- Model: Built-in AI systems that interpret signals, generate responses, and make predictions
- tool: Integration layer that connects AI to enterprise systems, such as APIs, protocols, and connectors
- Context: Before making decisions, information agents need to understand the entire business picture, including customer history, product catalogs, and supply chain networks.
- Government: Policies, controls, and procedures that ensure data quality, security, and compliance
This framework helps identify where reliability gaps emerge. When an enterprise agent fails, which quadrant is the problem? Is the model intended wrong? Is equipment unavailable or broken? Is the reference incomplete or contradictory? Or is there no mechanism to verify that the agent did what it was supposed to do?
Why is this a data problem, not a model problem?
It is tempting to think that as models improve, reliability will also improve. Nevertheless, model capability is increasing rapidly. estimated cost is fell about 900 times in three years, Hallucination rates are decliningAnd AI’s ability to perform long tasks doubles every six months.
Tooling is also gaining momentum. Integration frameworks like Model Context Protocol (MCP) make it dramatically easier to connect agents to enterprise systems and APIs.
If the models are powerful and the tools are maturing, what is hindering adoption?
To borrow from James Carville, “It’s data, stupid.” The root cause of most misbehaving agents is misaligned, inconsistent, or incomplete data.
Enterprises have accumulated data debt for decades. Acquisitions, custom systems, departmental tools and shadow IT have left data scattered across silos that rarely agree. Support systems do not match what exists in marketing systems. Supplier data is replicated across finance, purchasing and logistics. Places have multiple representations depending on the source.
Drop a few agents into this environment, and they’ll perform amazingly at first, because each is given a curated set of systems to call upon. Add more agents and the cracks grow, as each person creates their own piece of the truth.
This dynamic has been seen before. When business intelligence became self-service, everyone started creating dashboards. Productivity soared, reports failed to match. Now imagine that event not in a static dashboard, but in AI agents that can take action. With agents, data inconsistency produces real business consequences, not just arguments between departments.
