ChamberOne

Pro bono and legal aid: using AI responsibly with limited resources

By the ChamberOne team 5 minute read

A legal aid clinic has a queue of people who need help and a handful of lawyers and volunteers to give it. Every hour spent typing up an interview or sorting a pile of letters is an hour not spent advising. It is easy to see why AI tools appeal. The reasons for caution are just as clear: the clients are often vulnerable, the stakes for them are high, and the supervisors are already stretched. The aim is to save time on the paperwork without lowering the standard of the advice.

Key points

  • AI helps most with the paperwork around advice: intake summaries, plain-language summaries, routine drafts and checklists.
  • A lawyer must still decide eligibility, advice and priorities. No unreviewed AI advice goes to clients.
  • Use approved tools with individual accounts. Check what happens to client data.
  • Keep supervision real with checklists, sampling and a team habit of reporting errors.

Why the stakes are different

Clients of legal aid services and pro bono clinics are often dealing with housing, immigration, family breakdown, debt or criminal charges. They may have little money, limited English, poor literacy or trauma. A mistake can cost them a home or their safety, and they have fewer ways to recover.

Standards do not drop because the service is free. The duties of competence and confidentiality are owed to every client. Volunteers and students, who may be new to the work, are supervised by lawyers who remain responsible. AI does not change this, and it makes supervision more important, not less.

Where AI saves real time

The tasks that suit AI in this setting are the administrative ones around the advice.

  • Intake: turning a client's account into a structured summary for the supervising lawyer, with the facts, the dates and the questions still open.
  • Summarising long letters, notices and court papers into plain language for the volunteer to check.
  • Drafting first versions of routine letters and forms, for a lawyer to correct.
  • Preparing checklists of documents to ask the client for.
  • Translating the gist of a document so a volunteer can understand its purpose, then confirming with a qualified translator where it matters.
  • Creating plain-language handouts, reviewed by a lawyer before use.

Where to be firm

  • Do not use AI to decide eligibility, prioritise urgent cases or advise on the merits without a lawyer making the call.
  • Do not give clients AI-generated advice directly, without review.
  • Do not let a tool supply legal authority that nobody has read.
  • Do not enter client details into free or unapproved tools.
  • Do not rely on machine translation for anything the client will sign or rely on.

Each of these rules protects someone who may not be able to protect themselves.

Where AI saves time
  • Structured intake summaries for the supervising lawyer
  • Plain-language summaries of letters and notices
  • First versions of routine letters and forms
  • Checklists of documents to ask for
Where a lawyer decides
  • Eligibility and priority of cases
  • Advice on the merits
  • Any legal authority, which someone must read
  • Anything a client will sign or rely on, such as a translation
Where AI helps and where a lawyer decides

Confidentiality with volunteers and shared computers

Clinics often rely on volunteers who come and go, share devices and work from different places. That makes control of data harder. Decide which tools volunteers may use, and make it simple to follow: an approved list, a short written rule and an account for each person, not a shared login.

Check where data is stored and whether the provider keeps or learns from it. Funders and partner organisations may also have conditions about client data, so read the terms of your grants. If the service holds sensitive information about immigration status or abuse, be more restrictive still.

Keep supervision real

A single supervising lawyer may oversee many volunteers. The risk is that the sheer volume leads to skimming. Build supervision into the process so that it does not depend on goodwill.

  1. 1

    Require the volunteer to tick which parts of a document came from a tool, so the supervisor knows where to look closely.

  2. 2

    Use a short checklist for each advice note: facts verified against documents, authorities read, deadline checked, client's questions answered.

  3. 3

    Sample files regularly and discuss what was found with the whole team.

  4. 4

    Make it safe to say that a tool produced something wrong.

Inclusion and language

AI tools tend to work better in some languages than others, and may handle names, dialects and cultural context poorly. A summary of an interview with a client who speaks a less common language may be less accurate than it looks. Where the language matters, use qualified interpreters for the interview and treat tool output as rough notes. Plain-language handouts made with a tool should be tested with real readers, for example by asking a few clients whether the text made sense.

Count the costs honestly

Many tools have a free or reduced price for charities, but free is not the same as cheap. Add the time to set them up, train people and check the output. Ask what happens when the free offer ends, and whether you can export your data and leave. Choose tools that you can sustain on your budget, or that you can drop without harm.

Think about training, too. Volunteers who learn to use tools carefully take a valuable skill with them into their careers. Volunteers who learn only to press a button do not. A short induction covering the approved list, the rules on client data and the checking routine costs an hour and pays back quickly.

A scenario: the housing clinic

Suppose a housing advice clinic sees fifteen tenants on a Saturday morning, each with a bundle of notices and letters. A volunteer takes the client's account while a coordinator uploads the letters to an approved tool, which produces a short list of the documents and dates. The supervising lawyer reads the list beside the originals, notices that one date is wrong and corrects it before advising. Without the tool the lawyer would have read every page anyway, but the list gave a structure and saved most of an hour across the morning. The advice, and the date, remained the lawyer's.

Giving clients a choice

Tell clients, in plain words, if you use technology that processes their information, and what that means. Offer a way to say no where it is practical. Some clients, particularly those who have had bad experiences with institutions, will have strong views. Respecting those views builds trust that the service needs.

Intake and client access in ChamberOne

  • Each firm has a public enquiry form with its own link.
  • Enquiries arrive with a name-based conflict warning. Add notes, then convert one to a client, a lead or a matter.
  • Clients sign in without a password and see their matters, the documents you shared and their invoices.
  • Clients see only what the firm chooses to share with them.

What it does not do

  • Nothing from the enquiry form becomes a client until someone at the firm converts it.
  • A conflict check result is a warning based on names. The decision to act stays with the lawyer.
  • AI output can be wrong or incomplete. It does not replace a lawyer's review.

This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.

Read next

Pro bono and legal aid: using AI responsibly | ChamberOne