AI's data problem is governance in disguise

 As AI becomes standard practice across Australia, organisations are making their expectations clear. 

Workiva's 2026 Midyear Executive Benchmark Survey of Australian finance, risk, and sustainability leaders found that 95 per cent said they want AI to deliver measurable ROI within the next year. It's an ambitious timeline, and in reporting, where accuracy and accountability are non-negotiable, many will be hitting the same wall. 

Of the executives surveyed, only 14 per cent said their organisation's data quality had been sufficient for AI use in reporting, while one in five (21 per cent) said data quality issues had significantly limited their use of AI, blocking deployment in key workflows. The most telling response was that 44 per cent said they'd worked around these issues, but it slowed progress.

It's tempting to read those numbers as a data problem. I'd argue it's a governance problem wearing a data costume. 

AI amplifies what it's given 

When reporting workflows lack a single source of truth, with the same data point living in three spreadsheets, a slide deck, and an email thread, AI isn't going to resolve inconsistency. It will only amplify it. Yes, outputs are produced faster, but they're also produced at greater scale, and with a confidence that makes errors harder to spot. 

The practitioners using it are already inclined to take that confidence at face value. In the same survey, 79 per cent said they were at least somewhat confident in the accuracy of AI outputs without human review.

AI is an accelerant. Point it at a well-governed reporting environment and it accelerates valuable insight, but point it at a fragmented one and risk runs rampant.

The 21 per cent who know they're blocked are, in a sense, in the safer position as they've recognised the problem and stopped before it has compounded.

The greater vulnerability lies with organisations that have normalised the workaround: a manual reconciliation, offline copies, the memory of a trusted analyst. These solutions might've kept things moving the first cycle around, but over time they erode traceability and, with it, certainty over which source figures fed a particular AI-assisted output. 

In reporting, knowing where the numbers come from matters. Financial statements, board papers, and sustainability disclosures, including the mandatory climate reporting now phasing in across Australia, must stand up to auditors, regulators, and investors. A number you can't trace is a number you can't defend, meaning an AI-assisted process that can't be audited turns a productivity gain into a liability. 

Better AI starts with better foundations

When AI underdelivers, the instinct is to look for a better model or a smarter tool. But closing the gap requires connected and controlled data infrastructure that gives AI something trustworthy to work with in the first place.

In practice, that means three things: 

• A single source of truth, with each figure living in one place and flowing through to every report, board paper, and disclosure that uses it, so a change made once is reflected everywhere 
• Control, with clear ownership, permissions, and version history, so there's no ambiguity about who approved what, and when 
• And lineage, built in rather than reconstructed after the fact, so any number in any output can be traced back to its source in minutes, not days

That foundation makes AI useful rather than risky. Its outputs inherit the trust built into the data, so teams can then use AI to draft commentary, flag anomalies, and reconcile data with confidence, because every step can be verified.

Earning the label

Many organisations now describe themselves as "AI-ready", but until their reporting environments have lineage and control built in, that's a claim their own systems can't yet back up. Audits suggest as much, with 39 per cent of Australian executives saying their most recent AI audit uncovered data lineage gaps or traceability issues.

There's a simple test anyone in reporting can run. Take one AI-assisted figure from your last board report and ask your team to trace it back to its source. If getting an answer takes days, relies on a spreadsheet only one person understands, or depends on someone's memory, you're not AI-ready. You're exposed.

Organisations that are genuinely AI-ready can show evidence. It's what gives teams confidence in their work, business leaders confidence in their decisions and reporting, and boards confidence to invest in AI as a real advantage.