Best AI Intake Software for Law Firms (2026): Honest Pricing and the Ethics Rule Vendors Skip

Every AI intake vendor sells the same promise: capture every detail, never miss a lead. For a law firm that promise points the wrong way. Under the ABA rules that govern prospective clients, the intake conversation that collects the most is also the one most likely to disqualify the firm from taking the case. The script is the product. The voice is the packaging.

The short answer. For most small firms, Smith.ai is the honest starting point, but not at its advertised entry price: custom AI prompting is an Enterprise-only feature, so a firm that needs an intake script shaped around its ethics obligations pays 500 dollars a month, not 150. Ruby is excellent and genuinely inclusive, but it bills per minute, and a real legal intake call is long enough that Ruby works out to roughly 36 dollars per answered call on its most popular plan. Building your own on a voice platform saves under 100 dollars a month at every volume we checked, so build for control, never for cost.

All prices below were read on the vendors' own pages on 5 August 2026. Where a vendor publishes nothing, we say so rather than repeating a number from a directory.

The rule that inverts the sales pitch

Under ABA Model Rule 1.18, anyone who consults your firm about possibly retaining it is a prospective client. Confidentiality attaches even if no relationship ever forms. Under Rule 1.18(c), if that person shares information that could be significantly harmful to them in the matter, the lawyer is disqualified from representing the opposing side, and that disqualification is imputed to the entire firm.

ABA Formal Opinion 510, issued 20 March 2024, is the part almost no vendor page mentions. It reads Rule 1.18(d)(2) and concludes that firm-wide imputation is avoided only where the lawyer took reasonable measures to avoid exposure to more disqualifying information than was reasonably necessary to decide whether to take the matter. The opinion's own worked example of a reasonable measure is telling the prospective client that no engagement has been agreed and asking them to limit what they say.

Read those two together and the standard intake upsell reverses. "Extended intake: six additional questions" is a product that sells you more of the exact thing Opinion 510 tells you to limit. An intake bot that asks more collects more risk, and it does it at scale, overnight, without a human noticing that the caller just volunteered the other side's settlement position.

ABA Formal Opinion 512, issued 29 July 2024, is the first ABA formal opinion on generative AI. It runs through competence (Rule 1.1), confidentiality (1.6), supervision (5.1 and 5.3) and reasonable fees (1.5), and it makes one point that changes procurement directly: boilerplate consent buried in an engagement letter is not adequate informed consent for putting client confidences into a generative AI tool. Lawyers are expected to understand how the tool handles data.

So the buying question is not which voice sounds most human. It is: can I control exactly what this thing asks, can I stop it asking more, and where does the transcript go.

Smith.ai, and the 3.3x jump nobody quotes

Smith.ai publishes two separate menus, an AI-first receptionist and a human-first one, and they price on completely different logic.

The AI Receptionist ladder: a Free tier at 0 dollars for 25 real calls a month, then 3.00 a call. Pro at 150 dollars covers 75 calls (2.00 a call), or 150 calls at 1.80, or 300 at 1.67. Enterprise at 500 dollars covers 300 calls at 1.67, 500 at 1.60, and custom above 1,000. Extra calls run 2.50 down to 2.10. It is month to month with 30 days notice, and the Free tier includes 25 test calls and 50 simulated calls through a Quality Studio, so you can hear the thing before it answers a real client.

Here is the catch that decides this for a law firm. On Smith.ai's own feature table, custom AI prompting and custom integrations are marked "Not included" on Free and on Pro. They are Included on Enterprise only. Everything in the section above says a law firm's intake script has to be deliberately shaped, and shaping it is precisely the feature the 150 dollar plan does not carry. The real entry price for a firm that takes Opinion 510 seriously is 500 dollars a month, a 3.3x multiple on the advertised one.

Credit where it is due: native Clio and Zapier integrations are included even on the Free tier, which is unusually generous, and it matters because Clio is where most firms already keep the matter record.

