How to Explain Every AI Decision to Regulators

Jul 28, 2026

👋 Hi, I am Mark. I am a strategic futurist and innovation keynote speaker. I advise governments and enterprises on emerging technologies such as AI or the metaverse. My subscribers receive a free daily newsletter on cutting-edge technology.

When a regulator asks why your AI approved that loan, declined that claim, or flagged that patient, you need three things: a logged decision trail showing what the model saw and what it decided, an explainable model where you can articulate why it made that decision, and documentation created before the inquiry showing that you tested the system before it went live. Organizations building this now treat regulation as routine. Those who are not will face crisis.

A decision trail means tracking every AI decision with the inputs the model used, the decision the model made, and the confidence level. If a regulator asks about a specific loan decision, you can show: this was the applicant's data, the model processed it, the model scored a 78 approval probability, a human reviewed it, a human approved it, and it went live. Most organizations do not track this. Starting now is not expensive. It is building the habit.

An explainable model does not mean you use only linear models. It means you understand what features the model is using to make decisions. For a loan approval model, which variables matter most: income, credit history, debt ratio, employment tenure? Document the top 10 features that drive decisions. If the model is a black box, you are exposed. Dr. Mark van Rijmenam advises: if you cannot explain a model's decision, it should not make that decision.

Documentation means writing down your testing protocol before you deployed the system. What data did you use to train it? What performance metrics did you measure? What demographic groups did you test? What edge cases did you check? Did you test for bias? What was the result? When a regulator asks, you produce documentation created months earlier. You do not create retrospective stories. That is obviously defensive. Prospective documentation is credible.

Build this capability now. Start with your most critical AI systems. Document their testing. Create decision trails. Explain their logic. Extend to all systems over the next 90 days. This becomes your operating norm. When regulation arrives, you are not scrambling to figure out what you did. You have documented it.

Build governance that satisfies regulatory scrutiny. Visit https://www.thedigitalspeaker.com/intelligence-age-scorecard/

About Dr. Mark van Rijmenam: Dr. Mark van Rijmenam is a world-leading strategic futurist and the creator of the Intelligence Age Scorecard, a diagnostic assessment built on the WAVE framework from his book Now What? How to Ride the Tsunami of Change. He helps Fortune 500 companies and governments navigate AI and emerging technologies across five continents.

This article was created with AI assistance and reflects the WAVE framework methodology. For the full research-backed analysis, take the Intelligence Age Scorecard.

Frequently asked questions

What three things do you need to explain an AI decision to regulators?

You need a logged decision trail showing what the model saw and decided, an explainable model where you can articulate why it made that decision, and documentation created before any inquiry showing the system was tested before it went live. Organizations that build this now treat regulation as routine, while those that do not will face crisis when scrutiny arrives.

Link to this question

What is a decision trail in AI governance?

A decision trail tracks every AI decision along with the inputs the model used, the decision it made, and its confidence level. For example, it would show an applicant's data, the model's score such as a 78 approval probability, whether a human reviewed and approved it, and that it went live. Most organizations do not track this yet, but starting is not expensive, just a habit to build.

Link to this question

Does an explainable AI model mean only using simple linear models?

No, an explainable model does not require using only linear models. It means understanding which features the model relies on to make decisions, such as income, credit history, debt ratio, or employment tenure for a loan model, and documenting the top ten features driving decisions. If a model is a black box, an organization is exposed, and if a decision cannot be explained, that model should not be making it.

Link to this question

Why does documentation need to be created before, not after, deployment?

Documentation should record the testing protocol before a system was deployed, including the training data used, performance metrics measured, demographic groups tested, edge cases checked, and bias testing results. When a regulator asks, producing documentation created months earlier is credible, whereas creating retrospective stories after the fact looks obviously defensive and undermines trust.

Link to this question