Virtual Receptionist for Law Firms: Rule 1.18, Real Prices and an Intake Script

The first question on your intake script is an ethics decision, and many firms let the vendor's default script make it for them. A virtual receptionist for lawyers answers your calls, screens and books new matters and routes the urgent ones, either with trained people or with an AI voice agent. What decides whether that helps or hurts is what the script asks in the first ninety seconds, because a caller who starts describing their matter may become a prospective client of your firm under the rules you are licensed against.

I have shipped 126 production AI systems, voice agents among them, and the intake script is the part of any intake build I would spend the most time on. Vendors sell tone and coverage and play you a friendly voice answering at 2am. They do not show you that the voice is collecting information from someone who may already count as your prospective client.

I skip the ten vendor listicle. This covers what those leave out: what a law firm virtual receptionist can legally create, what the main services cost in 2026 on their own published prices, and how to write a script that does not manufacture a conflict. I am an engineer, not a lawyer, so read it as a practitioner's guide to the operational side and check your own state's rules.

What a virtual receptionist for lawyers actually does

A virtual receptionist for lawyers answers calls your firm cannot take, captures caller details against a script you define, collects what a conflicts check needs, books consultations into your calendar, and escalates genuine emergencies to whoever is on call. The good ones connect to your practice management system so the call lands as a record rather than a voicemail nobody plays back.

In 2024 Clio hired a research company to contact 500 law firms by phone and email with a client inquiry. Shoppers reached 52% by phone, only 40% picked up when called, and 48% were unreachable even after messages were left (Clio Legal Trends Report). Email did worse: 33% of firms responded. When a call did connect, 41% offered rate information and 12% could estimate the total cost.

A service that picks up fast and asks the wrong questions is a faster way to create a problem.

Receptionist, answering service or AI: which one you are actually buying

A vendor will happily use all three names for one product, and a "legal answering service" or "law firm answering service" is often the same thing as a legal virtual receptionist under a different heading. One vendor's legal page draws the line like this: an answering service takes messages when your office cannot pick up, while a virtual receptionist acts as your front desk, screening callers, transferring calls, running intake and booking consultations (Abby Connect, legal answering). That vendor then sells both, plus an AI receptionist, on one plan. Treat the labels as marketing and compare what the line can do.

The quickest test is three questions: who answers the call, who wrote the script, and what may the line do without a person approving it. Billing is a clue at best. Smith.ai bills per call for both its human and AI plans, Ruby, Posh and LEX Reception by the receptionist minute, Abby by a pooled minute that covers human and AI, and NextPhone by a flat monthly price. If you are shopping for after hours message taking, my guide to the legal answering service covers billing models, overflow rules and the questions to put to a provider. If the AI option is what you are weighing, AI receptionist for law firms covers what an agent may never do and a test script to run before go live.

The rest of this article covers what all three share. My reading, and a question for your own counsel, is that the rule does not care what the voice is made of: a message taker who asks "what is this about" and writes down a paragraph is collecting that information for the firm, and so is a model that does the same.

The Rule 1.18 problem nobody quotes you

Under ABA Model Rule 1.18, a person who consults a lawyer about the possibility of forming a client and lawyer relationship is a prospective client, and the lawyer cannot use or reveal what they learned, even when no relationship follows. If the person shared information that could be significantly harmful to them in the matter, the lawyer cannot represent someone with materially adverse interests in the same or a substantially related matter, and that disqualification passes to every lawyer in the firm unless an exception applies (ABA Model Rule 1.18).

Now hold that against your own intake script.

Vendors describe it differently. One legal page says its receptionists capture "the caller's matter, contact details and the qualifying questions you set" (Abby Connect). Contact details and qualifying questions are fine, and opposing party names are what a conflicts check needs. But capturing "the matter" can turn into "tell me what happened", which generates disqualifying information from a caller nobody has screened yet.

The official comment is specific. A consultation is likely to have occurred when a lawyer invites information about a potential representation without "clear and reasonably understandable warnings" that limit the lawyer's obligations, and a person provides it (Comment [2] to Rule 1.18). Comment [2] also carves out a person who volunteers information unprompted, so a caller who leaves facts on a voicemail is a different case from a script that asks for them. Comment [4] tells a lawyer to limit the first consultation to information that reasonably appears necessary to decide whether to take the matter. Comment [5] adds that a lawyer may condition a consultation on the person's informed consent that nothing disclosed will stop the firm representing someone else. An intake script is the natural place for all three ideas.