One asymmetry worth knowing before you pick a menu. On the human-first plans Smith.ai states plainly that there is no charge for spam, and its FAQ confirms that wrong numbers and unsolicited sales calls do not count so long as your forwarded lines pass Caller ID. On the AI plan, every answered call counts toward your quota by default, with an allowance to strip up to 10 percent of calls per billing cycle for selected spam. Known spammers are filtered before the AI picks up, but the AI tier is still the one that can bill you for junk. If your firm advertises anywhere that attracts robocalls, price that in.

The human-first ladder, for comparison: 30 calls for 300 dollars, 90 for 810, 300 for 2,100. That is 10.00, 9.00 and 7.00 per call, with overage at 11.50, 10.50 and 8.50. There is a 30-day money-back guarantee capped at 1,000 dollars and 10 percent off for a 12-month subscription.

The legal add-on stack, and the discount it quietly eats

Smith.ai's a-la-carte menu is where legal intake actually gets priced, and it is the most transparent list in this category. Per call: conflict checks 0.50, extended intake 1.50 (six additional questions), custom extended intake 0.25 per question, third-party intake 1.00, appointment booking 1.50, call recording and transcription 0.25, SMS or Slack notification 0.50 plus 0.25 for each additional ping, Microsoft Teams notification 0.50, text and email follow-up 0.50, dedicated Spanish line 1.00, accept payments 1.00 (LawPay, TrialPay and CPACharge are named), complex call routing 1.50, custom calendaring 1.00, and accept collect calls 0.50, an add-on whose description explicitly names attorneys who work with inmates. Per month: business caller ID for outbound 10, each additional transfer destination 15, each inbound tracking number 5. Free of charge: lead qualification, a dedicated number, new client intake at five short-answer questions, and your first CRM integration, with additional CRMs at 0.50 a call.

Now build the stack a personal injury or family firm actually needs. Conflict check 0.50, extended intake 1.50, appointment booking 1.50, recording and transcription 0.25, and an SMS ping so someone knows a live one just came in, 0.50. That is 4.25 per call on top of the plan rate.

Loaded, the human ladder becomes 14.25, 13.25 and 11.25 a call. And that produces the finding worth the whole section:

The published volume discount from the smallest plan to the largest is 30 percent. With the legal add-ons on, it is 21 percent. The add-ons are flat per call, so they do not scale with volume the way the plan rate does. Roughly a third of the discount you scaled up to earn is absorbed by the features that make the service usable for a law firm in the first place.

And the sharpest detail in this whole comparison: the conflict-check add-on appears only on the human-first menu. The one intake step that is unique to practising law is the one step the AI tier does not sell. If a conflict check has to happen before you take details, that single line item decides which menu you are on, regardless of what the AI can do.

Ruby, and what per-minute billing does to a long call

Ruby prices by receptionist minute: 50 minutes for 250 dollars, 100 for 395, 200 for 720, 500 for 1,725. That is 5.00, 3.95, 3.60 and 3.45 a minute. The pricing page was last modified 18 March 2026.

Ruby's inclusiveness is real and should not be glossed over. It states there are no additional fees for activation, onboarding, setup, customization or coverage during certain periods, that AI enhancements are included at no extra cost on every plan, and that scheduling, bilingual answering, client intake, payment collection, lead qualification and outbound assistance are available at every tier. HIPAA-compliant services are available. The only difference between plans is the number of minutes. Nothing on the Smith.ai menu above is charged separately here.

The crossover. Smith.ai's 90-call human plan is 9.00 a call. Ruby's 200-minute plan is 3.60 a minute. Those two cost the same at 2 minutes 30 seconds of talk time. Load the Smith.ai side with the 4.25 legal stack and it is 13.25, which moves the crossover to 3 minutes 41 seconds.

Legal intake is not a two-and-a-half minute call. A first conversation with an injured caller who is reading dates off a letter, spelling an insurer's name and describing a timeline runs 8 to 12 minutes. At a 10-minute call, Ruby's most popular 200-minute plan answers 20 calls a month at 36 dollars each.

Set that beside the comparator: Smith.ai's 30-call human plan, fully loaded with the legal add-on stack, is 427.50 a month for 30 answered calls. Ruby's 720 dollar plan costs 292.50 more and answers 10 fewer.

