Key takeaways
- The phone is still the most common way people talk to their legal adviser, so unanswered calls cost work and goodwill.
- AI call answering is good at answering every call, taking structured details, applying urgency rules and booking consultations.
- It must not give legal advice, decide whether the firm takes a matter, accept instructions or handle payment details.
- The firm stays accountable for what an automated assistant says, so scripts, testing and review matter more than features.
What an AI receptionist is
An AI receptionist is an automated service that answers phone calls in natural speech, works out why the caller is ringing, and then does something useful: takes details, answers a simple question from approved information, books an appointment, routes an urgent call or passes on a message. Unlike older phone menus, the caller simply talks. Unlike a voicemail box, the call ends with something recorded and actioned.
For a law firm, the attraction is obvious. Fee earners are in meetings, at court or on other calls, and reception is stretched at lunchtime and after 5pm. The question is not whether automation can answer a phone, but what it should and should not do once it has.
Why the phone still matters
Despite email and portals, the phone remains central. The Legal Services Board's Legal Needs Survey found the telephone was the most common way people communicated with their main adviser on a contentious issue (28%), ahead of face to face (27%) and email (24%)[1]. In the 2026 consumer tracker, 66% of consumers said they would be happy to use the telephone for an initial meeting with a lawyer[2].
People also want more contact than they get. Some 17% of consumers would have preferred to speak to a legal professional more than they did, rising to 24% of those served by telephone and 25% of those served by email or online[2].
Source: Legal Services Consumer Panel, 2026[2]
Show the data
| Item | Value |
|---|---|
| All consumers | 17% |
| Served by telephone | 24% |
| Served by email or online | 25% |
What it can do well
The SRA has recognised the potential. Its Risk Outlook on AI notes that legal chatbots can respond quickly to enquiries from potential clients and potentially offer a 24/7 service, gathering information and triaging cases for urgency[3]. In practice, a well-configured service can:
- Answer every call, including at lunchtime, after hours and when several calls arrive at once.
- Take the details the firm has decided it needs: name, contact details, practice area, a short description and any deadline mentioned.
- Apply the firm's urgency rules, for example a court hearing tomorrow or someone at a police station, and pass those calls to a person straight away.
- Book or request an initial consultation in the right fee earner's diary, following the firm's booking rules.
- Answer simple questions from approved information: opening hours, directions, documents to bring, where published prices can be found.
- Send the team a clear written note of every call, so nothing depends on someone remembering a voicemail.
What it cannot or should not do
- Give legal advice or comment on the merits of someone's case.
- Decide whether the firm will act, including conflict decisions, which need a person reviewing the conflict search.
- Accept instructions or tell a caller they are now a client.
- Discuss an existing matter in detail without the firm's identity checks for existing clients.
- Take card or bank details, or confirm any change to payment arrangements.
- Handle a distressed or vulnerable caller alone beyond taking essential details and getting a person involved quickly.
- Make promises about outcomes, timescales or costs beyond the firm's published information.
These limits are not technical weaknesses to be fixed in the next release. They reflect professional duties. The SRA is clear that firms remain responsible and accountable for outputs from AI they use, including errors in advice given by a third-party chatbot[3]. Every limit above is a place where an error could become a regulatory problem.
Which calls suit automation
| Call type | Suitable for automation? | Notes |
|---|---|---|
| New enquiry, routine | Yes | Take details, book or request a consultation, send a note |
| New enquiry, urgent | First step only | Recognise urgency and transfer or alert the duty fee earner immediately |
| Existing client wanting an update | Partly | Take a message for the fee earner; do not discuss the matter without checks |
| Other side, court, agent or lender | Partly | Route to the right person or take a detailed message |
| Caller in distress or at risk | No, beyond essentials | Hand to a person straight away, or give emergency contact routes if nobody is available |
| Payment or bank detail query | No | Route to accounts with call-back checks |
| Complaint | Partly | Log it and pass to the complaints lead the same day |
What callers think
Callers are open to automation but want people in charge. The main barriers to using AI to access legal services are lack of trust in AI technology (51%) and lack of human oversight (49%)[1]. In the Legal Services Board's 2026 public panel, 81% agreed AI legal tools should never act automatically for a vulnerable person in high-stakes situations[4]. A good service says it is automated at the start, explains what it can do, and makes it easy to reach a person.