The rule also gives you a path out, and it is built around restraint. Rule 1.18(d)(2) preserves the representation where the lawyer who received the information "took reasonable measures to avoid exposure to more disqualifying information than was reasonably necessary" to decide whether to represent the person, is timely screened from the matter and apportioned no part of the fee, and the prospective client gets prompt written notice. Rule 1.18(d)(1) is the other door: informed consent, confirmed in writing, from both the affected client and the prospective client. That path assumes a lawyer was exposed, so the safest design is one where nobody at the firm needs it because the script never collected the facts. Restraint at intake is evidence, not etiquette: you have to be able to show you took it.

Here is the pattern I would expect. A firm buys an answering service, the vendor supplies a generic intake script written for home services, and it opens with an invitation to describe the problem. For a plumbing company that is the correct script. For a family law practice taking a call from one spouse, where the other spouse may call next week, it is a risk to check against your vendor contract.

The fix is short. Your first ninety seconds should collect identity and adversity, not narrative: caller name, opposing party name, any other names involved, matter type at the category level, and jurisdiction. Then stop. Book the consultation and let a lawyer take the facts.

What a virtual receptionist for lawyers costs in 2026

On Smith.ai's published rates, human answering for law firms runs from $300 to $2,100 a month, and AI answering runs from $150 to about $500 for matched call volumes. Smith.ai prints both human and AI prices, so I could compare them on one vendor's numbers rather than a vendor average nobody can check.

On the human plans, 30 calls is $300 a month, 90 calls is $810 and 300 calls is $2,100, with overage at $11.50, $10.50 and $8.50 per call (Smith.ai receptionist pricing). On the AI plans the monthly price is the included calls multiplied by the per call rate: 75 calls at $2.00 is $150, and 300 calls at $1.67 is $500 (Smith.ai AI receptionist pricing). I read both pages again on October 23, 2026 and every figure is unchanged. The AI page also lists a free plan of 25 calls a month with $3.00 a call after that.

Two of those AI figures need showing. The cheapest paid AI plan includes 75 calls for $150, so 30 calls still cost $150, which is the $5.00 a call. At 90 calls you pay that $150 plus 15 extra calls at the published $2.50 overage rate, or $187.50. A firm that only wants to test the idea can stay on the free plan: 30 calls would be 25 included plus 5 at $3.00, or $15.

Now the caveats, because a cost table with none is a sales sheet. The gap narrows at low volume, where you mostly pay for availability rather than labor. And the per call rates assume the AI completes the call, so ask what happens to the price when a call transfers to a human. I covered the full stack in my AI voice agent pricing breakdown and the straight comparison in AI answering service versus human answering service.

Legal virtual receptionist services compared

Here is a shortlist to start from for a US firm, judged on what decides the Rule 1.18 question: whether you control the intake script, whether conflict names reach your practice management system, and who actually picks up. Every price comes from the vendor's own pricing page, in USD, read on October 23, 2026. If your firm is in Australia, the local services, AUD prices and Privacy Act rules are in my separate comparison for Australian firms.

The billing units make a straight price ranking misleading, so here is one workload on each: about 100 calls a month at three minutes each, or 300 minutes. The totals are my arithmetic on the published plans and overage rates, using the cheapest combination each vendor publishes.

Only Smith.ai prices a conflict check as a line item. For every other service, ask in writing whether the receptionist runs the names against your system or simply records them, because recording is the part a vendor can do and clearing is still your job. Then ask whether you can see and change the exact script and read what the agent actually asked. Smith.ai's human plans include five short answer intake questions, which is tight once you list names, matter category and jurisdiction. Ask to see the script before you compare prices.

The supervision duty is yours, not the vendor's

When you hand intake to an outside service, you do not hand over the ethical obligation, and that is easy to miss when you compare vendors on price alone. Model Rule 5.3 requires a lawyer with managerial authority to make reasonable efforts to ensure the firm has measures giving reasonable assurance that a nonlawyer's conduct is compatible with the lawyer's professional obligations, and a lawyer with direct supervisory authority has the same duty over that person (ABA Model Rule 5.3). The comments say this reaches nonlawyers outside the firm, name an internet based service that stores client information as an example, and say the lawyer should give directions appropriate under the circumstances (Comments [1] and [3] to Rule 5.3).

