The assistant prepares. The lawyer approves. Why that is the right order
What AI actually does in a law firm, and what it should never do: the difference between an assistant that prepares a draft and a system that acts on its own, and why we chose the first.
There are many promises around AI in law firms, and most of them frighten exactly the people they are meant to convince. "The system files by itself", "works while you sleep", "handles all your messages". A lawyer reading that hears one thing: someone else is deciding in my name.
We built CaseFlow on the opposite principle.
What the smart assistant does
It does the grunt work, at your request. A court notice arrived? It proposes to file it into the case, add the hearing to the calendar and prepare a draft update for the client. Need a statement of claim? It prepares a draft from the template with the case data. End of the month? It prepares the charges from the time entries.
And it stops. Every result waits in the proposals inbox marked "Awaiting your approval". The lawyer reads, corrects if needed, and clicks "Approve". Only then is anything saved, sent, signed or charged.
Why the click matters
Because the click is the moment responsibility stays where it belongs. A lawyer signs what leaves the office under their name, and no system should take that away from them. Not a good system either. Not at two in the morning either.
There is a second, less discussed advantage: a lawyer who approves every action knows at every moment what is happening in their office. No surprises in the morning.
What it means in practice
- The assistant proposes. It does not execute.
- The firm decides what the assistant may prepare, and every action is written to the audit log.
- No proposal is approved by default. A proposal you did not approve simply waits.
It is less impressive in a slide deck than "an agent working on its own". It is a lot quieter at seven in the morning.