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

AI by industry

AI Intake for Family Law Firms: Questionnaires, Court Dates, Checklists

A plan for AI intake at family law firms: questionnaires, court date calendars and document checklists an agent can run, with attorneys approving every filing.

Be Human typographic cover with the title: AI Intake for Family Law Firms: Questionnaires, Court Dates, Checklists
Conceptual illustration created with AI · Be Human

AI intake for family law firms works best when an agent handles the repeatable parts of a new inquiry: collecting structured facts through a questionnaire, tracking court dates on a shared calendar, and assembling a document checklist for each matter. The attorney still decides whether to accept the matter, what advice to give, what to file and when. This article explains how to scope those tasks, where the approval lines belong, and what to verify before any agent touches a client.

Key takeaways

  • Limit the first agent to intake logistics: questionnaires, reminders, calendar entries and document lists. Legal judgment stays with attorneys.
  • Write the questionnaire from the attorney's own intake notes, then have the agent ask only the questions that the matter type requires.
  • Give every court date a single owner, a source document and a confirmation step before it reaches the calendar.
  • Keep the agent's permissions narrow: read client forms, write calendar drafts, send status messages from approved templates, and never file or sign anything.
  • Judge the agent by whether the attorney receives complete, accurate intake packets that need fewer corrections each week.

What an intake agent can handle and what it cannot

A family law practice receives a steady flow of inquiries that look similar on the surface but differ in urgency, jurisdiction and risk. Many of them arrive outside office hours, come from people in the middle of a crisis, and contain facts given out of order. An agent is useful in that gap because it can respond promptly, ask consistent questions, record what the prospective client said, and tell the staff what is missing.

The agent should not evaluate whether a person has a claim, estimate outcomes, recommend a strategy, or tell someone whether a deadline applies to them. Those are legal judgments. The line is simple to state and needs to be written down: the agent gathers, organizes, reminds and drafts, while attorneys decide, advise and sign. A useful test is to ask whether a wrong answer would change what the client does next. If it would, a licensed person reviews it before the client sees it.

Be Human's managed approach starts with that boundary. A first engagement defines one useful role, the memory and tools that role needs, the approvals required before any outbound action, and a success criterion that the firm can observe. For intake, the role is usually narrow: receive an inquiry, run the questionnaire, check the answers for completeness, and prepare a packet for the attorney or intake coordinator.

How to build the intake questionnaire

The questionnaire is the core artifact. A good one reflects how the attorney actually interviews a prospective client, not how a generic form looks. Start with the existing intake notes, the questions that staff ask on the first call, and the reasons past matters were delayed or lost. Then separate the questions into three groups: identity and contact details, matter facts, and urgency signals.

A practical sequence for building it:

  • Collect the last ten or so intake notes from matters the firm handled well and from matters that stalled. Mark what information was missing at the start.
  • Group the questions by matter type, such as divorce, parenting arrangements, support modification, or prenuptial agreements, so the agent asks only what is relevant.
  • Mark sensitive questions explicitly, such as questions about abuse, finances and children, and define how the agent should handle a client who does not want to answer them.
  • Add a safety rule that routes any message describing immediate danger to a human at once, using a pre-approved script that points to emergency services.
  • Write a short closing message that explains what happens next, who will review the information, and when a person will reply, without promising a result.
  • Test the questionnaire with staff who have never seen it, then compare the packet the agent produces with the packet the attorney would have requested.

The questionnaire should also avoid inviting the prospective client to over-share. Asking for every message exchanged with a former partner, for instance, creates a large privacy burden and rarely improves the first decision. Ask for the documents needed to assess the matter, and say what can wait until after an engagement.

Court date calendars the agent can maintain

Court dates are where small errors cause large problems. The safest pattern is to have the agent create calendar drafts from documents a human has already identified, never from a client's memory alone, and to require a named person to confirm each entry. Every entry should carry its source, such as the order, notice or docketing screen, and its owner.

What the agent can track:

  • Hearing dates, conferences and deadlines that a staff member has entered or confirmed, with reminders to the responsible attorney and paralegal.
  • Gaps where a matter has an active case but no next date recorded, so a person can check the docket.
  • Documents that must be filed or exchanged before a known date, marked as tasks for a person rather than actions the agent takes.
  • Changes such as continuances, rescheduled conferences or new counsel, flagged for review so the calendar does not silently drift from the court record.
  • Time zone and location details for hearings, especially for firms that cover multiple courts or handle remote appearances.

The agent should never calculate a legal deadline on its own and present it as settled. Deadlines depend on jurisdiction, rules, service dates and local practice. If the firm wants the agent to propose a deadline, the rule set should come from attorneys, be stored as a reviewed reference, and show its source so that a person can check it. The agent's job is to make the calendar easy to verify, not to decide the law.

Document checklists by matter type

A document checklist turns a vague request into a concrete list. Clients often do not know which records matter, and staff spend time chasing them by hand. An agent can produce a checklist for each matter type, then track what has arrived, what is incomplete and what is still outstanding.

Illustrative scenario: a prospective client writes that they are separating and want to understand the next steps. The agent asks the intake questions for the relevant matter type, notes that the client has children and a shared property, and produces a checklist covering identity documents, existing agreements, property records, income statements and a short timeline. It marks three items as received, two as partial, and one as missing. It sends the client a message with the missing items and explains that an attorney will review the packet. The attorney then decides which documents are actually needed and whether the matter is accepted.

