LuMay builds AI voice agents, including one discussed below. Competitor prices come from each vendor's public pages, reviewed October 13, 2026.
Quick answer: An AI receptionist is usually cheaper per call and answers every call instantly, around the clock, in multiple languages. A human law firm answering service costs more but brings judgment, empathy, and reliable handling of unusual calls. Most firms get the best result from a hybrid: AI for volume and after-hours intake, humans for escalations and sensitive matters.
Vendor prices change, so confirm them before you buy.
Why phone coverage is a revenue and ethics issue
Missed calls remain the most common intake leak. In Clio's 2024 secret-shopper study, only 40% of law firms answered a prospective client's call. Counting callbacks, 52% were reachable. A caller who hits voicemail at 9 p.m. often dials the next firm on the list. The Clio Legal Trends Report covers the research.
Coverage is also an ethics question. Whoever answers, a person or software, handles prospective-client information your firm must protect and supervise.
What each option actually is
AI receptionist
An AI receptionist is a voice agent. It converts speech to text, decides what to say using a language model and your rules, and replies in a natural-sounding voice.
It can ask intake questions, book consultations, send texts, transfer calls, and write a summary into your systems. It follows the workflow you configure on every call.
Law firm answering service
An answering service uses trained human receptionists who answer in your firm's name. They follow your script, take messages, run intake questions, and patch urgent calls through.
The line is blurring. Ruby now adds optional AI features to its live plans, and Smith.ai sells AI-first, hybrid, and human-first options.
At a glance: AI receptionist vs answering service
Factor | AI receptionist | Human answering service |
|---|---|---|
Pricing | Per call or per minute; lowest unit cost at volume | Per call or per minute; highest unit cost |
Hours | 24/7 by default | 24/7 at major providers |
Peak capacity | Many simultaneous calls | Depends on staffing |
Intake consistency | Same questions on every call | Varies by agent and shift |
Empathy and judgment | Limited and rule-based | Strong |
Languages | Many, vendor-dependent | Often English and Spanish |
Escalation | Rule-based, with a call summary | Judgment-based patching |
CRM data | Structured fields and transcripts | Often messages or notes |
Setup work | Flow design and integration testing | Script writing and onboarding |
Best for | Volume, after hours, multilingual intake | Sensitive, high-touch calls |
1. Cost: setup, subscriptions, usage, and total ownership
How pricing works
Human services charge by the minute or by the call. AI platforms charge per call, or a base subscription plus per-minute usage.
Published pricing as of October 13, 2026 (sources: Ruby pricing, Smith.ai pricing):
Option | Pricing model | Entry price | Included | Other tiers and overage |
|---|---|---|---|---|
Ruby (live receptionists) | Per minute | $250/mo | 50 minutes | 100 min $395; 200 min $720; 500 min $1,725 |
Smith.ai (human-first) | Per call | $300/mo | 30 calls | $11.50 per extra call; 90 calls $810; 300 calls $2,100 |
Smith.ai (AI-first) | Per call | $0 Free / $150 Pro | 25 / 75 calls | $3.00 per extra call on Free; about $2.00 per call on Pro |
LuMay Voice Agent | Base plus per minute | $249/mo Basic | 500 minutes | $0.23/min overage; Starter $499; Growth $999 |
Costs hiding in the fine print
Per-call add-ons (human): Smith.ai charges extra for booking ($1.50), recording and transcription ($0.25), conflict checks ($0.50), and a dedicated Spanish line ($1.00).
Connector minutes (AI): LuMay bills $0.10 per active minute for each connector, such as CRM or calendar, used on a call.
Plan gates: LuMay includes call recording and transcripts only on Growth and above. Its Basic plan has no connectors.
Pass-through fees: SMS carrier charges, number porting, and telephony are often billed separately.
Your team's time: Writing scripts, testing flows, and reviewing calls is a real cost in both models.
Ruby and Smith.ai's human plans state no setup fees. LuMay offers a 14-day free trial and quotes custom enterprise integrations separately.
Cost scenarios
These scenarios combine published prices with one labeled assumption. They exclude taxes, telephony, and internal staff time. Assumption: an average call lasts four minutes. Your call logs will show your real figure.
Scenario A: Small firm, 50 calls a month (about 200 minutes), with booking and recordings
Ruby: 200-minute plan = $720/month.
Smith.ai human: $300 + 20 extra calls at $11.50 + booking on 20 calls + recording on 50 calls ≈ $573/month.
Smith.ai AI: Pro plan covering 75 calls = $150/month.
LuMay: recordings require the Growth plan = $999/month plus connector minutes. LuMay is not the low-cost choice at this volume.
Scenario B: Growing firm, 300 calls a month (about 1,200 minutes), 24/7, CRM and calendar
Ruby: exceeds the largest published plan, so it needs a custom quote.
Smith.ai human: Pro $2,100 + booking on 100 calls + recording on 300 calls ≈ $2,325/month (about $27,900 a year).
Smith.ai AI: about $500/month at the 300-call tier.
LuMay Growth: $999 + two connectors active two minutes per call (1,200 connector-minutes at $0.10) ≈ $1,119/month (about $13,400 a year).
