Most MDM books begin with the golden record. This one begins with a harder question: what happens when software can reason about that record, challenge it and act on it?
Agentic MDM: Master Data Management for the Autonomous Enterprise is structured as an architectural journey rather than a catalogue of AI features. It starts by rebuilding the foundations autonomous systems depend on — identity, relationships, hierarchies, quality, freshness, provenance and confidence — then introduces agents only after those foundations are explicit.
The book turns the problem in both directions: Agents for MDM, where agents help discover, match, enrich and steward master data, and MDM for Agents, where enterprise agents consume governed master-data context while making live business decisions.
At the centre of the architecture is one principle:
An agent may reason about enterprise truth without automatically gaining the authority to change it.
That principle becomes operational through Proposal → Policy → Commit, separating what an agent recommends from what policy permits and what the enterprise ultimately commits. Human authority, least privilege, evidence, rollback and accountability remain explicit rather than disappearing inside an orchestration workflow.
The architecture then expands into runtime trusted context, MCP, APIs, events, headless MDM, context packaging and multi-agent coordination. Later chapters deliberately attack the design with false merges, stale context, identity ambiguity, security failures, authority escalation, evaluation gaps and weak decision evidence.
The production question is therefore not simply whether an agent is accurate. It is whether a consequential decision can be evaluated, governed, observed, reconstructed and reversed.
Asterion Industrial Group provides a recurring synthetic enterprise thread throughout the book, allowing supplier identity, stewardship, runtime context, security, authority and governance decisions to evolve rather than resetting the scenario in every chapter.
The final chapters move from architecture to adoption: operating model, real-world implementation evidence, architecture failure patterns and a 90-day path for taking one bounded Agentic MDM workflow into controlled production.
Inside the Book - Part I — When the Golden Record Becomes the Starting Point: Redefine MDM for machine action, establish what “agentic” actually means and introduce the Agents-for-MDM ↔ MDM-for-Agents dual loop.
- Part II — Building Agent-Grade Master Data: Engineer identity, relationships, quality, freshness, provenance, confidence and the Agent-Grade Master Data Contract.
- Part III — Agents for MDM: Apply agents to discovery, matching, enrichment and stewardship while governing change through Proposal → Policy → Commit and explicit human authority.
- Part IV — MDM for Agents: Serve trusted context at runtime through MCP, APIs, events and headless MDM, then package that context for enterprise and multi-agent workflows.
- Part V — Trusting the Agent Enough to Let It Act: Establish least-privilege access, evaluation, observability, evidence and auditability before expanding authority.
- Part VI — Putting Agentic MDM into the Enterprise: Build the operating model, examine real-world evidence, challenge architecture failure patterns and execute a 90-day adoption roadmap.
An optional companion pack adds practitioner exercises, worked solutions, schemas, fixtures and governance artefacts for readers who want to test the architecture rather than only read about it.
