What the partner sees when the associate used AI.
This is the review screen a supervising attorney opens. An associate logged AI use against a matter; one entry came back flagged. Walk the five steps to sign it off, or send it back. Sample data, running in your browser.
The review queue.
Follow the prompts below the panel. Every action here is one a real reviewer takes, and the record it writes is append-only.
Sample matters and names are fictitious. The real screen runs on your firm's own logged entries. Candor stores the fact of each AI use, never the client text behind it.
Not that the AI was right, but that a lawyer took responsibility.
The flag is a prompt, not a verdict
Candor surfaces citations that may need a second look and the AI's own hedging. It doesn't verify accuracy or decide compliance; the attorney's sign-off carries the weight, and confirming the citation is still their job.
Every decision is dated and attributed
Who reviewed it, when, what they decided, and any note, kept as an append-only entry. A reversal references the original; nothing is overwritten.
This is the artifact for the carrier
When a malpractice carrier asks how AI is used at the firm, this loop (logged use, a flag, an attorney sign-off) is the record you hand them.