Is it current?
Information has an expiration date. Yesterday's inventory count might be fine for a report and dangerous for an order.
The definition
Context-Verified Decision Intelligence (CVDI) is an approach to AI-supported decisions in which an action may proceed only if the context behind it is verified at the moment of action: current, traceable to its source, and reliable enough for the stakes. The verification happens outside the AI model, and both permitted and stopped actions are recorded.
In plain English: before AI is allowed to act, the system checks that the information behind the decision is fresh, comes from a real source, and is strong enough to trust. It also keeps a record of what it allowed and what it stopped.
Why it matters
Most of the AI conversation is about whether the model is smart enough. That's the wrong question once AI starts taking actions: sending the email, changing the record, placing the order. The question that matters is whether the AI knew what it needed to know at the moment it acted.
“A genius working off last month's numbers can still be wrong.” — Jimmy Standaert
The three questions CVDI asks before an action
Where did it come from?
If you can't say where a fact came from, you can't say whether to trust it. “The AI said so” is not a source.
Is it strong enough for the stakes?
“Probably right” is fine for a draft email. It's not fine for a wire transfer.
Two principles sit underneath those questions. The check lives outside the AI, so the model doesn't get to grade its own exam. And the stopped decisions are recorded too, not just the ones that went through, because that second list is where you find out whether your controls actually work.
What CVDI is not
| Common approach | What it checks | What it misses |
|---|---|---|
| Guardrails | What the AI says or does | Whether the information it acted on was any good |
| Outcome-only evaluation | Whether things turned out well | Whether it was a good decision or a lucky one |
| Post-hoc audit trails | What happened, afterward | Stopping the bad action before it happens |
| Model self-judgment | Whether the AI thinks it's right | Everything. It will almost always say yes, confidently. |
CVDI doesn't replace any of these. Guardrails and audit trails are still needed. CVDI adds the check they skip: what the AI knew, at the moment it acted.
Where the term comes from
Jimmy Standaert, Fractional CAIO and founder of RiGi Group and WhiteWater.ai, introduced the term to name the category WWAIOS™ (the WhiteWater AI Operating System) is being built around. It's part of the WhiteWater.ai Field Guide, a living glossary of the terms behind accountable AI, written for people who don't speak AI.
Use the term. Argue with it. Cite it as: Standaert, J. (2026, October 30). Context-Verified Decision Intelligence (CVDI). WhiteWater.ai. https://whitewater.ai/cvdi/
Download the 2-page CVDI definition (PDF)
Want to know whether your AI workflows check what the AI knew? Talk to Jimmy.