AI Receptionist for Law Firms
What an AI receptionist can safely do for a small law firm's intake, the ethics lines it must never cross, and the arithmetic to tell if it pays.
A personal injury lead does not wait. Someone who was rear-ended this afternoon will search, call the first firm, then the second, then the third — and very often signs with whoever calls back first. If your phone goes to voicemail at 6:40pm, the case may already be gone by the time you listen to it.
That is the real case for an AI receptionist at a law firm. It is not about replacing your intake person. It is about not losing the caller in the hours when nobody is there. Here is what it can safely do, what it must never do, and how to work out whether it pays.
Why law firms are different
If you have read our pieces on AI receptionists for clinics and dental offices, the cost structure is the same. What changes for a law firm is the risk.
A single call can be worth a lot. For contingency practices especially, one signed case can be worth more than a month of a receptionist's salary. That makes a missed intake call unusually expensive, and it is why firms are right to care about after-hours coverage.
Every sentence the agent says is a professional-responsibility question. A clinic's agent must not triage. A law firm's agent must not give legal advice, must not appear to accept a case, and must handle what callers tell it as confidential. These are not product features. They are ethics rules, and they apply to the firm.
The facts a caller shares create duties. Under ABA Model Rule 1.18, and its state equivalents, a prospective client is owed confidentiality even if you never take the case, and what they told you can conflict you out of representing someone else. An agent that cheerfully collects a caller's whole story is creating risk, not value.
What it should and shouldn't handle
| Safe to automate | Keep with a lawyer or intake person |
|---|---|
| Hours, location, practice areas you handle | Whether the caller "has a case" |
| "Do you handle car accidents / divorces / estates?" | What the case might be worth |
| Collecting conflict-check facts and a callback number | Deadlines and statutes of limitations |
| "Do you already have a lawyer for this?" | Fee agreements and signing the client |
| Booking a consultation request | Anyone in crisis or in danger: route immediately |
| Answering in Spanish | Anything the script did not anticipate |
Two rules are non-negotiable. The agent never gives legal advice, including the one callers ask most: "how long do I have?" A wrong answer there can cost someone their claim. And the agent never implies the firm has taken the matter. It should say plainly, early in the call, that it is an automated assistant, that it is not a lawyer, and that someone from the firm will follow up.
Collect less, not more
This is the part legal tech marketing rarely mentions, and it is the most important design decision.
The instinct is to have the agent take a full account so the lawyer is fully prepared. Resist it. A detailed account from a prospective client can disqualify the firm from representing the other side of that matter, and it has to be protected whether or not you take the case.
The safer design collects just enough to act:
- Caller's name and callback number
- The names of the other people or companies involved, for the conflict check
- The type of matter and roughly when it happened
- Whether they already have a lawyer for it
- The best time to call back
Then it stops and hands off. Your intake person runs the conflict check and has the real conversation. The caller still gets an immediate response, which is what wins the case, and the firm has not taken on facts it cannot use.
Spanish-speaking callers
For personal injury, workers' compensation, and immigration practices in Arizona and Texas, a meaningful share of callers would rather speak Spanish, and many after-hours calls come from people who could not call during their own working day.
An agent that answers fluently in Spanish, collects the same five facts, and passes them to your team in English is one of the most practical uses of this technology for a small firm. It is also one where a human answering service is often weakest, because bilingual after-hours staff are expensive.
Legal answering service or AI receptionist?
Many firms looking for an attorney answering service, a legal answering service or an answering service for lawyers want exactly what the name says: a person. That can be the right call. A trained person reads a frightened caller better than any agent, and if your calls are few and each one is valuable, paying per minute or per call is reasonable.
The trade-offs are cost at volume, several calls arriving at once, and finding Spanish coverage after hours, which is where an AI receptionist tends to be stronger. The duties do not change with the choice. A person or an agent answering for your firm must not give legal advice or imply you took the matter, and what callers share stays confidential. See published prices compared at two call volumes.
What it costs
- The build: $1,200 to $4,000 for a narrow first version. The range is set mostly by whether it only takes requests or also connects to your case management or scheduling software.
- Usage: a few cents per conversation, usually single-digit dollars a month for a small firm.
- The recurring layer: the phone number, the telephony provider, and any integration fee your case management software charges.
Two questions to ask any vendor: Where are the transcripts stored, and who can see them? and Will the AI model provider use our callers' conversations to train its models? The answers should be written down, because confidentiality is your duty, not theirs.