Rule 5.3 does not care whether your assistant is a person in a call center or a voice model on a server. If it acts for your firm, you owe reasonable efforts to keep its conduct compatible with your obligations. Rule 1.6(c) adds a duty to make reasonable efforts to prevent the inadvertent or unauthorized disclosure of, or access to, information relating to the representation of a client (ABA Model Rule 1.6), and the ABA's release on Opinion 512 notes that other rules extend similar protections to former and prospective clients' information, and Rule 1.18(b) covers what a prospective client tells you. A recording or transcript is that kind of information from the first call.

For AI tools the ABA's first ethics opinion on generative AI points the same way. Formal Opinion 512, released on July 29, 2024, says lawyers and firms using generative AI must "fully consider their applicable ethical obligations" (ABA announcement), and it builds on Model Rules 1.1 on competence, 1.6 on confidentiality and 5.3 on nonlawyer assistance (UNC School of Law Library).

That puts four questions into your vendor diligence, and I would not sign without written answers. Where is call audio stored, for how long, and who can access it. Is any call content used to train a model. Can you get the intake script in writing, change it, and see a log of what the agent actually asked. And does the service record the call, which brings in your state's recording law. In California, recording a confidential communication without the consent of all parties can bring a fine of up to $2,500 per violation (California Penal Code section 632). I have not surveyed the other states, so check yours. A vendor who cannot answer the third question is asking you to put your name on a script you cannot see.

What California's 2026 guidance signals about agentic AI

The State Bar of California has written guidance for AI that takes actions on its own. It issued its Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law in November 2023. The 2026 version replaces it and, at the request of the California Supreme Court, addresses the unique challenges of agentic AI (State Bar of California).

The guidance does not leave you guessing whether this applies to answering services. Listing what counts as agentic, it gives four examples, and the last one is client intake that the system runs on its own. The guidance then draws the line: lawyers must not deploy agentic systems in a way that lets the system make "substantive legal determinations", communicate legal advice, prepare and file pleadings or otherwise act in a representative capacity without "meaningful lawyer supervision and review". It also says the greater the system's autonomy, the greater the lawyer's obligation to put oversight in place.

A voice agent that reads a script and takes a message is a tool. One that decides which callers to reject, books matters into your calendar and sends the confirmation is acting for the firm, which is where the guidance asks for supervision. If you are choosing a vendor this year, the ability to constrain what the agent may decide on its own will age better than the ability to make it sound human.

Where AI answering fits for a law firm, and where it does not

AI answering is a good fit when call volume is high, questions are repetitive, and the goal of the call is to qualify and book rather than to advise. It is a poor fit when the caller is likely to be in crisis, when the matter type carries a short deadline, or when the first call is effectively the consultation.

Where it works well

• High volume practices where most calls are status checks, directions or scheduling
• Overflow during business hours, when the fifth simultaneous caller is lost
• After hours coverage where the realistic alternative is voicemail, not a person

Where I would not use it

• Criminal defense intake where the caller may be in custody or under time pressure
• Practices where callers are routinely distressed and a wrong tone costs you the matter
• Any firm without a conflicts database the agent can actually check against
• Matters with imminent statutory deadlines, where a booking two days out is a risk

The pattern I would pick is not either or. It is AI answering first for identification, conflicts data and booking, with an immediate transfer rule for anything the script flags. That way the cheap path handles the volume and a human handles the judgment, which is the human in the loop pattern applied to a phone line. I use the same split for home services voice agents, though the flag conditions are completely different.

What an AI receptionist for a law firm can safely automate

Read against Rule 1.18, the safe line for an AI receptionist is the one I draw for a human: identity and adversity, never narrative. An AI receptionist for a law firm can answer every call, give hours and directions, take the caller's name, the opposing party's name and any other names involved, record the matter category and jurisdiction, book the consultation and send the confirmation. It should not ask what happened, tell the caller whether they have a case, predict an outcome or turn a caller away on the merits. The first creates the disqualifying information the rule is about, and the rest come close to the "substantive legal determinations" California's 2026 guidance says need meaningful lawyer supervision. Write the stop rule into the agent's instructions, then test it by trying to tell it your story. The longer version, with a go live test script, is in AI receptionist for law firms.

How I would set this up for a small firm

Write the screening rule before you look at a single vendor, because the script matters more than the vendor. Many firms do it backwards: they pick on price and coverage, then inherit the vendor's default script.

1. List your matter types and mark each one as bookable, screen out or transfer immediately.
2. Write the ninety second script. Identity, opposing party and other names, matter category, jurisdiction, contact details. No facts of the matter.
3. Write the stop rule in plain language. If the caller begins describing the substance, the agent says the lawyer will take those details at the consultation, and moves to booking.
4. Connect the conflicts check. Names captured at intake need to reach whoever clears conflicts before the callback, not after it.
5. Decide your escalation triggers and test them by calling your own number, at night, as an unhappy person.
6. Log everything and review week one. Read the transcripts and look for any question you did not intend to ask.

