The short answer: what an AI receptionist does for a law firm
An AI receptionist for a law firm answers your phone around the clock, asks callers a set of intake questions, books consultations and sends you a summary. It can cost as little as about $30 a month for a small call volume, and up to roughly $400 a month for an unlimited plan. It cannot give legal advice, decide whether someone is a client, or clear conflicts of interest. Those stay with you or your staff.
The best fit is a solo lawyer or a firm of two to ten people where calls go to voicemail during hearings, depositions and after 5pm. The worst fit is a firm whose calls are mostly existing clients asking about live matters, where a human who knows the file is the whole point.
What it costs
Published prices vary widely, and most come from the vendors themselves, so treat them as a starting guide. A 2026 roundup of legal receptionist tools from CloudTalk, itself a phone vendor, lists these figures:
- Dialzara: from $29 a month for 60 minutes, with overage at $0.48 a minute. Higher tiers start at $99 a month for 220 minutes.
- Smith.ai AI receptionist: from $95 a month, with per-call rates of roughly $1.60 to $1.90 inside a plan.
- Smith.ai with human backup: from $292.50 a month for 30 calls on annual billing.
- LegalClerk.ai: an unlimited-calls plan at $400 a month, no setup fee.
Two things drive the real number. The first is call volume, because most plans charge by minutes or calls. The second is whether a human is in the loop. Hybrid services cost several times more per call than pure AI, and for some practice areas that is money well spent.
Compare that to a full-time receptionist. Salary alone is a much larger monthly figure, and it covers roughly 40 hours of a 168-hour week. An AI service is not a like-for-like replacement, but the gap explains why small firms are looking at it.
What an AI receptionist handles well
Modern AI voice agents hold a natural conversation, so callers are not pushed through a menu of "press 1" options. If you want to see how this works in general, Zapplon explains it on its page about AI voice agents and receptionists. For a law firm, the dependable jobs are these.
Answering every call, including after hours. A potential client with a car accident, an arrest or a landlord dispute is usually calling several firms. The one that picks up tends to get the consultation. An AI answers on the first ring at 9pm on a Sunday.
Collecting intake details. Name, phone number, email, a short description of the issue, the date it happened, the other party's name and how they found you. You decide the questions, and the answers arrive as a text or email summary or go straight into your case management system.
Screening by practice area. If you only take family law in Orange County, the agent can politely tell a caller with a patent question that you are not the right firm. It can also give a referral to your state bar's lawyer referral service if you set that up.
Booking and rescheduling consultations. Connected to your calendar, it offers real open slots and sends confirmations. It also handles reminders, which reduces no-shows.
Answering routine questions. Office hours, address, parking, whether you offer free consultations, what documents to bring, and which languages your staff speak. Many AI agents can also speak Spanish and other languages, which matters in California.
Routing urgent calls. If a caller says they were arrested last night or have a court date tomorrow, the agent can send you a text immediately or transfer the call to your mobile.
What it cannot handle
This is where law firms differ from a dental office or a plumber. The limits are not just technical. They come from your professional duties.
Legal advice. The agent must not tell a caller whether they have a case, what their deadline is or what they should do. Even a friendly "that sounds like a strong claim" can be read as advice and can create expectations. A well-configured agent says that a lawyer will review the details and call back.
Deciding who is a client. Forming an attorney-client relationship is your decision. The agent should never accept a matter, quote a fee for a specific case or say "we will represent you." Use clear wording, both in the call script and your website, saying that talking to the firm does not create a client relationship until a signed engagement.
Conflict checks. An AI can capture the names of the opposing party and witnesses. It cannot reliably judge whether a conflict exists. Someone at the firm has to run those names before you take the matter.
Emotional, high-stakes calls. Someone calling about a death, a custody emergency or a criminal charge may be frightened. AI can be polite and quick, but some callers will want a person, and the agent needs an easy way to transfer them. Test this before going live.
Existing client matters. A client asking about the status of their case needs someone with file access and judgment. An AI can take a message and flag it, and that is about all it should do.
