For managing partners, practice heads and the advocate who runs the firm’s WhatsApp

AI agents for law firms: every enquiry answered, every document in before the first meeting.

An agent that answers enquiries within a minute without giving advice, collects the facts and the parties, runs the conflicts check, books the consultation with the right advocate, chases the documents, and hands the advocate a one-page intake summary. It never advises, never solicits, and never speaks to a court. Advocates keep every judgement.

Thirty minutes, free. You leave knowing which workflow to start with, what a pilot covers and what it costs.

PilotAI Agent Pilot: one workflow, 2–3 weeks, $4,500 (₹3,50,000) fixed. What the pilot includes.

Who this is for: Firms and chambers with three to forty advocates, enquiries arriving by WhatsApp, referral calls and the website, matters tracked in a practice tool or a very long spreadsheet, and juniors who spend their first hour each day asking clients for the same documents again.

The week this fixes

If this is your Monday, keep reading

The enquiry waits for a junior who is in court

A legal notice arrives at 8 pm; the client messages the firm’s number; someone sees it the next afternoon. By then two other firms have replied, or the client has called a cousin. The matters a firm loses are the ones it never heard about.

The first consultation happens without the papers

The client arrives with a photo of page two of a six-page agreement. The advocate spends the meeting listing what to bring next time, and the real consultation is the second one, a week later, if the client comes back.

Conflicts are checked from memory and documents are chased by hand

Whether the other party is an existing client depends on who remembers. Which documents are still outstanding on which matter lives in a junior’s WhatsApp, and leaves when the junior does.

One workflow, before and after

From enquiry to a prepared first consultation

Trigger: A WhatsApp message at 8:10 pm: "I received a legal notice from my landlord about vacating. Need urgent help."

Today

  1. ClientMessage sits on the office phone overnight. Messages a second firm at 8:30.
  2. JuniorReplies at 2 pm next day between hearings: "Please share the notice." Client sends a photo of page one.
  3. JuniorAsks for the agreement, the rent receipts and the notice date over three more messages across two days.
  4. PartnerConsultation on day four. Half the documents missing. Learns in the meeting the landlord is a company the firm advised in 2023.
  5. FirmMatter opened on day six, if the client has not gone elsewhere. Nothing about the enquiry is recorded anywhere searchable.

With the agent

  1. AgentReplies within a minute, in the client’s language: it cannot give legal advice, but it can get the right advocate ready. Asks the matter type, the parties’ names, the notice date and the deadline it mentions, and the city.
  2. AgentFlags the deadline as time-sensitive, runs the conflicts check against the firm’s matter list, and finds the landlord company was a client in 2023. Routes to the partner for a decision instead of booking.
  3. PartnerClears the conflict on their phone in two taps (the earlier matter was unrelated and closed). The agent proceeds.
  4. AgentOffers three consultation slots with the tenancy advocate, books one for tomorrow, sends the document checklist (full notice, agreement, receipts, prior correspondence), collects each as it arrives, and nudges for what is missing.
  5. AgentWrites a one-page intake summary into the matter file: facts as stated, parties, dates, documents received and missing, conflicts result. The advocate opens the consultation with the papers read.

What stays with a person

  • Every word of legal advice, opinion, strategy and fee negotiation. The agent says "the advocate will advise" and means it.
  • Conflict decisions. The agent finds the match; a partner clears or declines it.
  • Anything urgent beyond scheduling: an arrest, a hearing tomorrow, a limitation period about to expire. Escalated to a named person immediately, with the facts collected so far.
What it saves

Intake and document chasing: a worked example

MeasureTodayWith the agent
First reply to a new enquiryNext working day; longer at the weekendUnder a minute, any hour
Junior hours on intake, per month~24 hours~7 hours
Junior hours chasing documents, per month~25 hours~5 hours
Conflicts check before the first consultationWhen someone remembersEvery enquiry, logged
Documents in hand at the first consultationSomeThe checklist, or a list of exactly what is missing

Worked example. The assumptions:

  • A twelve-advocate firm receiving about 120 enquiries a month by WhatsApp, calls and the website
  • About 12 minutes of a junior’s time per enquiry to reply, collect the basics and set up a consultation, averaged over the ones who never reply
  • About 60 document requests a month across active matters, about 25 minutes each including the reminders and the filing
  • The agent completes about 70% of intake and 80% of document collection within the rules; the rest reach a person with the facts already gathered

Hours are the visible saving. The matter that never reached the firm because nobody replied that evening, and the second consultation that never happened, are the larger numbers, and a pilot measures enquiry-to-consultation rate and documents-complete rate against the previous quarter rather than quoting an uplift here.

Put your own numbers in the automation savings calculator, or get a price range from the AI agent cost estimator.

The agents

Usually one to start, three when it is working

Each agent has a small set of tools it is allowed to call and nothing else. That list is the guardrail, and it is written down here so you can see it.

