AI intake or an answering service: what should a small firm actually buy?

The short version

• Clio's secret shopper study found 48 percent of firms essentially unreachable by phone, and only 33 percent replied to an email at all.
• An answering service fixes pickup. It does not qualify the matter, collect documents, or write anything into your case management system.
• A quarter of firms reply to an online lead within five minutes. The median is thirteen minutes.
• Published answering service prices start at 250 to 300 dollars a month for a small bundle of calls or minutes.
• If your intake criteria are not written down, neither option works, because there is nothing for a receptionist or a system to apply.

What problem are you actually trying to fix?

These two products get compared as if they are alternatives. They are not. An answering service puts a human on the end of a phone that was ringing out. AI intake takes an enquiry that already arrived and does something structured with it. If you buy the second when you needed the first, your phone still rings out, and you have a well organised record of the enquiries that happened to come in by web form.

So the question is not which is better. It is which failure you actually have. There are three, and they need different answers. Nobody picks up. Somebody picks up but asks the wrong questions. Somebody asks the right questions and nothing is written down anywhere afterwards.

How many enquiries are small firms actually losing?

Clio ran a secret shopper study for the ninth edition of its Legal Trends Report, published in October 2024. Researchers contacted 500 firms as prospective clients. Forty percent answered the phone, down from 56 percent when the same exercise ran in 2019. Counting firms that called back later, 48 percent were essentially unreachable by phone.

Email was worse. Thirty-three percent of firms replied at all, down from 40 percent in 2019. Of the firms that did reply, 84 percent managed it within eight hours, which is the encouraging half of the finding. The discouraging half is that only 18 percent of those replies contained clear next steps or any information about cost.

On web leads specifically, Hennessey Digital's 2025 study of lead form response times found 25 percent of firms responding in under five minutes and a median response time of thirteen minutes. Fifty-six percent responded within an hour. Seventy-four percent responded within seven days, which is another way of saying 26 percent never responded at all.

Read those together and the shape of the problem is clear. The failure is mostly not slow qualification. It is no contact.

What does an answering service do, and what does it cost?

A trained person answers in your firm's name, takes a message, and either patches the call through or sends you the details. Some will follow a short script you supply. The good ones sound like your firm. The point of the product is that the phone stops ringing out, and at that it works.

The prices are published, so you can size this without a sales call. Ruby lists four receptionist plans: 250 dollars a month for 50 minutes, 395 for 100, 720 for 200, and 1,725 for 500. Smith.ai prices by call rather than by minute: 300 dollars a month for 30 calls, 810 for 90, and 2,100 for 300, with per-call overage above each bundle.

Note what the units tell you. Fifty minutes is roughly two or three minutes a working day. A small firm that is genuinely missing calls will move up those tiers quickly, and the cost scales with volume because the thing you are buying is a person's time.

What each one is actually for

What does AI intake do that an answering service does not?

Three things, and they are all downstream of contact rather than instead of it. It asks the same qualifying questions every time, including at eleven on a Saturday night, so two enquiries of the same type get screened the same way. It collects what you need from the client rather than noting that you will need it. And it writes the result into the system you already run, so nobody rekeys a message on Monday morning.

The third one is the quiet value. The Clio finding that only 18 percent of replies included next steps or cost is not a speed problem. Those firms replied. What arrived was a reply that left the prospective client with nothing to do next, which is a content problem, and content is the thing a system is actually good at being consistent about.

What it should not do is decide anything that binds the firm. Qualifying, collecting and scheduling are safe. Giving legal advice, accepting a matter, or sending an engagement letter unattended are not, and a system that does them at 2am with nobody reviewing it on Monday is a liability rather than a saving. That boundary belongs in writing before anything is built.

Which one should a solo or two partner firm buy?

If you are missing calls, buy the answering service. It is the cheaper fix, it is available this week, and it addresses the failure the data says is most common. Nothing about AI changes that arithmetic.

If your calls are answered and your problem starts after that, an answering service will not help you, because the gap is in what happens to the enquiry once someone has taken it. That is where intake automation earns its place.

Several firms need both, and that is a legitimate answer too: a human on the phone, a system behind it that qualifies, collects and records. It is also the most expensive answer, so it is worth being sure you have both failures before paying for both fixes.

What breaks if your intake criteria are not written down?

Everything, and it breaks quietly. A receptionist cannot apply a screening rule that exists only in a partner's head. Neither can a system, except a system will apply something anyway, consistently, and you will not notice what it chose until a matter you did not want is in the diary.

This is the part of an intake project that takes the time, and it has nothing to do with technology. What makes a matter worth taking. What disqualifies one immediately. Which conflicts have to be checked before anything is scheduled. What a caller must never be told. Firms that already have this written down find both products straightforward. Firms that do not will find the writing is the project.

The useful consequence is that the work is not wasted either way. Criteria written for a system are the same criteria you would hand a new paralegal.

When is the honest answer neither?

When you answer your own phone and reply the same day. A solo practitioner who already does that has the advantage the data describes, and buying anything to protect it is cost without benefit.

When the enquiries you are missing are not ones you want. More intake is not better intake, and a firm whose real problem is that too many unqualified people reach it should fix the targeting rather than the answering.

And when nobody has decided what a good matter looks like. In that case the first purchase is an afternoon with a whiteboard, not a subscription.