Callers who need extra help
Some callers find automated services hard to use: people with hearing or speech difficulties, callers whose first language is not English, older callers unused to talking to a machine, and anyone who is upset. Design for them from the start. The assistant should speak clearly and at a steady pace, allow pauses, repeat information when asked and accept a simple request such as 'person' or 'reception' at any point to reach a human or leave a message for one.
Give callers another route too. Publish an email address and the option to request a call back, and make sure the website explains how to contact the firm without using the automated line. Review calls where the caller gave up part-way through; they often show where the script is confusing.
What a well-scoped service looks like
The best results come from treating automated answering as a member of the intake team with a tightly written job description, not as a clever phone line. Ava, for example, is an AI Employee for solicitors that answers new enquiries, follows the firm's approved intake questions, routes urgent matters, books or requests initial consultations and hands over a clean note; links into practice management systems are still planned. Whatever service you look at, ask the provider to show each of these working on your own scripts, and to show exactly where the assistant stops.
Questions to ask a provider
- How does the assistant introduce itself, and can we set that wording?
- How do we set the questions it asks, per practice area, and who can change them?
- How are urgent calls recognised, and what happens if nobody answers the transfer?
- What stops it giving legal advice, and how do you test that?
- Where are recordings and transcripts stored, for how long, and who are your sub-processors?
- How do we review calls, and how quickly can we correct a script?
- What happens during an outage?
Example: a call at 7.40pm
Example (fictional): a caller rings a family firm at 7.40pm. The assistant says it is the firm's automated assistant and cannot give legal advice. The caller explains she has received court papers about child arrangements with a hearing in eight days. The assistant takes her details, the other party's name for the conflict search and the hearing date, recognises the date as within the firm's urgent window and books her into the first available consultation slot the next morning. The duty solicitor receives a note flagged urgent. At 9.05am the solicitor runs the conflict check, calls the client and confirms the appointment.
Running a pilot
Start with overflow and after hours
Let the assistant take calls reception cannot answer, so you compare it with voicemail rather than with your best receptionist.
Write and test the scripts
Test every call type in the table, including attempts to get advice and a distressed caller scenario.
Review every call for two weeks
A supervisor reads each note and listens to a sample of calls, correcting scripts the same day.
Measure
Track missed calls, consultations booked, urgent calls transferred and any complaints or concerns.
Decide
After a month, expand, adjust or stop, based on what the review showed.
For the comparison with staff, read AI receptionist vs human receptionist for solicitors, and for the difference between answering calls and completing intake work, see AI receptionist vs AI employee. What to tell callers is covered in telling clients when you use AI, and other low-risk uses in AI for practice management and admin tasks.
Frequently asked questions
Can an AI receptionist give legal advice?
It should not. It can take details, answer simple questions from approved information and book consultations, but advice must come from a solicitor. The firm is accountable for anything the assistant says.
Is AI call answering suitable for a law firm?
For routine new enquiries, overflow and out-of-hours calls, it can work well when scripts are tightly written, urgent calls go to a person and every call is reviewed during a pilot.
Should an AI receptionist say it is automated?
Yes. It should say so at the start, explain what it can and cannot do, and offer a way to reach a person.
Can an AI receptionist handle urgent calls?
It can recognise urgency using the firm's rules and pass the call to a person immediately. It should not try to deal with the substance of an urgent matter itself.
References
Legal Services Board (2024). Legal Needs Survey 2023: Summary report (opens in a new tab) ↑
Legal Services Consumer Panel (2026). How consumers are using legal services (Tracker Survey 2026) (opens in a new tab) ↑
Solicitors Regulation Authority (2023). Risk Outlook report: The use of artificial intelligence in the legal market (opens in a new tab) ↑
Legal Services Board (2026). AI in Legal Services: Research Report (LSB Public Panel) (opens in a new tab) ↑