AI doesn't have a trust problem. Your data does.

As adoption accelerates, governance lags

Organizations are already facing this governance gap reality.

Deloitte reported in 2025 that while AI pilots race ahead, many organizations are only beginning to build AI governance frameworks -- and some haven't started at all -- leaving internal audits to be pulled in late, after decisions have already been made.

The gap widens as systems grow more autonomous. In 2026, Deloitte reported that 74% of companies plan to deploy agentic AI within two years, but only 21% have a mature governance framework in place for those autonomous agents.

We are handing decisions over to systems we cannot yet explain.

The model isn't the problem

For data and analytics leaders, the issue is simpler than it appears. Auditability is not a model feature. It's a property of your data. Strip away the jargon, and every AI-driven decision should answer three questions:

• Where did this decision originate? (Lineage)
• Who is responsible for this decision? (Ownership)
• Did the decision create value? (Outcome)

If you cannot answer these questions, no amount of model documentation, explainability tooling or AI governance policy will save you.

Trusted AI rests on reliable data products that are governed, owned, versioned and traceable. Accountability cannot be delegated to a machine; someone must stand behind the decision. The audit trail provides the proof.

AI exposes cracks in the data foundation

Many of these data management weaknesses have existed for decades, long before GenAI arrived. We got away with these shortcomings, such as missing lineage and unclear ownership, because people sat between the data and the decision, quietly absorbing the ambiguity.

In a recent piece, I argued that many organizations' AI struggles stem from metrics that no one had linked to a business outcome. Here, the symptoms are different, but it's the same disease: decisions disconnected from reliable data.

AI is not the cause. It is the stress test, exposing weaknesses in data management that have been there all along.

Organizational charts won't solve this problem

Some believe the solution is to appoint a new executive -- such as a chief auditing officer -- to link the four seldom-connected functions: data governance, AI governance, internal audit and value measurement.

But fragmentation lets decisions slip through without transparent tracking. Adding a new title alone doesn't create new capabilities. Without changing existing workflows, this results in governance theater -- more dashboards and committees reviewing decisions long after they're made.

The real solution isn't a new officer but an operating model that builds accountability into trusted data products from the outset, so the audit trail is created as decisions are made, not reconstructed during a forensic effort after something goes wrong.

Why accountability fosters value

Auditability is not a compliance tax. It is a value enabler.

You cannot scale what you cannot trust, and you cannot trust what you cannot trace.

This is where auditability meets economics. A trusted data product is an appreciating asset. An AI model is a decision asset. The audit trail connects the two, creating a feedback loop that enables organizations to learn and adapt. Without it, you cannot tell which data products create value, which recommendations improve decisions or which models deserve more investment. Traceability provides visibility and ensures AI value compounds rather than leaks away.

Before you scale your next AI initiative, ask one question: If a regulator, a board or a customer asks why AI made a certain decision, can you trace it back to the data behind it, the person responsible for it and the outcome it was intended to achieve?

If you can't do this, the issue was never the AI, but the audit trail you never built. Successful organizations won't have the smartest models but the most reliable decisions. The real question isn't whether you can afford to build that trail; it's whether you can realistically and responsibly scale AI without it.

AI may automate decisions, but accountability still belongs to people.

Bill Schmarzo, "The Dean of Big Data," teaches AI-driven innovation at Iowa State University and advises organizations on data science, AI and data monetization. He is a former executive at Dell Technologies, Hitachi Vantara and Yahoo. He has written books on data-driven innovation and applied AI strategy.