AI Intake for Personal Injury Firms: What to Automate and What to Keep Human

An injured person or a family member may contact several firms while trying to understand what to do next. A firm that responds well has an advantage over one that leaves the caller wondering whether anyone received the message. Responding well means listening, explaining the next step and avoiding promises nobody has reviewed.

A fast answer is not the same as a useful answer. If a caller repeats painful details to several people, gets an uncertain callback promise or hears an automated case prediction, the intake process has failed even though the phone was answered. Give callers a clear route to a person and carry the collected information through that handoff.

Define response expectations by availability. During office hours, the assistant might offer an approved consultation slot or transfer to intake staff. After hours, it might gather a limited set of facts and explain when someone will review them. Do not promise a same-day attorney review unless the firm has the staffing and process to deliver it.

An AI receptionist should support that process, not hide a staffing gap. Check what the caller hears when a transfer fails, when a consultation slot disappears and when an inquiry needs immediate human attention. Those moments reveal more than a smooth demonstration of a routine call.

Start with tasks your intake team already performs from written instructions. Keep the questions plain, explain why you need the information and avoid collecting documents or sensitive details that are not necessary for the first review. Counsel should approve the wording, storage and handling of the answers.

• Answering around the clock, identifying the firm and explaining that the caller is speaking with a virtual assistant.
• Collecting what happened, the date and location, the injuries described and treatment received so far.
• Asking whether there is a police or incident report, which insurance is involved and whether the caller already has a lawyer.
• Screening against the firm's written case criteria to route an inquiry for review, without declaring that it is accepted or has no merit.
• Booking an approved consultation and telling the caller which documents the firm wants them to bring.
• Following up on incomplete intakes and sending agreed reminders through the firm's reviewed contact process.

Let callers say they do not know an answer. An unknown incident date or missing report is information for the reviewer, not a reason for the assistant to invent a value. Preserve uncertainty in the intake record. Distinguish what the caller said from what the system inferred, and do not treat a generated summary as a verified fact.

Use booking rules that reflect the real schedule. Define which consultation types can be booked directly, whether staff confirmation is needed and how cancellations are handled. Reminders should tell callers how to reach the office if they cannot attend, rather than leaving them to reply to an unmonitored number.

If someone cannot complete intake, save only what your approved policy permits and create an assigned follow-up task. Do not repeatedly restart the entire questionnaire. The next person should see which questions remain unanswered and which contact preferences the caller gave.

An attorney or an authorized person under the firm's supervision should decide whether the firm takes a case. Automated routing can organize inquiries, but the firm remains responsible for review. Keep a path for exceptions so an unusual inquiry does not disappear because it falls outside a preset category.

• Anything that sounds like legal advice, including deadlines, likely compensation, liability or what the caller should say to an insurer.
• Decisions about accepting or declining representation and the reasons communicated to the caller.
• Fee agreements, explanations of their terms and signing procedures reviewed by the firm.
• Conflict check decisions, including how possible matches and incomplete names are resolved.
• Sensitive calls involving a death in the family, catastrophic injury or a distressed person who needs a prompt human response.

Write a short route for sensitive calls instead of requiring the caller to finish every question first. The assistant can acknowledge the request, collect a callback number when appropriate and connect the person to the designated staff member. It should not imitate a therapist, offer medical advice or pressure someone to describe traumatic details.

Have counsel define what staff can say when a caller describes an immediate safety concern. Intake is not an emergency response service. Approved instructions should direct people to the appropriate authority when necessary, and the system should avoid delaying that step while gathering business information.

Have the assistant identify itself as a virtual assistant at the start. It must never give legal advice or predict an outcome. Have counsel approve language making clear that no attorney-client relationship exists until the firm agrees in writing. That message is a boundary for the process, not permission to handle a prospective client's information carelessly.

Treat what callers share as confidential and collect only what intake needs. Decide who can access records, how long they are kept, how exports are handled and how incorrect information is corrected. Ask the provider to document where call information goes and how its AI providers handle it. Do not upload a real caller's information just to experiment with a demonstration.

Define handoff triggers in writing: requests for advice, questions about representation, possible conflicts, sensitive situations and anything the assistant cannot understand. Include a fallback when the intended person is unavailable. An unanswered transfer should create a visible follow-up task, not a silent end to the intake.

Test interruptions, unclear names, callers who decline a question and requests to speak with an attorney immediately. Review the resulting record as well as the conversation. A calm response is not enough if the summary changes the caller's meaning or marks an incomplete intake as complete.

Ask counsel to review ABA Formal Opinion 512 on generative AI tools, issued July 29, 2024. It addresses duties such as competence, confidentiality, communication and supervision when lawyers use AI tools. Duties to prospective clients also apply to information shared during intake. States issue their own guidance, so check yours.

Bring a concrete workflow to that discussion instead of a general question about whether AI is allowed. Show the opening disclosure, intake questions, handoff rules, record storage and follow-up messages. Let counsel identify changes required for your practice and the jurisdictions in which your lawyers work.

• The exact statements the caller hears before providing information and before a consultation is booked.
• The providers that receive intake information, the written data terms and the people who can access records.
• The points where staff review a possible conflict, decide on representation or respond to a request for advice.
• The process for reviewing call recordings, correcting errors and dealing with a caller's request about their information.

Keep approved versions of the script and messages. Assign responsibility for changes and test them before release. If your practice or provider changes, return to counsel with the changed workflow. This guide does not determine which rules apply to your firm or replace a review of your own circumstances.

Only contact people who reached out to the firm or gave consent. Include an opt-out and have counsel review outreach, because lawyer advertising and solicitation rules limit how firms can contact potential clients. An intake record is not permission to add someone to an unrelated campaign.

Make each message serve a specific intake purpose: confirming a requested consultation, asking for a missing detail or reminding someone about documents. Identify the firm, avoid discussing sensitive facts in a message somebody else might see and record the person's preferred channel. Stop the automated sequence when they opt out, ask for a person or say they already have counsel.

A voice AI agent needs approved rules for who may be contacted, what it may say and when a person takes over. Have counsel review consent, opt-outs and the actual call script before use. Review unanswered attempts and replies so automation does not keep contacting somebody whose situation changed.

Keep a clear record of the request or consent that supports contact. Limit access to that record and keep it with the intake according to the firm's policy. When contact preferences are unclear, send the question to staff rather than letting the assistant assume permission.

Track response time, intake completion rate, consultations booked and qualified cases that came from after-hours inquiries. Define each measure before comparing periods. An answered call, a completed questionnaire and a qualified case are different events, and qualification should reflect your firm's human review.

Separate abandoned intakes from inquiries your firm could not serve. Review why callers stopped: confusing questions, failed transfers, unavailable appointments or a request for a person. Check a sample of records against the conversation under your approved privacy process. Do not count duplicate calls as separate prospective clients.

Review the handoff queue with intake staff. Every pending inquiry should have an owner and a next action. If automation answers well but nobody reviews the records, it has moved the bottleneck rather than solved it. Your personal injury intake workflow should make that responsibility visible.