The pattern is consistent. Per-call AI is cheapest, and human service costs most but buys judgment. Platforms like LuMay sit between them and add multiple agents, connectors, and governance controls.
Compare cost per qualified lead, not list price. A cheaper option that books fewer consultations ends up costing more.
2. Coverage: business hours, after hours, weekends, and 24/7
Both models can cover every hour. Ruby advertises 24/7 live answering, including weekends and holidays, at no extra charge. Smith.ai prices a 2 a.m. call the same as a 2 p.m. call.
The real difference is peak capacity. Human services staff to forecast volume, so a surge can mean longer waits. AI agents take simultaneous calls up to platform limits. LuMay states its inbound agent handles more than 1,000 concurrent calls.
Most firms choose one of three patterns: after hours only, overflow when the front desk is busy, or a full front door that answers every call first. The hybrid models below show how to combine them.
3. Call handling: intake, qualification, scheduling, and multilingual support
Intake and qualification
Human receptionists handle messy conversations well. They can calm a rambling caller and notice what the script missed. Smith.ai's human plans include five intake questions, with more as a paid add-on.
AI asks every required question on every call and stores answers as structured data. LuMay's legal AI receptionist template collects accident details, injuries, treatment, insurance, and police report status, and confirms whether the caller already has a lawyer.
Scheduling
Both models can book consultations. Ruby includes scheduling, while Smith.ai's human plans charge per booking. AI agents book directly through calendar connectors. LuMay supports Google and Outlook calendars, depending on plan.
Multilingual support
Ruby offers 24/7 Spanish and bilingual answering. Smith.ai offers a dedicated Spanish line as an add-on.
LuMay states it offers 80+ voices across 25+ languages and can switch language mid-call. Test any multilingual flow with native speakers and real legal vocabulary before launch.
4. Escalation, live transfers, human intervention, and failed-call recovery
Escalation is where firms lose clients or create risk. Decide in writing which calls must reach a person immediately.
Typical urgent calls
A caller or relative reporting an arrest or detention.
A safety concern in a domestic violence matter.
A hearing, filing, or limitations deadline within days.
A medical emergency, which should be redirected to 911.
An existing client with an urgent problem.
How each model escalates
Human services patch calls to on-call staff. Smith.ai can alert your team by SMS, Slack, or Teams so someone accepts or declines the transfer.
AI agents escalate on rules. LuMay triggers handoffs on caller sentiment, unrecognized intent, VIP flags, or keywords, and passes a live summary so callers don't repeat themselves. Its legal template includes a 911 redirect and attorney escalation.
Smith.ai's AI receptionist can also transfer to a live receptionist at any hour, which is a built-in hybrid.
Failed-call recovery
Transfers fail. Attorneys are in court, phones are off, and software mishears names. Plan for it:
Detect the trigger: a keyword, sentiment, or the caller asking for a person.
Attempt a warm transfer to the on-call contact.
Try a backup contact if nobody answers.
Set honest expectations: tell the caller when they will hear back, and never promise representation.
Create a priority task and alert, with the transcript attached.
Review every failed escalation weekly and fix the cause.
Read back names and phone numbers on every call. A wrong callback number loses the client.
5. CRM integrations, call summaries, confidentiality, and data security
Integrations and summaries
Smith.ai includes its first CRM integration free, with options such as Clio, HubSpot, and Salesforce. Its AI plans include recordings, transcripts, and summaries, and Ruby advertises AI-powered transcripts on live plans.
LuMay offers native connectors for HubSpot, Salesforce, Zoho, calendars, and Twilio SMS, plus REST API and MCP access. Its legal template lists Clio, PracticePanther, and Filevine, so confirm during scoping whether your system connects natively or needs custom work.
LuMay can generate a transcript and structured summary covering intent, actions taken, and follow-ups. Test field mapping before launch. Summaries only help when they land in the right matter record.
Confidentiality and vendor diligence
Prospective clients' information is protected even if you never take the case. ABA Formal Opinion 512 notes that Rules 1.6, 1.9(c), and 1.18(b) all apply when lawyers use generative AI. It also tells lawyers to vet AI providers much as they would any outsourced service. Ask every vendor, human or AI:
Is our call data used to train models?
How long are recordings and transcripts kept, and can we delete them?
Who can access calls, and where is data stored?
Will you notify us of a breach or a subpoena for our data?
Which subprocessors touch our calls?
LuMay states it holds SOC 2 Type II and ISO 27001 certification. Its security page describes SSO, role-based access, database-level tenant isolation, PII masking, audit trails, and private-cloud or on-premises deployment. Ruby publishes a data processing addendum, and Smith.ai states it uses no overseas agents.
6. Legal safety: supervision, consent, and the limits of automation
Attorney supervision
Under Opinion 512, managing lawyers must set clear policies on AI use. Supervising lawyers must make reasonable efforts to ensure staff and outside providers comply, under Rules 5.1 and 5.3. The same duty applies to an answering service.
No legal advice
Opinion 512 cautions against letting generative AI stand in for a lawyer's own legal judgment. Script AI agents and human receptionists alike to:
State that they are not attorneys.
