AI for law firms. Client intake, prepared for review

AI for law firms is often discussed in terms of research and drafting. This article focuses on client intake: a task that takes time every week and involves confidential information from the start. An inquiry arrives through a web form, email or a message recorded by your team. Someone prepares the summary, conflict search, first meeting, engagement letter and document request. A model can help prepare that material, with a lawyer deciding at every step.

The model runs through a provider the firm chooses under a data processing agreement. The firm assesses what that arrangement means for its confidentiality obligations. This article explains the information it would need for that assessment.

The tools have different roles. An API integration connects the website, mailbox, calendar, practice management software and document templates. Robotic process automation, or RPA, fills gaps where a system has no API. An AI agent uses a language model to read free text: the inquiry, attachments and notes taken by a person during a call. The model reads and drafts. It does not decide.

Client intake. What the work costs today

An inquiry arrives with the story out of order and dates missing. Someone reads it, asks for missing facts, searches the client list for the other party and finds a meeting slot. They copy an engagement letter, change the names, request documents and file the thread.

Multiply that work by the inquiries received each week, including those you turn down, and intake becomes a part-time job that no one was hired for. The wait between the inquiry and the first human reply is also where prospective clients can go elsewhere.

AI for law firms. Intake task by task

Read the inquiry. Prepare the summary

A model would turn a web form, email or transcribed message into a structured summary using your format. It would list the parties, matter type, key dates, requested help and missing facts. The lawyer would read the summary first and the original second.

Missing facts. Questions for human approval

A model would use your intake questions to draft a reply in the firm’s wording. It would request the specific facts and documents needed for that type of matter. A person would approve the reply before sending it.

Conflict searches. Prepare the lawyer’s review

An API integration would search clients, former clients and adverse parties in your practice management software. The model would suggest name variants. The search results would go to the lawyer, who would perform the conflict check and make the decision.

First meetings. Calendar rules and reminders

A calendar integration would offer slots under the lawyer’s rules. It would send a confirmation and reminder, including the video link or address. A request to speak to someone first would go to the firm’s team.

Engagement documents. Your templates and terms

An integration would prepare the engagement letter from your approved template, with the fee terms you set. A model would help draft the document request list for that matter type. The lawyer would review, sign and send the documents.

Open the matter. Use information already collected

An API integration would create the client record, matter record, folder structure and first deadlines in your practice management software. It would use the information already collected, avoiding duplicate entry.

Status replies. Drafted from the matter notes

“What is happening with my case?” would produce a draft based on the matter file. The responsible person would read, correct and send it. Nothing would reach the client without that step.

The legal assessment, advice, decision to accept or decline a matter, and first conversation with a worried client stay with the lawyer. The model prepares the material.

Confidentiality and privilege. What the model would read

An intake inquiry is confidential before the person becomes a client, and it may be privileged once they do. The firm interprets its professional rules on confidentiality and privilege with its own adviser.

In a custom integration, the model would read the inquiry and notes supplied to it and propose a summary or draft. It would use a business provider chosen by the firm, under a data processing agreement that excludes training and identifies the storage region. The provider account would be the firm’s.

Every read would be logged so the firm could show what was processed, by which service and when. A consumer AI subscription on a lawyer’s laptop does not, by itself, establish those controls. Its terms can permit retention or use of submitted material for training.

Whether the custom setup meets your obligations of professional secrecy is a question for the firm and its regulatory adviser; we put the mechanism in writing so that the question can be answered.

You check this with your bar association and your own counsel.

The people who message the assistant are told, in the first reply, that they are dealing with an automated assistant, and how to reach a person.

Where AI is the wrong tool. Connect structured information directly

Booking meetings, opening matters and filling templates are structured tasks. An API integration can move information from a form, calendar or record without a model reading it.

The model has a role where the input is free text: the inquiry, attachments or notes taken during a call. If your intake already arrives as a completed form without attachments, an integration can handle it without AI.

How it could work. In your law firm

If your team prepares intake manually, today’s cost would include reading inquiries, collecting missing facts, preparing conflict searches and filling engagement templates.

Say you run a law firm handling family, property and small business matters. Inquiries arrive through the website, a shared mailbox and messages recorded by the front desk. The integration would collect them into one intake queue.

Each inquiry would have a summary, missing-facts list and search results from your records, ready for a lawyer to review. A click would send approved follow-up questions in your wording. Another would offer meeting slots.

Once the lawyer accepted the matter, the engagement letter and document request would be drafted from your templates. The matter would be opened in your practice management software using the information already collected.

Client status questions would arrive with a draft reply and the relevant notes attached. The model would use the provider the firm chose under a data processing agreement. Each week, the queue would show inquiries received, matters opened and the time taken for the first human reply.

Choose the setup. Compare the intake options

  • A practice management suite with AI features. This fits a firm already using one suite, satisfied with its intake module and comfortable with its data processing terms. It is quick to switch on, with per-user fees and limited control over the model.
  • A custom integration. This fits a firm with several tools, a particular intake process, or clients and a regulator asking where information goes. The provider is chosen by the firm under a data processing agreement. The workflow is yours, with no per-user fee for the custom integration.
  • Change nothing yet. This fits a sole practitioner with few inquiries, where the person reading them is also the person deciding.

We use AI assistants day to day in our own company. They prepare work, and people check and approve it before anything is sent or published. The scenario applies that principle to client intake at a law firm.

AI for accountants examines another workflow where professional judgment remains with the person reviewing the draft. Document automation describes this type of integration.

Questions. Short answers

Is this AI giving legal advice?

No. The model reads, summarizes, drafts and prepares. A person at the firm reads and sends every message to a client or prospective client. A lawyer makes every decision about a matter. The model’s output is a draft by construction, not by policy.

Where does client data go?

The inquiry and notes the model needs to read go to the business provider the firm chooses. The data processing agreement must exclude training and identify the storage region. Every read is logged. Whether that arrangement meets your professional obligations is for the firm and its adviser to assess.

Does it work with our practice management software?

Yes, if its API supports the required work. The main products used by small firms in the UK, US and Canada offer APIs. Where a product has no API, a software robot can move the information, with more maintenance when the interface changes.

What does an AI intake integration for a law firm cost?

It depends on the channels and systems connected and the model provider selected. We do not publish prices. After consultation, you receive a written proposal with a fixed price for the agreed step. If the proposal suits you, we sign the contract and start work.

How to start. Measure the intake work

For one month, count the inquiries and measure the time from inquiry to first human reply. Record the minutes spent on inquiries that became matters and those that did not. These measurements show whether intake is the bottleneck and what an integration would need to improve.

The paid audit examines work across the firm’s processes and produces a written report with priorities, estimated costs and risks.

No software brands are named here. Your practice management, calendar and document tools are the starting point. The model and its hosting are chosen in the proposal, with their licenses and data terms and input from the firm’s regulatory adviser.

Book the free consultation

During the free 30-minute consultation, we establish whether the audit is the right step. The audit is paid work.

Got a question we haven't written about yet

we answer, and if it is useful to others, we write the article