Step five is where most of the design work sits. Three branches cover the first sort of every call: an urgent caller goes to a live person at once, a routine caller gets names and a booking, and an existing client gets a message to the team instead of an intake. The firm decides what counts as urgent, in a short written list: a custody hearing within days for a family practice, an arrest for a defense practice, a deadline inside a week for any of them. The transfer destination must be a phone someone answers, not a voicemail box, which is why escalation paths get tested at 2am rather than assumed.

For booking specifically, see what breaks when AI writes to your calendar in virtual receptionist appointment booking. For the comparison against hiring, see how much a virtual receptionist costs.

Is a virtual receptionist right for your firm?

Answer these six questions before you request quotes. It is the gate I would run before quoting or buying anything, and sometimes the answer is that the firm needs to fix who answers the phone between 9 and 5, not hire a vendor.

• Are you losing more than a handful of calls a week, and do you know the number rather than guessing it?
• Can you name the three matter types that make up most of your inbound calls?
• Do you have a conflicts database that someone can check the same day?
• Is your first call usually a screening conversation, or is it usually advice?
• Can you write down what the agent must never ask, in one sentence?
• Will someone at your firm actually read the call transcripts in the first month?

If you answered no to the third or the fifth, fix that before you buy anything. A service that captures names you never check is a cost with no benefit, and a script you cannot constrain is the risk this whole article is about. If you answered yes to most, this works, and the AI path is worth pricing seriously.

If you would rather have the intake flow designed, tested and monitored for you, that is what I build: see AI voice agents for law firms and how I work with firms. Bring your current intake script to the first conversation and I will read it with you.

Frequently asked questions

What is a virtual receptionist for lawyers?

It is an outside service that answers your law firm's phone, follows an intake script you define, captures caller and opposing party details, books consultations and escalates urgent calls. Trained people, an AI voice agent, or a mix of the two can staff it.

What is the difference between a legal answering service and a virtual receptionist?

The words overlap, and vendors use both for one product. In general an answering service takes a message and passes it on, while a virtual receptionist runs intake against your script, screens callers and books consultations. For a law firm, what matters is who writes the script and whether it invites the caller to describe the matter.

Can a virtual receptionist do a conflicts check?

It can collect what a conflicts check needs, which is the caller's name, the opposing party's name and any other parties involved. On Smith.ai's human plans a receptionist can also check the names against your CRM for $0.50 a call. Clearing the conflict is still your decision, and the value of the service collapses if nobody at the firm acts on the result.

Does using an answering service create a lawyer and client relationship?

Answering the call does not create a representation, but it can create a prospective client under Model Rule 1.18, which carries a confidentiality duty and a disqualification risk if the caller shares information that could be significantly harmful to them. That is why the script should stop at identity and adversity rather than inviting the caller to explain the matter.

How much does a virtual receptionist for law firms cost?

Smith.ai's published human plans run from $300 a month for 30 calls to $2,100 for 300, with overage of $8.50 to $11.50 per call. Its AI plans for the same volumes run about $150 to $500 a month, plus a free plan of 25 calls. Ruby starts at $250 for 50 minutes. Ask what a transferred call costs.

Is AI answering ethical for a law firm?

The Model Rules do not prohibit it, but Formal Opinion 512 says lawyers using generative AI must fully consider their applicable ethical obligations, building on competence, confidentiality and nonlawyer assistance. In practice you supervise the tool and the script: know what it asks, where recordings live and whether call content trains a model.

Do I have to tell callers they are speaking to an AI?

I found no single national rule to point you to, and requirements vary by state, so check your own jurisdiction instead of trusting a vendor's assurance. Beyond any specific rule, I would disclose by default: a caller who works out mid conversation that the voice was synthetic is unlikely to trust the firm more for it.

Is a virtual receptionist better than hiring someone?

At low call volume a part time person you already employ is, in my view, usually cheaper and better, because they know your matters. Outside services win on coverage: nights, weekends, holidays and simultaneous callers. Compare against the hours you need covered, not a full time salary.

This article describes professional responsibility rules in general terms and is not legal advice. Model Rules are adopted with variations by each state, so check your own jurisdiction's version and any applicable AI disclosure or call recording requirements before you change your intake process.