Banks Rush to Deploy AI Agents While Struggling With Data Quality and Context
A new study reveals that financial institutions are rolling out AI agents faster than they can explain their reasoning, with most organizations lacking the data infrastructure and contextual frameworks needed to support these systems at scale.

Technology leaders across banking, insurance and government sectors acknowledge significant shortcomings in the data powering their artificial intelligence systems, according to research released Tuesday by IDC and commissioned by Quantexa, an AI software vendor. The study, which gathered responses from 366 senior technology executives at large financial and government institutions, found that 43% believe their data requires either substantial improvement or a complete redesign.
Financial services firms are deploying autonomous agents into their operations at a pace that outstrips their capacity to document and explain how those agents reach their conclusions, the research indicates. Context—encompassing the full range of instructions, rules, knowledge and memory that agents rely on to reason and act—remains largely fragmented across organizations. Despite expectations that agents will handle roughly one-third of organizational work within the coming year, fewer than 9% of surveyed companies have successfully embedded context throughout their decision-making and AI workflows.
Organizations have invested in the building blocks of context, but those capabilities sit in pockets
Stewart Bond, VP of data intelligence and integration software at IDC
Data Fragmentation Undermines Competitive Position
Fragmented data systems are eroding banks' ability to compete with consumer-facing AI tools. According to McKinsey & Company's Global Banking Annual Review 2026, nearly three-quarters of consumers seeking AI-driven financial guidance turn first to OpenAI's ChatGPT, Anthropic's Claude or Google's Gemini rather than their own bank's offerings. Agents are already reshaping financial transactions: Meta's Muse agent accesses consumer account balances and transaction history via Plaid, while PayPal has integrated its wallet functionality into ChatGPT's payment system.
Even as banks and insurers assemble the foundational elements needed to support agents and AI capabilities, implementation remains scattered. The Quantexa study found that only about half of organizations have deployed multiple context capabilities, and these are typically confined to specific use cases rather than applied organization-wide. Knowledge graphs—which map the logical pathways AI systems use to make decisions—represent the least widely adopted context capability, with only around 17% of companies reporting broad implementation.
Bridging data silos emerges as one of the most significant obstacles when organizations attempt to construct a comprehensive context layer. Without access to appropriate data, particularly information about data lineage and decision histories, financial institutions cannot trace how their agents arrived at specific conclusions. This gap directly undermines confidence in AI-driven decisions.
Lack of data and decision lineage, not model complexity, is what most limits trust in AI decisions, and a real-world relationship view is the context organizations are missing most
Stewart Bond
To rebuild trust and move context capabilities beyond isolated pilots to enterprise-wide deployment, IT leaders must establish data lineage and decision data as fundamental design requirements, the study recommends. Business executives, meanwhile, should clarify which decisions agents will make autonomously and which require human oversight, while also specifying the contextual information agents need to operate effectively within their organizations.
Vendors Address Context as Scaling Challenge
Technology vendors are increasingly recognizing context as a critical bottleneck for enterprises seeking to scale agentic AI systems. On Tuesday, during Microsoft Ignite, Salesforce and Microsoft unveiled Copilotforce, a joint offering designed to integrate Salesforce customer context and AI capabilities into Microsoft's software ecosystem.
AI is only as powerful as the data and context that grounds it
Judson Althoff, CEO of Microsoft Commercial Business


