Artificial Intelligence is the most hyped technology of the last 35 years. While past innovations like the cloud required software vendors to push for adoption, AI has organizations moving so fast that experts are actually advising them to “slow down” and build a proper foundation.
At the recent BLUEPRINT 4D conference, Doug Cosby, the original inventor of Oracle’s enterprise data tools, shared why Enterprise Data Management (EDM) is the “money shot” for any successful AI strategy. If your AI can’t tell the difference between two similar data points, it isn’t the AI’s fault—it’s a data hygiene problem.
Moving Beyond the “Data Glob”
Cosby argued that AI doesn’t fail because the models are bad. It fails because enterprise data is inconsistent, disconnected, poorly described, and governed differently across systems.
Traditional Master Data Management (MDM) often fails because it tries to force every department into a single “glob” of data. This “Big Bang” approach requires everyone to agree on everything upfront, which usually stalls projects for years.
Modern EDM uses a “bubble” or “app model” approach. This allows different departments—like HR, Finance, and Marketing—to maintain their own versions of the truth while remaining linked through a governed system. This agile, “baby-step” method means you can implement EDM alongside your AI projects rather than holding everything up for a massive cleanup.
Why Cross-System Relationships Matter
Those governed links highlight one core concept: AI needs relationships. Enterprise systems rarely align perfectly. Product structures differ between operational systems and analytics environments, and planning hierarchies often intentionally diverge from production structures.
Humans navigate those differences naturally because they understand the business. AI needs those relationships explicitly mapped.
That’s where EDM becomes foundational. It maintains the governed links between systems so AI can understand how entities relate across the enterprise. Without those mappings, AI can answer questions inside a single application, but cross-functional “why” questions become much harder to solve.
For example: Why didn’t a hierarchy update correctly? Why is planning missing a department? Why did one system approve a change another rejected? Why are reporting totals inconsistent?
Enterprise AI can answer those questions, but it requires context, governance history, and connected enterprise data — not just raw records.
The Surprising AI Problem Nobody Talks About
One of the most interesting moments of the session came when Cosby described a customer struggling with AI validation because they had dozens of nearly identical account descriptions. AI couldn’t reliably determine which “Cash and Cash Equivalent” account users actually meant.
The fix wasn’t complicated; they cleaned up the descriptions. That was it. By making descriptions more intentional, unique, and business-specific, they dramatically reduced AI confusion and improved response quality.
It sounds almost too simple, but Cosby stressed that descriptive clarity is becoming one of the biggest drivers of AI accuracy. AI systems rely heavily on metadata, naming conventions, and business context to understand intent. Weak descriptions create weak AI outcomes.
The Two Pillars of AI Success: Links and Descriptions
AI is essentially a sophisticated text-prediction engine. A human can often infer meaning from context, but AI cannot. If descriptions are vague, hierarchies are inconsistent, or systems aren’t properly linked, AI starts guessing. That’s where hallucinations, conflicting answers, and unreliable outputs begin creeping into the business.
To make it work for your business, Cosby identifies two non-negotiable requirements:
- Intent-Driven Descriptions: AI struggles when objects have vague names. As Crosby’s anecdote pointed out, the customer had multiple accounts named “Cash Equivalent,” causing AI to hallucinate or pick the wrong one. By using EDM to create clean, unique, and intent-driven descriptions, organizations have seen hallucinations drop significantly.
- Cross-Pillar Links: Again, AI is most powerful when it can answer “why” across different systems. To do this, it needs to know that “Department 100” in your HCM system is the same as “D100” in your Finance system. EDM manages these links so your AI agents can navigate the entire enterprise.
Practical AI Agents in EDM
Cosby highlighted three “prongs” of AI integration that are changing how data is managed:
- Productivity Agents: These help build requests automatically. Instead of manual dragging and dropping, a user can ask an agent to “move all cost centers in Department 5 to Department 6,” and the AI builds the request for review.
- Cross-Pillar Agents: These agents use the “links” mentioned above to answer complex business questions by pulling data from multiple independent “bubbles” like Oracle, SAP, or Workday.
- External Scrapers: EDM is beginning to “scrape” data from third-party tools like JIRA, Service Now, or even email inboxes. It can identify a request for a new customer or cost center, initiate it in EDM, and notify the original tool once the governance process is complete.
Governance is Your Safety Net
Using AI to build data requests or move cost centers sounds risky. However, because EDM uses a request-based governance model, it doesn’t matter if a human or an AI initiates a change. Every action must still pass through your established workflows and policies before it ever touches your production systems.
As you build your AI roadmap, remember: the intelligence of your AI is capped by the quality of your data foundation.
The Future of AI Is Business-Led Data Governance
Perhaps the biggest takeaway from the session was that AI readiness is really data readiness.
Organizations don’t necessarily need a massive rebuild before adopting AI. But they do need cleaner descriptions, stronger governance, better mappings, and business-owned data processes. Because at the end of the day, AI is only as smart as the enterprise data feeding it.
And as companies continue rushing toward AI adoption, the organizations that invest in enterprise data foundations now will likely be the ones that see the most value later.
Want more?
Explore more content and resources to help you get the most from your Oracle investment:
- Visit the Quest Learn Library for blogs, how-to demos, and on-demand sessions.
- Connect with peers in one of our Quest Cloud Community User Groups to swap stories, ask questions, and share tips with other users facing the same challenges.
Not a Quest member yet? Join today and tap into the ultimate Oracle customer network.