Intake agent

On the firm’s WhatsApp number and website. Responds to inbound enquiries only. Explains it cannot advise, collects matter type, parties, dates and urgency, runs the conflicts check, books the consultation with the right advocate from the firm’s own routing rules, and writes the intake summary. Hands over the moment a question needs an advocate.

  • classify_matter
  • collect_parties_and_dates
  • conflicts_check
  • flag_urgency
  • schedule_consultation
  • prepare_intake_summary
  • escalate_to_advocate

Document collection agent

Sends the checklist for the matter type, receives files on WhatsApp or by upload, classifies each against the checklist, asks again for what is missing or unreadable, files everything in the matter folder with a consistent name, and tells the junior only what is still outstanding.

  • send_checklist
  • classify_document
  • check_completeness
  • request_missing
  • file_to_matter
  • report_outstanding

Matter update agent

Drafts, never sends: the status update after a hearing from the advocate’s two-line note, the deadline reminder a week before a limitation date, the fee reminder on an unpaid invoice. The advocate reads and sends each one.

  • draft_status_update
  • log_hearing_outcome
  • remind_deadline
  • draft_fee_reminder

Guardrails

  • No legal advice, ever. Questions of law, strategy, likely outcome or "what should I do" get the same answer: the advocate will advise, here is the earliest slot. The refusal is tested, not hoped for.
  • No solicitation. The agent responds to people who contacted the firm and never initiates contact with a prospective client, which keeps the firm inside the Bar Council of India’s rules on advertising and solicitation (Rule 36) and their equivalents elsewhere.
  • Confidentiality and privilege by design: client data stays in the firm’s own systems, nothing is used to train any model, the data-flow diagram is a one-page document the firm can show a client, and access is logged.
  • Conflicts check before any booking, against the firm’s matter list, with every match routed to a partner rather than decided by the agent.
  • Urgency escalation: arrest, custody, a hearing within 48 hours, or a limitation period in the message goes to a named advocate immediately, by call if needed, not into a queue.
  • Every message, document and decision is logged against the enquiry or the matter, so a complaint or a Bar Council query has a transcript.

Systems it talks to

  • WhatsApp Business API, the website enquiry form and the firm’s intake email
  • Practice management: Clio, MyCase or PracticePanther abroad; Legodesk, Provakil or the matter spreadsheet in India
  • The matter list for conflicts checks, wherever it lives today
  • Google Drive, OneDrive or the document management system for matter folders; Google Calendar or Outlook for advocates; Tally or Zoho Books for fee reminders
Proof, and its limits

What we have built, and what we have not measured yet

The document-chasing agent on the demos page is this design built for a CA firm: a checklist, documents arriving out of order, classification, and a reminder for what is missing. The lead-qualifying demo is the intake half. We have not yet published a law-firm result and will not quote one until a firm agrees to publish theirs; the reference builds show the mechanics, and the pilot measures the rest on your own enquiries.

FAQ

Questions law firms ask us

Will the agent ever give legal advice?

No. It is designed so that it cannot: questions of law, strategy, outcome or "what should I do" are recognised and answered with a booking, not an opinion, and the refusal is tested against a list of the ways clients ask. It collects facts and documents and schedules the advocate, which is what a well-trained receptionist does, and nothing a receptionist may not.

Is this allowed under Bar Council of India rules?

The agent responds to people who contacted the firm; it never initiates contact with a prospective client, never advertises and never makes claims about outcomes or the firm’s standing. That keeps it on the right side of Rule 36 as the Bar Council currently applies it to firm websites and enquiry handling. The firm’s own policy on what the agent may say is written into the rulebook in week one, and nothing goes live without a partner signing it.

What about confidentiality and privilege?

Client facts and documents stay in systems the firm controls. The model provider is used under terms that exclude training on your data, the data-flow diagram shows a client exactly what is read and where it goes, and access is logged. Under the Digital Personal Data Protection Act the firm is the data fiduciary either way; the agent gives it a cleaner answer than a junior’s personal WhatsApp does.

Which systems does it work with?

WhatsApp Business API and the website form on the intake side; Clio, MyCase and PracticePanther have APIs, Legodesk and Provakil are reachable with some work, and a firm running on a spreadsheet and Drive can start there. Conflicts checks run against whatever matter list exists today; a pilot often tidies that list as a side effect.

What does it cost?

The AI Agent Pilot covers the intake agent on WhatsApp for one practice area for a fixed fee in two to three weeks, measured on enquiry-to-consultation rate and documents-complete rate against your last quarter. Document collection across matters and the update drafts are scoped after. The cost estimator gives a range for your setup.

Next step

Tell us about your firm

How many advocates, which practice areas, where the enquiries arrive, which matter system. We reply with what a pilot on your intake would cover, and what it would refuse to do.

Or start with a number: the AI agent readiness quiz takes three minutes and gives you a score and a 30-day plan.

Or send the details

Name and email are all we need. The rest helps us reply with something specific.