AI or a live answering service for a law firm.

In Clio’s 2024 study, researchers phoned 500 US firms as prospective clients. About 40% answered live, down from 56% in 2019, and about 52% made contact once callbacks were counted. A separate 1,200-call audit of small and mid-sized firms, placed between 10am and 4pm, found 34.8% unanswered. CallRail’s lead data puts legal at 28% of tracked leads missed.

A prospective client with a fresh problem calls the next firm on the list. Our missed-call calculator for law firms puts a number on it. At its defaults, a firm taking 30 calls a weekday, with about a third unanswered during business hours and nobody on the phones after six, misses 7,184 calls a year, 4,200 of them after hours. Some callers leave a voicemail or ring back. 5,064 never come back. That is about 355 matters a year, worth $726,737.

The defaults come from the 1,200-call audit for the daytime miss rate, an answering service’s call logs for the after-hours share, and LawPay’s retainer data for the value of a matter, set at $2,500. Put in your own call volume and fee per matter to get your firm’s number.

Measure both options against that loss. Answering nearly every call recovers about $695,409 a year at the same defaults. Either service costs a small fraction of that. What separates them is how much of the loss each one recovers.

Vendor price lists put a traditional legal answering service at $300 to $1,200 a month, with live per-minute rates of $0.75 to $1.50 and dedicated receptionists at $500 to $2,500 a month. AI answering plans are quoted at $25 to $160 a month. These are vendor figures, several from AI vendors, so the live numbers may be high and the AI numbers low.

The price lists leave out two costs. For a live service, the minutes you use beyond the plan, at a higher rate. For an AI agent, the work of connecting it to your case management system so it can run a conflict check and book a consultation rather than take a message. The second is a build, quoted up front.

A message taken at 9pm and returned the next morning recovers some of the loss. By then the caller may have booked with another firm. A consultation booked while the caller is on the line recovers more. That difference, more than the monthly price, is what decides between the two.

A live service answers in your firm’s name, takes the caller’s details and the nature of the matter, and relays a message. Some will book into a shared calendar from a script. The operator does not know your practice areas and changes from shift to shift.

An AI intake agent answers at any hour and takes many calls at once. Connected to your systems it can ask the matter-type and conflict questions you set, screen out what you do not take, book the consultation into your calendar, and open the intake record. It cannot give legal advice, and it should not try. Anything that needs judgment, a distressed caller, an existing client with a problem, an opposing party, goes to a person with the context attached.

The ABA’s Formal Opinion 512, July 2024, is the first formal guidance on lawyers’ use of generative AI. It says the existing rules apply: competence (Rule 1.1), which includes understanding the benefits and risks of the technology you use; confidentiality (Rule 1.6); communication (Rule 1.4); reasonable fees (Rule 1.5); and supervision of the people and services the firm relies on (Rules 5.1 and 5.3). A caller asking about a new matter is a prospective client, and Rule 1.18 protects what they tell you even if the firm never takes the case. State bars have issued their own opinions and some go further.

On consent, Opinion 512 draws one line. Before confidential information goes into a self-learning tool, one whose later users could see it, the lawyer needs the client’s informed consent, and a clause in the engagement letter does not count. A tool that trains on nothing and shows call content to no one outside the firm and its vendor sits on the other side of that line.

For the vendor itself, bar opinions on cloud services set the standard at reasonable care. New York’s Opinion 842 asks for an enforceable duty of confidentiality and security, a look at the provider’s security, and the ability to move the data or have it deleted. So the questions for any intake vendor, live or AI, are the same:

• Where does call content go, and who can see it?
• Is any of it used to train a model?
• Who else touches it: subcontractors, the phone provider, the model provider? Are they bound by the same terms?
• How and when will you tell us about a breach?
• At the end of the engagement, do we get the content back, deleted, or both?
• Is the call recorded, and with what consent?

Keep any vendor that cannot answer those in writing off your intake line.

Our answers. The agent works in the firm’s own systems and runs on our Google Cloud, or in the firm’s own Google Cloud account if it prefers. Our voice provider keeps call content on its system under a business associate agreement with us, and the agent writes it to the matter in your case management system. We sign a confidentiality agreement with the firm, and our US-based partners and contractors work under the same terms. Nothing trains a model, and the agent logs every action. At the end of the engagement we return the call content, delete it, or both, as you choose. How we handle law firm data.

• If most of your missed calls are after hours and your intake is simple, either option recovers most of the loss. Price them both.
• If intake needs a conflict check and a booking, only an agent connected to your system does that on the call. A live service takes a message you act on later.
• If your callers are often in distress, family law, criminal defense, personal injury, keep a person reachable and use the agent for overflow and after hours.
• Whatever you choose, measure the answer rate before and after. Your phone system reports it.