The arithmetic, with a worked example
You need: missed intake calls per week, the share that are potential cases, the share of those you would actually sign, a conservative recovery rate (assume a third), and your average fee per signed matter.
weekly recovered value =
missed calls × potential-case share × signing rate × recovery rate × average fee
Two illustrations. Both use made-up numbers, chosen to show the shape of the decision, so substitute your own:
Firm A, personal injury. Misses 10 intake calls a week, mostly evenings and weekends. About 30% are potential cases, and the firm signs 1 in 4 of those. With a hypothetical average fee of $6,000 per signed case: 10 × 0.30 × 0.25 × ⅓ × $6,000 = $1,500 a week. A $3,000 build pays for itself in about two weeks of recovered cases. The case is strong, and so is the case for doing it carefully.
Firm B, estate planning. Misses 3 calls a week. Most callers are not in a hurry, half are potential clients, and the firm signs 2 in 3 of those. With a hypothetical average fee of $2,000: 3 × 0.50 × 0.67 × ⅓ × $2,000 = about $670 a week on paper, but the recovery rate is the weak number here. Estate planning callers leave voicemails and wait for a call back, so an agent recovers far fewer of them. For this firm, a reliable voicemail-to-callback routine may be enough.
The difference between the two is not the technology. It is whether your callers shop around in the first hour.
When it is not worth it
- Your callers wait. If your practice area does not involve urgent, shop-around callers, the speed advantage disappears.
- Nobody calls back fast. The agent wins the first minute. If the follow-up takes two days, you have still lost the case.
- You cannot supervise it. If nobody at the firm will read the transcripts, do not deploy it. Supervising the tool is your professional obligation, not an optional extra.
New to this, or still comparing? Our guide to AI receptionists for small business covers the ways to buy one, and answering service vs AI receptionist puts published prices side by side.
How to start
Cover evenings and weekends only. Have the agent identify itself, collect the five facts, and promise a callback within a set time, then make sure that callback happens. Read every transcript for the first month, with a lawyer reviewing what the agent said. You will learn what callers actually ask, and you will catch anything the script handles badly before it matters.
If you want to see the other half of the intake journey, our law firm landing page case study shows a personal injury site built around a four-step "do I have a case?" intake that ends by passing the visitor into the firm's contact form, so a person makes the real call. For a working conversational agent and its handoff, see the clinic AI assistant case study.
We are not lawyers and this is not legal advice. Check your state bar's rules on advertising, technology, and prospective clients before going live. When you want an intake agent scoped to your firm, our practical AI service starts with the same question this article does: how many cases are you losing after 5pm?
Running a trades business instead? HVAC flips several of these rules, starting with who decides what is urgent: see AI receptionist for HVAC companies.
Frequently asked questions
- How much does an AI receptionist cost for a small law firm?
- A narrow intake build — after-hours and overflow calls, basic qualification questions, and a clean handoff to your intake person — typically runs $1,200–$4,000 to set up, plus a few cents per conversation and your phone provider's monthly fee. Connecting it to your case management software adds cost depending on what that software allows. Be wary of a flat 'legal AI receptionist' subscription quoted before anyone has asked how many intake calls you actually miss.
- Can an AI receptionist give legal advice?
- No, and it must be built so that it cannot. It should not tell a caller whether they have a case, what it is worth, or how long they have to file. Those answers are legal advice, they can be wrong, and a caller may rely on them. The agent's job is to take the facts your intake team needs, say clearly that it is not a lawyer and that no attorney-client relationship has been formed, and get the caller to a person quickly.
- Does an AI receptionist create an attorney-client relationship?
- It should be designed so it never appears to. The agent must not say or imply that the firm has taken the matter. Separately, information a prospective client shares can create duties for the firm even if you never take the case — under ABA Model Rule 1.18 and your state's version of it — which is why a good intake agent collects only what you need for a conflict check and a callback, not the whole story.
- Is it ethical for a law firm to use AI for intake?
- Generally yes, with supervision. The ABA's Formal Opinion 512 (2024) addresses lawyers' use of generative AI and stresses competence, confidentiality, and supervising the tool like you would a non-lawyer assistant. Your state bar may have its own guidance. In practice that means you know what the agent says, you review its transcripts, the data is kept confidential, and a lawyer makes every decision about taking a case.
- What should a law firm's AI receptionist ask a caller?
- Only what your team needs to act: the caller's name and callback number, the other parties involved (for the conflict check), the type of matter, roughly when it happened, whether they already have a lawyer for it, and the best time to call back. It should not collect a detailed account of what happened. That conversation belongs with a person at your firm.
- Should a law firm use a legal answering service or an AI receptionist?
- Use a person if your calls are few, each is valuable, and callers often arrive distressed. Consider an AI receptionist for evenings, weekends, several calls at once, or Spanish coverage. Many firms could use both. Either one answers on your behalf, so it must not give legal advice or imply you took the matter, and someone at the firm has to supervise it.