Data readiness

AI-ready data: the five tests

Most boards assume their data is ready for AI. It is not. Five simple tests show you exactly where the gaps are, and they take an afternoon, not a year.

By Zakir Hoosen, Director, Fuzzelogic Solutions. Board-level guidance in plain English.

Your data is either ready for AI or it is not. There is no halfway. A board that assumes the first without checking the second spends money on projects that look brilliant in a slide and fall apart at the first handoff between teams. This guide gives you five tests. Run them and you will know more in an afternoon than most organisations discover in a year of pilots.

Why data is the problem, not the technology

The technology works. That is no longer the debate. The question is whether the data feeding it is in a state that allows machines to learn from it, act on it, and be held accountable for what they do with it. Most boards skip this step because data feels like a plumbing issue. It is not. It is a governance issue. If your data is wrong, your AI will be wrong, and someone will have to explain that to a regulator, a customer, or a court.

Nearly two thirds of organisations either know they are not ready or do not know whether they are. Both positions are dangerous. The first is a known problem. The second is an unknown one. A board that does not ask the question cannot manage the answer.

The five tests

Fuzzelogic defines AI-ready data as data that passes five tests. Each one is simple to state and hard to fake. Run them against any dataset you plan to feed into a system that will make decisions.

1. Reachable. Can the data actually be found when the system needs it?
2. Trustworthy. Do you know it is accurate, current, and complete?
3. Explainable. Can someone trace where it came from and why it says what it says?
4. Changeable. Can it be updated when the business changes without breaking the system?
5. Governed. Has someone decided who may access it, who may change it, and who answers when it goes wrong?

These five are not optional. If even one fails, the system that depends on it will fail too. The question is not which ones you can skip. The question is whether you can pass all five, today, for the data you plan to use.

Reachable

Reachable means the data exists and a system can get to it. Sounds obvious. It is not. I have seen organisations that know exactly where the data lives, but it sits behind a process that takes three people and a spreadsheet to unlock. If the system needs the data in real time and the data requires a human handoff, it is not reachable. It is available, but it is not ready.

The test is simple. Ask the team: if the system needed this data right now, how would it get it? If the answer involves a person, a manual step, or a delay, you have found a gap.

Trustworthy

Trustworthy means the data is accurate, current, and complete enough to act on. Incomplete data does not just give you partial answers. It gives you confident wrong answers. A system trained on outdated customer records does not know they are outdated. It treats them as fact.

Two thirds of the companies that are already ahead of the pack say data is the thing holding them back. That is not a technology problem. It is a trust problem. The data exists. It just cannot be relied on.

The test: can you put a number on how current this data is? If the answer is "we think it is recent" or "it gets updated sometimes," you are not ready.

Explainable

Explainable means someone can tell you where the data came from, how it was collected, and why it says what it says. This matters for two reasons. First, accountability. If a system makes a decision based on data, and someone asks why, "the computer said so" is not an answer that holds up in any boardroom, court, or regulator's office. Second, accuracy. If you cannot trace the data, you cannot check it.

The test: ask the team to show you the source of one specific piece of data the system uses. If they can point to a system, a record, and a process, you are in good shape. If they point to a spreadsheet, an export, or a shrug, you have work to do.

Changeable

Changeable means the data can be updated, corrected, or replaced without breaking the system that uses it. Businesses change. Customer records change. Regulations change. If the data structure is so rigid that a change in one place causes errors in ten others, the system is not ready for the real world.

The test: ask what happens if a key field changes. If the answer is "we would need to rebuild the pipeline," you have a fragility problem. If the answer is "we update the record and move on," you are closer to ready.

Governed

Governed means someone has decided who owns the data, who may access it, who may change it, and who answers when something goes wrong. Without governance, every other test is temporary. You can fix the data today and have it drift next month because no one owns the process.

One in five organisations has no governance at all. That means one in five organisations has no answer to the question "who is responsible when this goes wrong?" For a board, that is not a gap. It is a liability.

The test: name the person who owns this data. If you get a name, you have governance. If you get a department, you have a start. If you get silence, you have a problem.

What the tests reveal

You already have AI in your business. You just do not know where. Once you run these five tests, you will see it clearly. The system your finance team uses to forecast. The tool your marketing team uses to segment customers. The reporting your operations team relies on. Some of it is ready. Some of it is not. The tests tell you which is which.

Eighteen trillion dollars of value is locked inside data that cannot be used as it is. The data exists. The value is real. But it sits behind systems that were not built for machines to use. The five tests show you where the locks are. Fuzzelogic helps you open them, or tells you honestly which ones should stay closed.

The honest answer

Your systems were built for a world before AI. Most can get there. We tell you which ones cannot. If the honest answer is that AI should not touch a process, we put it in writing rather than build it anyway. That is what the tests are for. Not to sell you something. To tell you the truth.

Start with the assessment. Two to four weeks, fixed price, and you own the verdict and the roadmap whether or not we build any of it. When you are ready to talk AI, call Fuzzelogic Solutions and ask for Zak. www.FuzzelogicSolutions.com | info@FuzzelogicSolutions.com | +44 (0)1624 618950

For more on why pilots stall, read why 95% of AI pilots return nothing. For the full assessment process, read data readiness assessment: how to assess data for AI. The full library is on our index. Our site explains how Fuzzelogic approaches AI for business.