Skip to content
be-human.ai
← Back to the journal

AI by industry

Superintelligent Agents for Family Law: Intake, Court Dates, Documents

An AI agent can handle family law intake forms, court date reminders and document requests, with conflict checks and advice confidentiality as hard limits.

Be Human typographic cover with the title: Superintelligent Agents for Family Law: Intake, Court Dates, Documents
Conceptual illustration created with AI · Be Human

An AI intake agent for a family law firm can collect structured facts from prospective clients, send the right forms, remind clients of confirmed court dates and request documents in a fixed order. It should never decide whether the firm can take a matter, interpret the law for a client or see privileged material beyond the role it was given. The practical design is a narrow agent with written limits, a conflict check that a named person must clear, and confidentiality rules enforced by system permissions rather than by a policy document.

Key takeaways

  • Start with intake forms, reminders and document lists. These tasks are repetitive and checkable, and they stay low risk when the agent is barred from giving legal opinions.
  • Treat conflict checks as a human decision backed by a complete list of named parties, never as an automated yes or no.
  • Keep advice and case files outside the agent's default access. Each step should see only the fields it needs.
  • Build court date reminders from dates your staff has verified, and show the source and confirmation time in every reminder.
  • Measure the agent by complete intakes, missing documents and reminder accuracy, and review those numbers with a lawyer every month.

Why family law intake needs a narrower agent than a general chatbot

Family law intake is where a firm first learns who is involved, what has already happened and which deadlines may be close. Those facts carry conflict and confidentiality consequences the moment they arrive. A general assistant that answers questions freely is the wrong tool for that first contact, because it will eventually say something that sounds like advice. A narrow agent with a written role, fixed fields and a human handoff is easier to test, easier to audit and easier to explain to the lawyer who signs the engagement.

The useful split is simple. The agent collects, organizes, reminds and requests. Lawyers and trained staff decide whether the firm can act, what the facts mean, what the court expects and what the client should do next. Every workflow in this article follows that split. If a task requires judgment about the law, it leaves the agent's queue and goes to a person.

Intake forms: what the agent may ask and what it must not

The intake form should be short, structured and identical for every prospective client, so the team can compare submissions and the conflict step has clean data to work with. A workable first set of fields includes:

  • Full name, date of birth, contact details and preferred language
  • Names of the other party and of any children, with each person's relationship to the client
  • Any former names or spellings the client has used
  • Whether a court case already exists, and the court or county named on any papers the client holds
  • Hearing dates the client already knows about, recorded as stated by the client and not yet confirmed
  • Which documents the client can provide now and which they cannot
  • Preferred consultation times and any accessibility needs

The hard limits matter as much as the fields. The agent does not ask the client to explain the merits of the dispute, does not say whether the client has a strong position and does not estimate outcomes or the cost of litigation. Long narratives belong in a human intake call or in a restricted notes field that only the assigned team can open. If the client describes an immediate safety risk, the agent stops the form, shows a fixed script with emergency contact information and alerts a named staff member at once. The script is written by the firm, not generated on the spot.

A practical rollout follows a short sequence:

  • Write a one-page scope for the intake agent that lists what it asks, what it stores and what it never says.
  • Have a lawyer and an intake lead approve the field list, then freeze it for the trial period.
  • Install the fixed safety script and name the urgent contact before any live client traffic reaches the agent.
  • Route every completed form to a staff queue, not to a client-facing summary.
  • Read a sample of finished intakes each week and record where clients were confused by a question.

Conflict checks are a stop, not a suggestion

A conflict check in a family matter must cover more people than the prospective client. Spouses, former partners, children, new partners, employers in some disputes and related businesses can all create a conflict. The agent's job is to assemble a complete list of named parties from the intake form and send it to the person who runs the firm's conflict process. It must never mark a matter as cleared, and it must never open the next stage of intake because the party list looks finished.

Set the rule in writing. Decide which intake stages can run before a conflict result is recorded, and which must wait. Many firms collect only contact details and the party list at the first stage and hold detailed questions until a named person records the result. Check that threshold against your jurisdiction's professional conduct rules and your own policy. The agent should enforce the pause by not displaying the later form sections, not by asking the client to wait while the detailed questions are already on the screen.

Names are the usual weak point. Clients use nicknames, former surnames and transliterated spellings, and they sometimes forget a relative until the second conversation. The agent should ask for alternative names explicitly, store each one exactly as the client wrote it, and pass all of them forward. The human reviewer decides which records match, using dates of birth and the firm's own files.

Illustrative scenario: a prospective client lists a spouse under a married name and mentions a former partner's first name only. The intake agent records both names, the children's names and the first name as written, holds the detailed questions and sends the party list to the conflict reviewer. The reviewer finds a former client with a similar name, confirms from date of birth in the firm's records that it is a different person, records the clearance and the agent resumes the form.

Court date reminders that do not rely on memory

Reminders are only useful when the date is right and the source is visible. The agent should send reminders only from hearing details that a staff member has entered or checked against the court's notice. It should never send a reminder based on a date a client typed into a chat window. Each record keeps the source, the time it was last confirmed and the name of the person who confirmed it. When a record is older than the firm's refresh rule, the reminder is held and the paralegal is asked to check it.

