The Dirty Secret About AI and Data Quality

“We can’t implement AI until our data is clean.” It is one of the most common statements heard in boardrooms and project meetings today. On the surface, it sounds sensible. Artificial intelligence depends on data, so better data must come first.

Except it does not. In reality, this argument has become the modern equivalent of “we’re not ready yet.” It delays projects by months, sometimes years, while organisations chase a standard of data quality that nobody can actually define. How clean is clean enough? Nobody seems to know.

Every supply chain runs on imperfect data. Supplier lead times are wrong. Minimum order quantities change. Production rates drift. Transport routes evolve. Customers order differently than expected. Third-party data arrives late or incomplete.

Even if an organisation spends millions cleansing its ERP master data, it still cannot control supplier performance, shipping delays, geopolitical disruption or customer behaviour. The irony is simple. The world changes faster than master data can.

The obsession with perfect data did not appear by accident. Traditional planning engines such as MRP are highly deterministic. They assume that lead times are correct, batch sizes are correct, calendars are correct, supplier constraints are correct, and routing information is correct.

If one of those assumptions is wrong, the planning output becomes progressively worse. The only answer was to continuously maintain the master data. This created decades of data governance projects whose sole purpose was keeping planning systems alive.

Modern AI changes the equation. It does not have to treat master data as absolute truth. It can observe reality. Instead of asking, “What does the database say the lead time is?” it can ask, “What has actually happened over the last six months?” Those are two very different questions. One usually produces a much better answer.

Transaction history contains an enormous amount of intelligence. Properly designed AI can calculate effective supplier lead times, actual production rates, realistic batch sizes, transport reliability, supplier consistency, demand behaviour and inventory movement patterns.

Instead of relying on values someone entered into an ERP system five years ago, AI can continuously learn from what is actually happening. That is not poor data quality. That is better data quality.

Perhaps the biggest misconception is that AI only consumes data. Good AI also improves it. Pattern recognition allows AI to identify inconsistencies that traditional software simply ignores, including duplicated suppliers, inconsistent part numbering, incorrect lead times, abnormal batch sizes, missing planning parameters, disconnected master records and unusual ordering behaviour.

Instead of failing because the data is imperfect, AI identifies patterns and intelligently repairs many of those issues. The AI becomes part of the data cleansing process.

This distinction matters. Many AI products simply consume whatever data they are given. If the underlying data is poor, their outputs are poor. Other AI platforms, including O8 Organic Planner, have been deliberately designed to recognise data inconsistencies, derive planning parameters from operational history, and compensate for many of the weaknesses found in traditional ERP master data.

When people say, “Our data isn’t good enough for AI,” the better question is: “Is your AI capable of working with enterprise reality?”

The real objective is not perfect data. It is better decisions. Supply chains have always operated with imperfect information. Experienced planners instinctively compensate for poor data every day. Artificial intelligence should do exactly the same.

The goal is not perfection. The goal is making consistently better decisions than today’s manual process. If AI can reduce inventory, improve service and eliminate thousands of hours of manual planning while working with imperfect data, should the project really wait another two years for a master data programme? Probably not.

One of the unexpected benefits of AI is that data quality stops being a one-off project. Instead, it becomes a continuous learning process. Every order. Every receipt. Every shipment. Every supplier interaction. Every planning cycle.

The AI continually refines its understanding of the supply chain. Instead of maintaining static assumptions, the planning model evolves with the business itself.

Perhaps organisations are asking the wrong question. Instead of asking, “Is our data clean enough for AI?” they should ask, “Is our AI intelligent enough to work with our data?” Those are fundamentally different questions. One delays progress. The other accelerates it.

At O8, we never expected enterprise data to be perfect. In fact, we designed our AI and machine learning tools on the assumption that it would not be. Organic Planner uses pattern recognition and operational history to derive critical planning parameters, identify inconsistencies and improve decision quality without depending on static master data alone.

The result is an AI platform that learns from reality rather than blindly trusting outdated assumptions. Because in supply chain planning, the objective is not perfect data. It is better decisions. And the sooner businesses recognise that, the sooner they can start realising the value AI was always meant to deliver.