Key Takeaways
- AI is the first technology shift where getting the technology right isn’t enough, because the hardest part now sits in the data underneath it.
- People used to cover for data gaps with judgment. AI agents act without that review, so the quality of the data becomes the quality of the decision.
- Organizations that treat data trust as a competitive advantage can defend what their AI does, and earn the right to give it more responsibility.
Every major technology shift in my career has required getting the technology right. The AI movement is the first one where that isn’t enough.
Let me be clear: the technology still matters. But the hardest part of AI now sits underneath it, in the data.
MIT Technology Review Insights finds that, on average, AI has access to only 45% of enterprise data. Gartner predicts 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. Data readiness is a critical part of addressing those challenges.
AI ambition is colliding with the reality of enterprise data, and AI-ready data is what separates the organizations that scale AI from the ones that end up walking it back.
What makes this moment different for enterprise AI?
For more than 30 years, at companies like Microsoft, Oracle, Sage, and now Precisely, I’ve watched technology shifts reorder whole industries. In each one, the technology was the hardest part, and people managed around the gaps. Analysts caught errors, teams reconciled breaks between systems, and critically, human judgment filled in whatever the data couldn’t.
That judgment was an invisible subsidy. Organizations rarely had to ask whether their data was ready, because people made it ready one decision at a time.
AI agents don’t work that way. They act on the data they’re given, instantly and at scale. There is no time in between for people to pause and ask whether a record looks wrong. That data sits across mainframe, on-premises systems, cloud platforms, and SaaS applications, largely disconnected and inaccessible. And some of that data is stale, inaccurate, or incomplete. Even good data often means different things in different places: one system’s “customer” or “product category” doesn’t line up with another’s. Put agents to work on that data, and you won’t get a trusted outcome.
With no human judgment in the loop, the quality of the data becomes the quality of the decision.
If you want trusted AI, you need trusted data.
Data readiness is a leadership decision
Too often, data readiness gets treated as technical cleanup that happens somewhere below the executive level and is seen as a one and done activity. I think it’s a leadership decision, and it has to be made before an agent is allowed to act.
Before AI acts on your data, you should be able to answer a few plain questions. Can AI reach it? Can you rely on it? Can you trace a decision back to the data and the policy behind it? Does it carry the context that decision needs? If the answer to any of them is no, AI won’t settle that uncertainty. It will carry it into decisions at machine speed.
Trust is the competitive advantage
Data trust can be a compliance obligation, or a competitive advantage. The organizations that treat it as an advantage will be able to let AI act on their data, and to defend what it does. They’ll move faster because they won’t second-guess their own systems. Everyone else will spend the next few years walking AI back.
Consider what autonomy changes. When an agent approves a claim, reroutes a shipment, or answers a customer, someone will eventually ask why. The organizations that can answer with evidence, not assurances, will earn the right to give AI more responsibility. The ones that can’t will keep a human in every loop and give back much of the speed AI was supposed to deliver.
I’m optimistic because the AI readiness gap can be closed, if the right data foundation is in place.
The next phase of enterprise AI will reward the organizations that take the steps to make their data trustworthy first.
