Matt Waxman, Chief Product Officer at Precisely.
There’s a tension building inside every enterprise right now, and most leaders haven’t named it yet. AI is giving employees something they’ve always wanted: instant, unfiltered access to data. Ask a question, get an answer. No waiting on the BI team. No static dashboards. No ticket submitted and forgotten. Just insight, on demand.
And once people experience that, they don’t go back. The problem is that this collides directly with the centralized, governed data infrastructure enterprises have spent decades building and for very good reason. As data volumes grow, the consequences of poor governance do as well.
That collision is what I’m calling the democratization trap. And it’s catching more companies off guard than anyone wants to admit.
The old model was intentionally built around a bottleneck. For years, the constraint on data access was a feature, not a bug. You didn’t pull a number without a data team validating it. You didn’t build a report without a governance process behind it. The BI team was slow, yes, but that slowness was doing real work. AI removes that checkpoint. When employees can query the entire data estate through a chat interface, controlled, human-reviewed data pipelines are increasingly sidelined. What used to move through a narrow, validated funnel now flows through all at once.
A Gartner report found that poor data quality costs organizations an average of $12.9 million per year. In a world where AI opens the floodgates, that number may be only the starting point.
What The Pattern Looks Like On The Ground
In conversations with organizations across financial services, manufacturing and healthcare, the same dynamic keeps emerging: companies racing to expand AI-driven data access while data quality, lineage and governance infrastructure quietly lags behind.
The urgency is real. The competitive pressure is real. So teams move fast. And the gap between what AI promises and what the underlying data can support keeps widening, invisibly.
The warning signs don’t announce themselves loudly. They show up like this:
• Shadow metrics. Different teams cite different numbers for the same KPI, each pulled by a different AI tool from a different dataset, with no one able to trace which source is accurate.
• Decision laundering. AI-generated outputs get used to justify decisions that would previously have required analyst review, stripping accountability from the process without anyone acknowledging it. The answer looks authoritative. No one checks where it came from.
• Silent abandonment. A team discovers one bad AI output and discreetly stops using the tools entirely. Adoption collapses, not with a complaint, but with a return to spreadsheets.
By the time any of these surface visibly, they’ve been building for months.
The Agentic Inflection Point Changes The Stakes
Today, AI mostly answers questions. A human asks, something answers, a person decides. That loop still has judgment in it. There is still a checkpoint.
Agentic AI removes that checkpoint too.
When agents act autonomously, adjusting inventory, triggering procurement, flagging credit applications, the data underneath those decisions stops being an analysis input and becomes an operational instruction. An agent making autonomous decisions on unverified data introduces risk at machine speed and scale.
According to McKinsey’s 2025 State of AI survey, 88% of organizations now use AI in at least one business function, up from 78% a year earlier. Meanwhile, 62% of organizations are experimenting with AI agents, but just 23% have scaled agentic systems in production. The move toward agentic workflows is accelerating, yet most enterprises are closer to this inflection point than their governance infrastructure is ready to handle.
That gap is the thing worth sitting with.
Three Things That Bridge The Gap
The answer isn’t to slow down AI adoption. The competitive reality makes that a losing position. The answer is to treat data infrastructure as the prerequisite it actually is, not the afterthought it has become.
This is the work many enterprise teams are in the middle of right now, and the best practices aren’t yet fully known. But here’s what’s emerging from organizations that are getting it right:
• Instrument before you open. Before expanding AI-driven access to new datasets or user groups, establish lineage tracking and data quality scoring at the source. If you can’t tell an AI agent where a number came from and when it was last validated, that number shouldn’t be in play.
• Rebuild governance for the volume AI creates. Traditional governance was designed for low-volume, high-control environments. AI multiplies query volume by orders of magnitude. Governance needs to keep pace through policy-as-code, automated quality checks and access controls that travel with the data, not just with the user.
• Make accountability part of the output, not just the access. Every AI-generated insight that drives a significant decision should carry a traceable audit trail. Not because regulators will require it, though many will, but because organizations that can’t explain how a decision was made cannot learn from it, defend it or correct it.
This is a point of view, not a finished blueprint. But the direction is clear enough to act on.
The Real Competitive Edge
The organizations getting this right are not moving slowly. They invested in data infrastructure early, treating it as a strategic priority before AI made that foundation urgent. Now they can move faster, and with more confidence. AI agents are coming whether the data foundation is ready or not. The question every enterprise leader needs to sit with is whether their data can actually support autonomous action at speed and scale.
Most organizations aren’t there yet. As agentic AI moves closer to production, that readiness gap is becoming harder to ignore.
The question I’m still working through: How do you build the governance layer fast enough to keep pace with adoption, without making it so heavy that it re-creates the bottleneck you were trying to move past? I don’t think that has a settled answer yet.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