Mishearing. Voice AI still makes mistakes with names, accents, background noise and spelled-out email addresses. A good setup reads details back to the caller for confirmation and keeps call recordings so you can check the original audio.
Confidentiality and ethics: check these before you buy
Calls to a law firm often include privileged or sensitive information, even at the first contact. Lawyers in the US have duties around client confidentiality and competence, and the American Bar Association issued Formal Opinion 512 in 2024 on lawyers' use of generative AI. Your state bar may have its own guidance, and the California State Bar has published practical guidance for lawyers using generative AI. Read your own bar's position rather than relying on a vendor's summary, and ask your malpractice insurer if you are unsure.
Put these questions to any vendor in writing:
- Where are recordings and transcripts stored, and for how long? Can you delete them on request?
- Is your data used to train the vendor's models? The answer you want is no.
- Who at the vendor can access call content? Ask whether human reviewers listen.
- What security controls are in place? Ask about encryption and whether they have an independent security audit such as SOC 2.
- Is there a signed data processing or confidentiality agreement?
- Does the agent tell callers they are talking to an AI? Some states have rules on disclosure and on call recording consent. California requires consent from all parties to record a call, so the agent should announce recording at the start.
If a vendor cannot answer these plainly, move on. A low price is not worth a confidentiality problem.
AI, human answering service or in-house receptionist
There is no single best option. The right choice depends on call volume, practice area and how much each missed lead is worth.
- AI receptionist: Best for firms with fewer than a few hundred calls a month, mostly new enquiries, and a need for coverage when nobody is available. It is cheapest and always on.
- Hybrid or human-backed service: Worth the higher price for criminal defense, immigration, family law and personal injury, where callers are often distressed and the cost of losing one case is high.
- In-house receptionist: Better when the front desk also greets clients, handles documents and manages the office. AI cannot do any of that in person.
Many firms use a combination: a human at the desk during office hours and AI for overflow, lunch breaks and nights. That keeps the personal touch where it matters and removes voicemail from the rest.
To decide, work out your own numbers. Count calls that went to voicemail last month, estimate how many were new enquiries, and multiply by your typical conversion rate and average fee. If five missed new-client calls a month would have produced even one signed case, a $100 to $400 monthly tool pays for itself. If your missed calls are almost all existing clients, it will not.
How to set it up without creating problems
A sound rollout takes one to three weeks for most small firms, mainly because writing the script and testing it takes more time than the technology.
- Write the intake script. List the questions per practice area and what the agent must never say.
- Set the escalation rules. Define which phrases trigger an immediate transfer or text, such as arrest, restraining order, deadline or hospital.
- Connect your tools. Link the calendar, your email or SMS, and your case management or CRM so nothing is retyped.
- Add disclosures. Include the AI and recording notice and a statement that no attorney-client relationship exists yet.
- Test with real scenarios. Call it yourself with an angry caller, a mumbled name, a Spanish speaker and a client asking about their case.
- Review calls weekly at first. Listen to a sample of recordings and fix gaps in the script.
Zapplon builds this kind of setup for small businesses, but the checklist applies whichever vendor you pick.
Frequently asked questions
Can an AI receptionist give legal advice?
It should not. A properly configured agent collects information, answers logistics questions and passes the matter to a lawyer. Any vendor that promises the AI will assess cases for callers is asking you to take on risk you do not need.
Is an AI receptionist allowed under attorney ethics rules?
Using one is generally not prohibited, but you remain responsible for confidentiality, supervision and clear communication with callers. The ABA's Formal Opinion 512 and your state bar's guidance are the starting point. This is general information, not legal advice, so check your own jurisdiction's rules.
How much does an AI receptionist for a law firm cost?
Published plans run from about $29 a month for small volumes to about $400 a month for unlimited calls. Hybrid services with human agents start closer to $300 a month for around 30 calls. Your cost depends on call volume and how much human involvement you want.
What to do next
Start with a one-month count of missed and after-hours calls, then shortlist two or three vendors and ask each the confidentiality questions above. Run a trial on your overflow and after-hours calls first, not your whole phone line. If the summaries are accurate and callers book consultations, expand from there.