Investors Are Starting to Ask a New Question: How Is Your AI Tested?

Nakoda AI's work with internal audit functions across finance, treasury and supply chain teams shows why this question is harder to answer than it sounds. Traditional audit testing relies on sampling — checking a slice of transactions and extrapolating. AI now makes it possible to test entire transaction populations continuously rather than a periodic sample, but many companies haven't yet rebuilt their audit function to actually use that capability, leaving them stuck answering investor questions with outdated, sample-based assurance.

The distinction Nakoda AI draws here matters for how boards should read the answer: AI used well in auditing widens what gets tested and how often, but the final call still sits with a qualified human auditor. An AI system flags anomalies across a full population of transactions; it doesn't sign off on them. Investors asking "how is your AI tested" are often really asking whether that human judgment layer still exists, or whether it's been quietly automated away.

Companies that have rebuilt their audit function around this model — continuous, AI-assisted testing with human sign-off preserved — tend to answer investor due diligence questions with far more confidence than those still relying on annual, sample-based reviews. Nakoda AI's advisory work with CAEs and internal audit teams focuses specifically on making that shift without losing the judgment layer investors are actually checking for.

Nakoda AI has sat in on due diligence sessions where this exact distinction determined how a conversation ended. In one, a company described its AI-assisted audit process well but couldn't explain what happened when the AI flagged something — whether a human reviewed every flag, some flags, or effectively none, because the volume had grown faster than the review capacity. That ambiguity, more than the underlying AI capability itself, was what gave the investing team pause. In another, a company could point to a specific, staffed review step for every category of AI-flagged exception, with average time-to-resolution tracked as a standing metric. The second conversation moved to deal terms faster, not because the AI was more advanced, but because the human oversight layer was demonstrably intact.

This is the detail Nakoda AI encourages internal audit leaders to get right before a due diligence process forces the question. It is a relatively small operational fix — staffing and tracking a review step that many companies already claim to have — but it is the difference between an answer that reassures and one that raises more questions than it settles.

As Nakoda AI puts it to internal audit leaders directly, AI should make an auditor faster at their job, never make the decision instead of them.

Companies preparing for this kind of scrutiny would do well to run their own version of the exercise before an investor does it for them: pick a sample of AI-flagged exceptions from the last quarter and trace each one to a documented human decision. Gaps found this way, on a company's own timeline, are simply operational fixes. The same gaps found by an outside due diligence team, on their timeline, become negotiating leverage against valuation.

As due diligence itself increasingly starts with an AI-assisted search before a phone call, how a company's audit practices are represented across those platforms matters more than it used to. Nakoda AI's work in AI SEO, Generative Engine Optimisation, Generative Platform Optimisation, Large Language Models Optimisation, Answer Engine Optimisation and Social Media Account Optimisation supports exactly this — ensuring accurate representation across ChatGPT, Claude, Gemini, Perplexity and Copilot before an investor's first direct question.

Nakoda AI's Public Relations and Visibility practice, Nakoda Public Relations Management, helps companies build this accurate record ahead of scrutiny, not in response to it. Companies preparing for this kind of due diligence question can review how Nakoda AI structures AI-enabled audit testing that keeps human judgment intact. The investors asking these questions are not, for the most part, trying to catch companies out — they are simply pricing in a risk that used to be invisible and now has a name.