Every client reminder should include the hearing type as recorded, the local time, whether attendance is in person or remote, what to bring and who to call with questions. The agent does not describe what will happen in the courtroom or what the client should say. It can point to the firm's own preparation guide, written and approved by a lawyer, and nothing more. Changes deserve their own process, because a corrected time that reaches the client without review causes more trouble than a late reminder.

Illustrative scenario: a paralegal updates a hearing start time after the clerk's office calls. The agent holds the old reminder and places the corrected time in the staff queue. The paralegal approves the change, and the client receives one message that names the new time and the date it was confirmed. Nothing is sent from the chat history alone.

Document requests and the paper trail

Document requests are the most repetitive part of a family matter and the easiest to lose track of. The agent should build a checklist from the matter type and the intake answers, then track each item as requested, received, unreadable or not available. Follow-up messages go out on a schedule a lawyer has approved. Each morning, the paralegal receives a short summary of items still missing and the reasons recorded so far. The pattern is similar to the document-request work described in AI for CPA Firms: Document Requests and Filing Checklists, where the checklist does the chasing so people can do the review.

Four rules keep this work safe:

  • Ask for documents only through the secure upload area, never by asking clients to email or message them in the chat.
  • Ask only for what the checklist lists. Decline unrelated personal records with a neutral message that points back to the list.
  • Store every upload in the matter folder with the same access rules as the case file, not in the intake inbox.
  • Keep the request list separate from any interpretation. Saying bring the bank statements for the listed period is a checklist instruction. Saying the statements show a certain position is legal analysis, and the agent does not produce it.

Documents can contain instructions too. A client might paste a message that tells the agent to ignore its rules or to reveal another client's status. Treat anything a client writes or uploads as data to be read, not as a command. The OWASP guidance on prompt injection prevention describes why this boundary has to be designed into the system rather than hoped for in the prompt.

Confidentiality and access are enforced by the system

Confidentiality should live in permissions. Give the agent the smallest set of tools for each step. The intake agent can read the intake form and write to the staff queue. It cannot open the case file, the billing records or the firm's email archive. The OWASP authorization guidance is a useful reference for writing these checks as explicit rules, so that access is decided on every request rather than assumed from where the agent sits.

Use this decision guide before adding any new step:

  • Does this step need the content, or only the fact that a step is complete?
  • Would a non-lawyer staff member reasonably see this under the firm's access rules?
  • Does the output leave the firm's systems, for example through a text message or an outside tool?
  • Does a lawyer need to approve this before the client sees it?

If the answer to the last question is yes, the step waits in a queue. Check vendor terms as well. Know whether client content is retained, where it is processed, who can read logs and how deletion works after a matter closes. Keep an access log so the firm can answer later who viewed a record and when. The NIST AI Risk Management Framework offers a structure for documenting these risks and the controls that address them before the system takes live traffic.

Running intake, reminders and documents as a managed service

Be Human designs personalized AI infrastructure for businesses, and a family law office would normally begin with one role: an intake and document coordinator. The first engagement defines that role, the memory it may hold and the tools it may use, the approval boundaries and what the agent must never say. It also sets one observable success criterion, such as the number of intakes that reach a lawyer complete without a second call, so the firm can judge the agent on evidence instead of impressions.

After launch, managed operations cover the work that keeps the role safe over time. That includes monitoring the staff queues, running recurring evaluations against a fixed set of sample intakes, reviewing permissions when staff change roles, maintaining the connection to the docket source and document storage, and responding to incidents such as a reminder sent with the wrong time. Human approvals stay in place throughout. The managed service makes those approvals easier to run and easier to audit, but it does not replace them.

The free Be Human AI diagnostic can help you find the first task worth handing to an agent in your practice.

Frequently asked questions

Can an AI intake agent tell a prospective client whether they have a case? No, and it should be designed so that it cannot. The agent collects structured facts, explains what the firm needs next and passes everything to a lawyer or trained staff member. Any question about merits, likely outcomes or what the law requires belongs to a person who can give advice under the firm's engagement terms.

Does a reminder from an agent replace the firm's calendar? No. The firm's verified calendar remains the record of truth, and the agent sends reminders from entries that a staff member has confirmed. Each reminder should show its source and the time it was checked, so a discrepancy can be traced quickly and corrected before a client misses a hearing.

How should a family law firm handle conflict checks when an agent is collecting intake data? The agent should compile the complete list of named parties, including former names and children, and send it to the person who runs conflict checks. It must not mark anything as cleared, and it should hold detailed intake stages until the firm's rule allows them to proceed. Confirm the timing of that pause against your jurisdiction's professional conduct rules.

Where should client documents be uploaded? Documents should go only through a secure upload area that is connected to the matter folder, with the same access rules as the case file. Clients should never send records through chat or ordinary email to the agent. Uploads that fall outside the checklist should be declined with a neutral message, and the vendor's storage, retention and deletion terms should be reviewed before go-live.

Your next step

See where an AI agent fits in your company first.

Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →