AI Receptionist for Law Firms: Costs, Risks, and Setup

What is an AI receptionist for a law firm?

An AI receptionist for a law firm is software that answers routine intake when your staff cannot. It greets a website visitor, captures their name, contact details, and the type of matter, and books a consultation. The safest versions handle after-hours web chat and scheduling, then hand off to a human. They are front-desk help, not lawyers.

Full disclosure: Visione Edge builds after-hours intake systems (our product is Visione Flow), so read this as an engineering-standards buyer's guide from a company with a stake — not a neutral explainer.

Legal-ethics review: [Reviewer name], [jurisdiction / bar no.], [date] — placeholder pending sign-off. Nothing here is legal advice; treat each rule below as a question for your own bar counsel.

What does an AI receptionist cost for a law firm?

Costs split three ways. Voicemail is $0 but captures nothing. AI answering runs from about $49 a month — Rosie's entry plan is $49 for 250 minutes (retrieved 2026-07-08). Live human receptionists cost more: Ruby's phone plans start at $250 a month for 50 minutes, and its live web chat starts at $143 a month for 10 chats.

The table below is a dated snapshot. Vendors change prices often, so treat these as market bands, not quotes, and check each page for today's numbers.

The pattern is clear. Voicemail costs nothing and returns nothing. AI answering is the cheapest live option — Rosie's $49-a-month entry plan — but most market AI receptionists, Rosie included, are voice-first. Human services cost roughly six to twenty times more: Ruby's $250 a month for 50 phone minutes, or $143 a month for 10 live web chats.

Note two things about this snapshot. Rosie answers by voice; Ruby uses live human receptionists. Smith.ai, another well-known vendor, gates its current pricing behind a contact form, so we do not quote a number we cannot see on the page. For a fuller cost breakdown, see our AI receptionist pricing guide.

What are the confidentiality and liability risks?

The real risk is not the monthly fee. It is that an AI at your front desk can create liability and breach confidentiality. A Canadian tribunal has already held a company responsible for its chatbot's words, and bar guidance now treats client information fed to AI as a confidentiality question. These are questions for your bar counsel, not settled answers.

Start with liability. In 2024, in Moffatt v. Air Canada, 2024 BCCRT 149, the British Columbia Civil Resolution Tribunal ordered Air Canada to pay after its website chatbot gave a customer wrong information about bereavement fares. The airline argued the bot was, in the tribunal's words, "a separate legal entity that is responsible for its own actions. This is a remarkable submission." Tribunal member Christopher Rivers rejected that: "It makes no difference whether the information comes from a static page or a chatbot," and the company "is responsible for all the information on its website." The award was small — $650.88 plus interest and fees — but the principle is not. Your intake bot's promises are your promises. We go deeper on this in AI chatbots and legal liability.

Confidentiality is the second exposure. ABA Model Rule 1.6 says a lawyer "shall not reveal information relating to the representation of a client" without informed consent, and "shall make reasonable efforts to prevent the inadvertent or unauthorized disclosure of, or unauthorized access to," that information. In July 2024, ABA Formal Opinion 512 applied this to AI. Because self-learning tools can leak what you feed them, "a client's informed consent is required prior to inputting information relating to the representation into such a GAI tool." The opinion adds that "merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient."

State bars echo this. Florida Bar Ethics Opinion 24-1 (2024) advises lawyers to protect client confidentiality under Florida's Rule 4-1.6 when using generative AI, to obtain the affected client's informed consent before a third-party AI tool handles confidential information, and it warns that a self-learning system may store client details and surface them to others in later inquiries. Your state may take a different line — which is exactly why this belongs with your bar counsel, not with a vendor.

The line to hold is simple. An AI receptionist can take the caller's name at 2 a.m.; it must never take the case.

What are four things a law-firm AI receptionist must never do?

Every safe deployment draws the same four lines. Each one comes from a real tribunal decision or a published ethics opinion, not from our own preference. If a vendor's demo crosses any of them, treat that as a design flaw rather than a feature — and as a reason to keep shopping.

1. Never give legal advice, predict an outcome, or state a deadline. In Moffatt v. Air Canada, the company was liable because the bot's statement was treated as its own — "it makes no difference whether the information comes from a static page or a chatbot." An intake bot that names a statute-of-limitations date can create the same exposure.
2. Never feed confidential case facts into a self-learning model without informed consent. ABA Formal Opinion 512 states that "a client's informed consent is required prior to inputting information relating to the representation into such a GAI tool," and Model Rule 1.6 requires "reasonable efforts to prevent the inadvertent or unauthorized disclosure of" client information.
3. Never promise representation or that the firm will take the matter. Moffatt shows a bot's promise can bind the business. A greeting that says "yes, we can handle your case" may mislead a caller and commit your firm to something no lawyer has reviewed.
4. Never let boilerplate consent stand in for real disclosure. Opinion 512 is explicit that "merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient." Consent has to be specific and informed.

How do you set up after-hours intake without taking on risk?

Keep the AI on the safe side of every line above. The pattern that works is narrow on purpose: a web-chat intake that greets an after-hours visitor, captures name, contact, and matter type, offers a booking link, and hands off to a human by morning. It schedules; it does not advise, promise, or store privileged detail in a learning model.

Start with web chat and booking, not voice. Most market AI receptionists answer the phone with a synthetic voice; a privilege-conscious firm can start narrower, with a chat widget that does intake and scheduling while a human reviews every conversation the next morning. That is the pattern our appointment-booking hub walks through in detail.

Capture the minimum. A name, contact, matter type, and preferred time are enough to book a consultation. The bot should not ask for — or record into any learning system — the facts of the case. That is what keeps you clear of the Rule 1.6 questions above.

Route phone only where it is genuinely safe. If you want after-hours phone coverage, the low-risk version is to route a missed call to a text or a booking link, not to have an AI voice discuss the matter. Phone answering is a separate integration with its own risks; do not assume a chat demo proves a voice can do the same job.

Always hand off to a human. The AI's job ends at "booked and captured." A person confirms the appointment, runs the conflicts check, and decides whether to take the matter.

Does the math work for a small firm?

Yes, if you recover even one otherwise-missed matter a year — but do the math honestly. The software cost is knowable; the return depends on intakes you are losing today, which no honest vendor can quantify for you. So compute it the safe way: cost first, then how little recovery it takes to break even at your own rate.

Take the cheapest live option. Rosie's entry plan is $49 a month, which is $588 a year (retrieved 2026-07-08). Ruby's live web chat at $143 a month is $1,716 a year; its live phone answering at $250 a month is $3,000 a year. Voicemail is $0 — and captures no one.

Now use your own billing rate instead of a borrowed statistic. Suppose your firm bills at $300 an hour. The AI chat option at $588 a year costs under two billable hours. Ruby's live chat at $1,716 costs under six. So the break-even question is not "how many leads will AI convert" — a number you should distrust from any vendor — but "will after-hours intake recover more than two to six billable hours across a whole year?"

For most firms that miss evening and weekend inquiries, one recovered consultation that becomes a retained matter clears the entire year. We cannot promise it will; we can show that the floor is low. Ignore anyone who quotes you "30–40% more leads" without your own numbers — that figure is folklore, not evidence.

What don't we know, and who should not do this?

We do not know your jurisdiction's exact rules, your matter mix, or how many after-hours inquiries you lose today. Anyone who claims to know those without your data is guessing. AI intake is not right for every firm, and some should not do it at all until specific conditions are met.

What we cannot tell you: whether your state bar treats a given tool as compliant — that is a question for your bar counsel — or your conversion lift, which no one can state honestly in advance.

Who should not do this yet:

• Firms whose matters are so sensitive that even a name-and-contact capture on a third-party system is a problem, until you have vetted the vendor's data handling against Rule 1.6.
• Firms with no one to follow up the same business day. An unattended booking is a missed client with extra steps.
• Solo or small firms unwilling to review intake logs regularly. If no human reads what the bot said, you cannot catch the day it crosses a line.

None of this is legal advice. Every rule above is a question to put to your own bar counsel before you deploy.

See what our intake demo actually does

If you want to see the safe pattern in practice, the Visione Flow demo is a scripted web-chat intake. It greets an after-hours visitor, captures their name and matter type, and books a consultation on your calendar — then hands off to a human. It does not give legal advice, and it does not answer phone calls with a synthetic voice. That is the whole point: a front desk that never tries to be a lawyer.

Book a walkthrough and we will show you the chat flow end to end.