As organizations give AI agents more responsibility, CIOs face a practical challenge: ensuring those agents can access the business knowledge people use to make decisions, with appropriate controls. Without that knowledge, agents can act quickly on incomplete assumptions.
At Accenture, I saw this gap repeatedly. Teams had the official number but still had to find the people who understood its assumptions. Decisions slowed while they pieced together knowledge the organization already had. At critical times a model was reconstructed as the landscape of context changed; manually, and frequently.
In July 2024, McKinsey reiterated an estimate that ineffective decision-making cost a typical Fortune 500 company roughly $250 million annually in management wages.
The forecast missed what the business knew
Consider a sales team relying on a $50 million deal to beat its quarterly target. The forecast looks promising, but it misses the buyer’s history of lengthy reviews. Former account team members knew this pattern, but it wasn’t captured in the CRM. The deal slips, leaving the team short of target.
That history should have shifted attention toward smaller, more certain expansions, despite the hit to profit. An agent recommending where to focus needs the same buyer history, available alternatives, and business priorities.
Business context extends beyond the recorded facts
Business context is the knowledge that changes how data should be interpreted or acted on: goals, constraints, exceptions, and experience; a multi-dimensional model. It spans business systems, account plans, policies, spreadsheets, and prior analyses. Some remains in people’s experience and must first be captured. Previously this was a huge corporate endeavor.
When teams reconstruct knowledge in private conversations or spreadsheets, the next team repeats the work. Agents serving different functions can inherit the same fragmentation and produce inconsistent recommendations. This all gets missed or backlogged in centralized systems.
Keep the warehouse and expand the context
Governed data remains essential. The CRM can remain authoritative for the recorded forecast while reviewed account history provides grounds to reconsider it. An authorized owner decides whether to update the official record.
A composite context landscape connects governed data with relevant business knowledge wherever it lives, as an evolutionary process, while preserving each source’s authority and permissions. Agents introspect over the knowledge sources and true decision traces to identify gaps for people to resolve. Capturing those clarifications with their source and scope helps subsequent teams and agents build on the work. Those nearest the domain remain responsible for governance. The composite structure is a local evidence based approach that evolves based on system evaluation results.
This is why we built QueryStory, to bring agentic analytics to the last mile of decision-making, where business teams turn analysis into action without requiring specialized teams or manual capture and governance. Agents gather context from systems, documents, and working files, including sources that haven’t been standardized for analytics. The composite context helps to deliver narratives, and interactive outputs to drive action. Traceable logic, cited sources, and human review make the reasoning inspectable. Conflict detection, and approval controls govern what gets published, as simple as giving the go-ahead. Spend less time assembling context, knowing what or where to source next, and achieve unparalleled basis for defensible outcomes.
What this changes for the CIO
For CIOs, the opportunity is practical: less time spent chasing information, writing metadata, or manual collection of ontological data, more consistent services to business line users, and a maximized value from central data infrastructure context and last mile data helps the ecosystem.
As agentic adoption expands, knowledge captured in one workflow should help the next team or agent make a better-informed decision. Treating business context as shared infrastructure can reduce repeated effort and support more consistent action across the organization. It gives CIOs a foundation for extending an agents’s responsibilities while preserving oversight and getting more value from existing data and governance investments.
Join our Gartner IT Symposium/Xpo session, Unifying Fragmented Business Context for AI Scale, to learn how to identify the context a decision requires, establish ownership and controls, and make that knowledge available to agents without custom-building a backend. Visit the QueryStory booth #525 in the Data & Analytics marketplace to explore the approach.