That is not a knock on Ruby's model, and it is worth being precise about why. Minutes are consumed only while someone is actually talking, so a firm whose calls are genuinely short, existing clients asking whether a document arrived, does better on minutes than on calls. Per-minute is not a worse model. It is a worse model for this call.

What building your own actually saves

The obvious response to all of the above is to build the intake agent yourself on a voice platform and keep total control of the script. The control argument is strong. The cost argument is not, and it is worth killing with arithmetic.

Retell AI's published components add up to 0.14 a minute all in for a US number: voice infrastructure 0.055, platform text-to-speech 0.015, the language model 0.045, US telephony 0.015 and PII removal 0.010, with no platform fee on the entry plan. A 10-minute intake call therefore costs about 1.40.

Against Smith.ai's AI tiers at the same volumes:

• 75 calls a month: Smith.ai 150 dollars, self-built about 105. You save 45.
• 150 calls a month: Smith.ai 270, self-built about 210. You save 60.
• 300 calls a month: Smith.ai 501, self-built about 420. You save 81.

Under 100 dollars a month at every tier, before you have paid a single person to build the flow, write the prompts, wire it to Clio, monitor it, or fix it at 11pm when it starts mishandling transfers. Nobody's time is worth less than that.

So build for control, and under Opinion 510 that is a genuinely good reason: you own the script, you decide what it will not ask, you decide where the transcript is stored and for how long. Just never build it to save money, because it does not.

One catch bites hardest here. Retell bills while the caller is silent or on hold, because the speech engine stays active. Legal intake is full of pauses while somebody hunts for a claim number, so a self-built agent's real minute count runs above the talk time you would estimate. We break down the wider platform market in our 2026 AI voice agent comparison, and we build these flows for firms through our AI automation team.

The vendors that publish nothing, and how to read them

Three names come up constantly in this category and none of them publish usable pricing. LexReception and Answering Legal return no readable pricing to an ordinary visitor, and the directories quoting them contradict each other: one lists an entry plan at 425 dollars for 150 minutes, another quotes only a per-additional-minute rate with no base plan at all. Lawmatics publishes nothing, and the directories disagree on whether it is priced per firm or per user, and on whether a three-seat minimum applies.

We are not going to pick one of those numbers and present it as fact. Treat every quoted figure, including every figure in this article, as unverified until it arrives from the vendor in writing on a quote with your firm's name on it.

That is not a throwaway caution. Search for Smith.ai pricing and you will find confident third-party roundups quoting 97.50 a month for AI-only at 30 calls, 292.50 a month for human at 30 calls, and overage between 9.75 and 11 a call, one of them citing an analysis of "500 plus billing patterns". Not one of those numbers appears anywhere on Smith.ai's own pages. The real AI entry is 150 dollars for 75 calls. The real human entry is 300 dollars for 30.

And here is the structural difference that makes legal harder than healthcare. In healthcare, the compliance artifact is a purchasable item with a price on it: the Business Associate Agreement, which as we found in our HIPAA-compliant AI receptionist comparison ranges from free to 30,000 dollars a year depending on vendor. You can shop it. In law there is no equivalent document. No vendor sells a "legal ethics agreement" SKU, because the obligations sit on the lawyer, not the vendor. That means the diligence cannot be bought. It has to be assembled, question by question, by the firm.

Five questions to ask before you sign

• Can I edit the script, and on which plan? If custom prompting sits two tiers above the price you were quoted, you were quoted the wrong price.
• Who runs the conflict check, and is it sold on the tier I am buying? Ask specifically whether it is on the AI menu or only the human one.
• Where does the transcript live, who can read it, how long is it kept, and can I delete it? Opinion 512 makes this your obligation, not the vendor's.
• Does the meter run during silence and hold? On per-minute and per-second models this is the difference between the estimate and the invoice.
• What does the thing I will do 30 times a month cost? Price the loaded call, never the headline rate.

If you would rather not run this procurement yourself, that is the work we do: we design the intake script around the firm's ethics constraints first, then choose the platform that can actually execute it. See AI intake for law firms, or how we approach consent and calling rules on TCPA-compliant AI calling.