Most AI tools give you a paragraph and ask you to trust it. Vera gives you a paragraph where every sentence carries a label telling you how much to trust it.
Grounded
A grounded sentence is backed directly by source. It links to the exact code line or document page it came from, so you can click through and check it yourself. If a change board asks where a claim came from, the answer is one click away.
Inferred
An inferred sentence is Vera's reasoning from what the source shows. For example: "Retiring this job would likely leave holds unreleased until someone releases them by hand." That conclusion follows from the code, but no single line states it. We mark it so you know it's worth a second look.
Gap
A gap means the source is silent. No document explains why a batch job exists. No code shows who owns an interface. Vera says so plainly, instead of making up something plausible.
We think this is the most important label of the three. A confident wrong answer is worse than no answer, because people act on it. A declared gap becomes a to-do list.
And when sources disagree
Sometimes the document says one thing and the code does another: a policy says holds release after 48 hours, the code releases after 72. Vera flags that as a disagreement and shows both sources side by side. Somebody needs to decide which one is right. Now they know they have to.
Why it matters
- Architects can sign off on evidence, not on trust.
- Auditors and risk teams can trace any claim back to source.
- Other AI tools that use Vera's model inherit the same labels, so they stop guessing too.