Answer
How do you show what an AI tool did with a client's information?
In most firms you cannot, because nothing was written down at the time. A run record fixes that by capturing metadata for every run — what kind of task it was, when, under which policy, with what outcome and who approved it — while never storing the content itself.
Why the honest answer is usually 'we would have to ask'
When AI use grows out of individuals being resourceful, there is no moment at which anyone decides to write it down. The work gets done, the draft goes out, and the only trace is in someone's chat history on their own machine.
That is not carelessness. Nothing in the process ever asked for a record, so nobody produced one.
What a run record holds
Counts, hashes, policy identities, timestamps, outcomes and opaque references. What kind of run it was, when it happened, which policy governed it, what the outcome was, and — where the workflow declares a review step — that a named person approved it.
It is metadata by architecture, not by convention. Several of the writers throw an error rather than persist anything content-shaped, which means the guarantee is in the code rather than in a policy document somebody could quietly change.
What it deliberately is not
It is not an archive of the documents. You cannot read a client's file out of it, which is the point.
It is not a certification, an audit, or evidence that a firm has met any particular obligation. It is a description of what the software did. What that is worth to you is your firm's judgement, and it remains your firm's responsibility.
FlowClave is a software workflow. This is not legal, privacy or compliance advice, and it does not make a firm compliant with anything. Your firm remains responsible for its own obligations.