Keep the checklist and the advice separate. The checklist says what to gather. It does not say what the documents prove, whether a court would accept them, or what the client should do with them. Label every checklist as prepared for attorney review, and store the version history so the team can see how the template changed.

For a firm that also handles process service or related filings, the status and proof-of-service records follow similar rules. The article on AI for Process Servers: Service Logs, Affidavit Drafts and Status Updates shows how a service log can stay accurate without the agent drafting sworn statements on its own. For firms with transcript or record requests tied to hearings, Superintelligence Prep for Court Reporting Firms: Transcript Orders covers a similar pattern of tracked orders and human confirmation.

Approval boundaries and attorney ownership

Every outbound action needs a defined owner. A helpful way to set this up is a permission table written in plain language. For each tool the agent can use, state what it may read, what it may draft, what it may send, and which human must approve each step. Then review the permissions on a schedule, because intake roles tend to expand quietly when staff ask for one more convenience.

The OWASP Authorization Cheat Sheet is a useful reference for this discipline. It describes checking access at each point where an action happens, rather than trusting that a user or system should be allowed everything once it is inside. For a family law agent, that means the agent should not be able to open a sealed file because it shares a folder with a permitted matter, and it should not send a client message that cites a document it was never authorized to read.

A reasonable approval model for intake:

  • The agent may read the questionnaire responses and the documents the client uploads through the approved channel.
  • The agent may create calendar drafts and checklist entries, which a named staff member confirms.
  • The agent may send status messages from attorney-approved templates, and every custom message goes to a person first.
  • The agent may never file, sign, accept an engagement, quote fees, or state a legal conclusion.
  • Any message that mentions a deadline, a safety concern or a conflict of interest goes to a human immediately.

Attorneys keep ownership of the file. The agent is a tool that prepares work and records what happened. When the agent makes a mistake, the firm needs a clear way to see the source of the error, correct it, and decide whether the template or permission should change.

Security, privacy and untrusted input

Family law intake contains some of the most sensitive information a person can share. The NIST AI Risk Management Framework offers a structure for thinking about these risks, organized around governing, mapping, measuring and managing AI systems. For an intake agent, the practical questions are concrete: what data is stored, who can see it, how long it is kept, and what happens if the agent is given false or hostile information.

Prompt injection is a particular concern. A prospective client, a third party or an uploaded document can contain text that tries to change the agent's instructions, such as a message telling it to ignore its rules or to send a file elsewhere. The OWASP Top 10 for LLM Applications and its prompt injection guidance both treat this as a design problem, not a matter of better wording. Keep system instructions separate from client content, treat all uploaded text as data, limit which tools the agent can call, and log every outbound action so that a person can review it.

Privacy rules depend on jurisdiction. Firms serving Brazilian clients should review the guidance from the Autoridade Nacional de Proteção de Dados, and firms working in the European Union should consider the obligations described in the EU Artificial Intelligence Act overview. Whatever the setting, the firm should document what the intake agent collects, why, where it is stored, and who may access it. Keep the retention period short for prospects who do not become clients, unless the firm has a clear reason to keep the record.

Starting the first month without overbuilding

The first month should be small enough to evaluate honestly. Choose one matter type, one intake channel and one checklist. Run the agent alongside the existing process for two or three weeks, comparing the packets it produces with the packets staff produce by hand. Look for missing questions, false urgency, unnecessary questions, and places where a human had to rewrite the agent's message.

Be Human's managed operations can then cover the ongoing work: monitoring the agent's outputs, running recurring evaluations against a fixed set of sample intakes, reviewing permissions, maintaining integrations with the firm's calendar and document tools, and handling incidents when something goes wrong. Improvements should be controlled. Each change to a questionnaire or checklist should be recorded, tested and approved by an attorney before it goes live.

For a firm that is not yet sure where to begin, the best first task is often the one that is repetitive, low in legal judgment, and easy to check. That is usually the intake packet or the court date reconciliation, not the client conversation itself. Take the free Be Human AI diagnostic to find the first task in your intake process that is worth handing to an agent.

Frequently asked questions

Can an AI intake agent tell a prospective client whether they have a case? No, and it should not be designed to try. The agent can ask structured questions, record the answers, and tell the client that an attorney will review the information. Any statement about the merits, likely outcomes or next legal steps belongs to a licensed attorney, so the agent's messages should stay within approved wording.

How does an agent keep court dates accurate without making legal decisions? The agent creates calendar drafts from documents that a person has identified, attaches the source of each date, and asks a named staff member to confirm it. It can flag gaps, such as an active matter with no next date, and remind the responsible team. It should not calculate deadlines as settled facts unless attorneys have reviewed and approved the rule behind the calculation.

What should a firm check before giving an intake agent access to client documents? The firm should confirm what data the agent can read, where it is stored, how long it is kept, and who can review its activity. It should also test the agent against prompt injection, because uploaded documents and messages can contain instructions meant to manipulate the system. Access should follow the principle of least privilege, with permissions reviewed on a regular schedule.

How long does it take to see whether an intake agent is useful? A useful first trial runs alongside the current process for a few weeks, with a fixed set of sample intakes and a clear success criterion such as fewer missing items in the attorney's packet. The firm should record every correction staff make and use those corrections to improve the questionnaire or checklist. If the packets do not get more complete or the corrections do not decrease, the scope should be narrowed or the approach changed.

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 →