Global – October 20, 2026 -- Capgemini identifies three data failures (fragmentation, stagnation and semantic inconsistency) as the main barrier keeping enterprise AI confined to pilots, particularly in financial services and other highly regulated industries.
Capgemini's Renjith Sreekumar says scaling problems lie in data, not models
Renjith Sreekumar, Head of Oracle Cloud & AI CoE for Financial Services at Capgemini, frames the issue as a shift from "does AI work?" to why it is so hard to scale. Early gains in analysis, content creation, workflow automation and customer experience remain limited to pilots, single functions or tightly controlled use cases.
Fragmented data leaves AI agents blind to account, transaction and risk inputs
Critical information is spread across operational systems, cloud platforms, data warehouses and departmental applications. A financial services agent may need account information, transaction history, service interactions, product eligibility, risk indicators and regulatory requirements. If those sit in disconnected environments, the agent is fast but poorly informed. Connecting them must still meet security, privacy and governance standards.
Batch processing delivers stale data to AI platforms
Data trapped in systems of record or moved through batch processes may be copied multiple times and stripped of business context before it reaches an AI platform. An AI system working from yesterday's information can miss a shift in customer behavior, an emerging operational issue or a new risk signal. Capgemini calls for bringing AI closer to current operational data.
Conflicting business definitions make AI outputs harder to validate
Business units may define customers, products or risk events differently, and the same KPI can be calculated several ways depending on the team or platform. Employees bridge such gaps through experience, but AI cannot do so reliably without a shared semantic foundation. Teams then debate numbers, and automation risks applying the wrong interpretation at scale.
Capgemini outlines a three-part foundation for scaling agents
The company says enterprise AI depends on trusted enterprise data, business context and semantics, and intelligent agents and automation. Delivering them requires coordinated work across data architecture, cloud modernization, governance, security, operating models and AI adoption.