Avoid judging case merit or predicting outcomes.
Never confirm that the firm will represent the caller.
LuMay's legal template opens with a non-attorney disclaimer. Make that standard on every channel.
Recording consent
Recording laws differ by state. California, for example, requires every party's consent to record confidential calls, according to the Reporters Committee recording guide.
Play a recording notice at the start of every call, and follow the strictest law that could apply to your callers.
AI disclosure and advertising rules
Opinion 512 flags lawyer-advertising rules for chatbots, citing Florida Bar guidance. Consider telling callers they are speaking with an automated assistant, and check your state bar's current position.
Conflict-screening limits
Receptionists and AI can collect adverse-party names and flag possible matches. Neither can clear a conflict. An authorized person at the firm must decide before the caller shares detailed confidential facts.
Automated assistance vs attorney judgment
Automation can | Only an attorney can |
|---|---|
Collect facts and documents | Evaluate the merits |
Flag possible conflicts | Clear conflicts |
Book consultations | Accept or decline representation |
Share approved firm information | Give legal advice |
Route urgent calls | Decide strategy and fees |
7. Reliability, limitations, and suitability
Where AI falls short
Background noise, weak connections, and strong accents cause transcription errors.
Poorly constrained agents can give inaccurate answers.
Integrations and telephony can fail, so fallbacks matter.
Some distressed or older callers won't engage with automation.
Ask for written uptime commitments and incident history. LuMay publishes a 99.5% SLA on Growth and 99.99% on Enterprise.
Where human services fall short
Per-call cost rises steadily with volume.
Wait times can grow during call surges.
Script adherence varies by agent and shift.
Notes are often free text rather than structured CRM fields.
Who fits which model
AI-first: high-volume practices with repeatable intake, such as personal injury or immigration, and firms with many multilingual callers.
Human-first: practices with emotionally charged calls, such as family law, or clients who expect a person every time.
Hybrid: most firms. AI handles volume and consistency; people handle judgment.
Hybrid operating models that work
AI front door, human escalation: AI answers every call; urgent or complex calls go to staff or a live service.
People by day, AI after hours: your receptionist covers office hours; AI covers nights, weekends, and holidays.
AI overflow: AI answers only when the front desk is busy.
Human service, AI back office: people answer calls; AI produces summaries and CRM entries.
Practical law firm examples
These are illustrative scenarios, not client case studies.
Personal injury, 11:40 p.m.: A rear-end collision victim calls. The AI captures accident details, books a 9 a.m. consultation, and texts a secure document upload link.
Criminal defense, Saturday: A parent reports a son's arrest. The keyword triggers an immediate transfer to the on-call attorney, with a backup contact if unanswered.
Immigration, weekday: A Spanish-speaking caller completes intake in Spanish, and the attorney receives a summary in English.
Estate planning: An older caller asks for a person. The agent transfers right away instead of pushing through the script.
Implementation considerations
Audit 30 days of calls: volume, length, after-hours share, and call types.
Define urgency tiers: what must reach an attorney now, today, or next business day.
Approve the scripts: an attorney should sign off on intake questions and disclaimers.
Configure disclosures: recording notice, non-attorney statement, and AI disclosure where appropriate.
Test integrations: confirm each field lands correctly in your CRM or practice management system.
Pilot one channel: start with after-hours or overflow calls for two to four weeks.
Supervise continuously: review sample transcripts weekly and keep your written AI policy current.
KPIs to track
Answer rate and average speed to answer
Caller abandonment rate
Intake completion rate
Qualified-lead rate and consultation booking rate
Consultation show rate and signed-matter rate
Escalation connection rate and callback time after failed transfers
Data accuracy, checked by sampling transcripts against CRM records
Cost per qualified lead and per signed matter
Complaints, and attorney overrides of automated decisions
For more on measuring answered versus resolved calls, see how AI voice agents help reduce missed business calls.
A decision framework
Answer these questions with your own data:
How many calls arrive each month, and how long do they last?
What share arrive after hours or on weekends?
How many callers are distressed or need judgment on the first call?
Which languages do your callers speak?
Which systems must receive intake data?
Do clients or regulators require specific data controls?
What is a signed matter worth to your firm?
Then apply these rules of thumb:
Under about 50 calls a month, mostly sensitive: a human service, or a low-cost per-call AI plan with live backup.
Over 200 calls a month, repeatable intake: AI-first with human escalation.
Deep integrations or strict data control: an AI platform with configurable connectors and deployment options.
Still unsure: pilot both on after-hours calls and compare cost per signed matter.
Where LuMay fits
The LuMay AI Voice Agent suits firms that want intake connected to their systems and governed by attorney-approved rules. It includes a legal receptionist template, rule-based escalation, and structured call summaries.
It is not the cheapest option for very low call volume. It fits best when you need several agents, CRM and calendar connectors, multilingual handling, or private deployment.
Comparing intake platforms for a personal injury practice? See our guide to the best AI client intake tools for personal injury law firms.
See it on your own calls: Test the LuMay Voice Agent with your firm's real intake script during a 14-day free trial, then decide whether AI-first, human-first, or hybrid coverage fits your practice.
This article is general information, not legal advice.
