Starter guide to agents

You may have tried a chatbot such as ChatGPT: you ask a question, it answers, and the next step is yours. An agent is given tools to act in addition to chat. It can take multiple turns on its own before handing back a finished result. You give it a goal, and it reads the file, works out what to do next, does it, and brings you the result with its sources.

The closest comparison is handing a file to a capable member of your staff. You say what you need and review what comes back, and the judgment calls remain yours.

The termsCopy Anchor Link

• Agent: an AI that works toward a goal in steps, choosing what to do next as it goes. Mona is the agent you work with directly. It knows the whole case, proposes what to do next, and does the work or delegates it.
• Subagents: specialists, each built for one job, that Mona can send. Trace checks one line of a statement against the evidence, and the Statement agent reviews a whole statement. You can start either one yourself.
• Background agents: agents that start on their own when you add a file. They sort and read what comes in, so Mona and its subagents have evidence to work from.
• Tools: the actions an agent can take, such as reading a document, searching the law, calculating support, or writing a work product. An agent can only do what its tools allow. Mona has no tool for changing the case's facts or for sending anything out of the firm, so those stay with you.

The agents lists each one and what it does.

What you can ask forCopy Anchor Link

Click Ask Mona and write as you would to someone on your staff:

• "Where does the case stand?" Mona writes a memo of the case, with sources, and lists what to do next.
• "Is anything missing from the other side's statement of net worth?" The Statement agent reads the whole statement and proposes lines it left out. Only you can add them.
• "Check this line against the evidence." Trace records what the evidence supports and cites its sources.
• "Draft a discovery demand for the statements we're missing." Mona names each account and the months missing from it.
• "Run a support analysis." Mona calculates support and cites every figure.

Why it helpsCopy Anchor Link

• It reads everything. People read what time allows. Agents read every page of every document and every transaction in every account, so a small transfer or a passing remark in an interview still gets read.
• It remembers the case. Mona writes what matters into the case memory: your client's objective, the findings, your decisions, and what comes next. A decision you record applies from then on.
• It shows its work. Every figure carries a footnote to its source. Click it and the page opens.
• It gives you back your hours. The reading and sorting move to the agents, so your time goes to strategy, clients, and negotiation.

Common questionsCopy Anchor Link

Can it get things wrong? Yes. You may have read about lawyers sanctioned for citing cases a chatbot invented. Mona reads the documents in your case, looks up law at its source, and lists the legal authorities still to confirm before a paper leaves the firm. Review its work before you rely on it, as you would anyone's.

Does it act on its own? Partly. Mona can read the case, write drafts, update the case memory, and start subagents without stopping to ask. Each step is recorded, and you can edit or stop it. It asks first before long or costly work. Trace's results reach the Net worth analysis as soon as it finishes, so review them there. Anything that leaves the firm, you send.

Who makes the legal call? You do. Mona will point out a transfer made two months before commencement. Whether it was dissipation is your call.

What about client confidentiality? Your case data is not used to train AI models, and each firm's data is walled off from every other firm's.

To try it, start with Your first case.