology is moving faster than the organisation's ability to absorb and operationalise it. The gap is increasingly one of execution, not intent."

She points to five recurring obstacles:

• Readiness gaps before workforce capability and operating discipline are established.
• Weak data foundations, including fragmented and low-quality enterprise data.
• Disconnected pilots that never become embedded into core workflows or linked to business KPIs.
• Governance inconsistency across teams.
• Low organisational confidence in AI-assisted workflows and accountability.

Nelson says stronger returns come when organisations invest equally in data readiness, governance, workflow integration and workforce enablement, rather than focusing primarily on technology.

From digital capability to AI-native work

Nelson draws a distinction between organisations that are digitally capable and those becoming AI-native.

"In practical organisational terms, the difference is simple, a digitally capable workforce uses technology effectively, while an AI-native workforce works with AI as an embedded part of how decisions are made, workflows are run, and value is created."

She says AI-native organisations differ across several dimensions:

• AI supports decision-making rather than simply automating tasks.
• Processes are redesigned around human-AI collaboration.
• Employees are expected to validate AI outputs, apply business context and understand governance.
• AI learning becomes part of everyday work instead of being treated as periodic training.

For Nelson, organisations become AI-native only when AI becomes part of the operating model rather than remaining an isolated capability.

Enterprise roles are changing across every function

AI is also reshaping how work is distributed across technical and business teams.

According to Nelson, developers are increasingly expected to manage AI orchestration, prompt engineering, model integration and governance alongside software development. Analysts are moving towards validating AI-generated insights and translating them into business decisions.

Managers, meanwhile, are taking responsibility for enabling human-AI collaboration, redesigning workflows and driving responsible AI adoption. Business teams are also becoming more involved in AI-assisted execution and operational decision-making, increasing the need for AI literacy beyond technical functions.

Human factors continue to slow adoption

Nelson believes the biggest barriers to meaningful AI adoption are behavioural and organisational rather than technological.

"The challenge is not access to AI, but whether employees feel clear about their role, confident in the outputs, and supported by the wider organisation to use it responsibly."

She highlights several factors slowing adoption:

• Uncertainty around changing responsibilities and performance expectations.
• Limited confidence in interpreting AI-generated outputs.
• Change fatigue as organisations manage multiple transformation programmes simultaneously.
• Fragmented implementation and inconsistent leadership communication.

Nelson also notes that Xebia Academy's Data & AI Literacy programmes identify 46% of organisations as viewing the absence of a strong data-driven culture as a major obstacle to AI success.

AI capability building needs to move beyond classroom learning

Traditional training models are no longer sufficient, Nelson says, because AI capabilities evolve continuously.

Employees increasingly need practical experience with AI-assisted workflows, governance frameworks, prompt engineering, model evaluation and decision-support systems. Different functions also require role-specific learning aligned to their operational responsibilities.

Nelson says organisations see stronger outcomes when AI capability building is integrated into real business workflows rather than delivered through isolated learning modules.

Scaling AI requires enterprise-wide discipline

Looking ahead, Nelson believes successful organisations will be distinguished by how effectively they integrate AI into everyday business operations.

She identifies four characteristics shared by organisations building AI-ready workforces:

• Integrated operating models where AI becomes part of routine execution and decision-making.
• Enterprise-wide capability building across technical and business teams.
• Shared governance and standardised frameworks supporting consistent implementation.
• Strong data foundations and coordinated human-AI collaboration that enable scale.

For Nelson, enterprise AI maturity will increasingly depend on organisational capability rather than technology alone. Organisations that build workforce readiness, governance and operating discipline together are likely to scale AI more effectively, while fragmented approaches risk limiting long-term business impact.