Solution Lab: Trusted Data Services for AI Agents: From Pilot to Production
Webinar Speaker: Donald Farmer, TDWI Research Fellow
Date: Thursday, December 3, 2026
Time: 9:00 a.m. PT / 12:00 p.m. ET
Business teams now build their own AI agents; for example, a procurement team can build an agent that checks supplier records and open commitments when a purchase order is raised. Agents like this act on enterprise data inside business processes, so if errors occur, they may be repeated in every case until someone notices.
Yet many of these agents stall between pilot and production. The causes are usually data problems. An agent may use one of two conflicting supplier records, report a figure that no one can trace, or show a user data they are not permitted to see. The data team cannot fix this by building custom tools for every agent because the requests multiply too quickly.
The capabilities that prevent these errors already exist: data quality, master data, lineage, and access policy. What is missing is a way to use them in agentic scenarios. However, when they are available as governed services in the agent platform’s tool catalog, the people building an agent can find data sources, check lineage and quality, and confirm that the data is approved for the planned use.
In this TDWI Solution Lab, research fellow Donald Farmer gives an independent view of this category, drawing on new TDWI research. Specialists from Informatica and Microsoft will demonstrate how to build an agent using trusted data services (including customer identification, data lineage, and policy management) to create agents in Microsoft Copilot or Foundry that you can confidently move into production.
Attendees will learn:
- Why agents built by business teams stall before production, and which data problems cause it
- How trusted data services help builders find, check, and confirm approval for data before an agent is built
- What changes when data quality, master data, and lineage are services agents can call as they run
- How business and data teams divide the work of building, reviewing, and monitoring agents