Design Principle: Reviewability

In audit, work only counts if someone else can check it.

When a partner signs an audit opinion, they are personally accountable for it, and that responsibility cannot be handed off. Before signing, the partner has to be satisfied that the evidence supports the conclusions, and they are expected to review the significant judgments the team made along the way. A manager does the same one level down. That review is how the profession assigns responsibility for the opinion. The documentation standards work the same way: a file has to be clear enough for another experienced auditor to pick it up and understand what was done.

We think any tool that produces audit work should be held to that same standard, and it matters even more for work produced by agents. A model can generate a testing conclusion. The harder part is making sure a reviewer can confirm that conclusion quickly enough to put their name behind it and place the work in their audit file.

That question is at the front of our minds. How do we make it effective and efficient for a practitioner to evaluate work produced by an agent and take responsibility for it in their file? Reviewability is the principle we are building Snap around because of it, from the start rather than as a layer added later. In practice, that means every testing line should carry its reasoning and a link to the source document it came from, so a reviewer can see how a conclusion was reached and check it against the evidence quickly.

We also think the review process itself can improve. Not every line needs the same scrutiny. Low-risk, high-volume testing can be summarized for a lighter review, while items that carry real risk are surfaced and flagged for closer attention. That would let a reviewer spend their time where it matters rather than spreading it evenly across everything.

To be clear about what we are aiming for: we are not trying to get partners to accept Snap's output on faith. We are trying to make verification fast, so the reviewer stays in control and can defend the file to anyone who asks, including an inspector.

If you are a partner or a manager who carries that responsibility, this is the standard we want to be held to. We would be glad to hear how you think about review, and what it would take to make a file you could confidently stand behind.