AI agents for client and patient intake

Intro — how this agent works

An intake agent answers a prospective client or patient, qualifies them against the criteria that decide whether they are worth taking on, collects the structured facts a case or chart needs, and either books the next step or declines politely. It is the same conversational machinery as a receptionist agent pointed at a much more valuable moment — which is why it is sold, and priced, completely differently.

The intake conversation, step by step

Why intake economics invert every other agent vertical

In support, the goal is the lowest cost per resolution — vendors compete at $0.90 to $2.00. In intake, cost per conversation is close to irrelevant. If a signed personal injury case is worth tens of thousands in fees, paying $20 to handle the call that captures it is trivially correct, and the failure mode is not expense but a missed lead: reported figures put the share of callers who never call back after an unanswered first attempt at around 85%.

Three consequences follow:

• Hybrid human fallback is worth paying for here and almost nowhere else. Smith.ai charging the same for AI-first or human-first is a rational product for this buyer.
• Speed to first response beats conversation quality. An adequate answer in 5 seconds outperforms an excellent one in 5 minutes.
• The structured record is the deliverable. A beautiful conversation that leaves a paralegal re-keying twelve fields has captured nothing.

Two vertical markets, barely overlapping

• Legal intake — dominated by plaintiff-side firms, especially personal injury, where case value justifies real spend. The AI wave here went straight past intake into case work: medical record analysis, demand letters and case valuation. EvenUp at $2B and Eve at $1B are the markers, with the wider plaintiff-law cluster — EvenUp, Eve, Supio, Darrow, Hona, Theo AI, CaseMark — representing over $700M of identified funding.
• Patient intake — a different problem: less about qualifying and more about collecting insurance, consents and payment before the visit, then writing it into the chart. Phreesia built a business on the in-office variant; Luma Health approaches it from communications and scheduling.

Where intake agents fail

• Conflict checks and jurisdiction. An agent that promises representation before a conflict check is a liability, not a productivity gain. Every legal deployment needs a hard gate here.
• Declining well. Most intake calls are not cases. Turning someone down without generating a complaint — or an unauthorised-practice problem — is harder than saying yes.
• Integration depth decides everything. If the agent cannot write into Clio, Filevine, MyCase or the EHR, staff re-enter the data and the ROI evaporates.
• Compliance is not optional. HIPAA for patient intake, privilege and confidentiality for legal, plus call-recording consent in two-party states.

Part 1 — the providers

Legal splits into intake-capture tools and the case-work AI that grew out of them. Healthcare splits into pre-visit intake and the voice agents that feed it.

Legal and healthcare AI vendors overwhelmingly quote rather than publish. Treat the custom rows as "expect a sales conversation", and verify Smith.ai tiers against current plans.

Part 2 — how each provider handles the work

Legal — capturing and qualifying the lead

• Smith.ai — the most-used answer for firms, and the clearest expression of intake economics: the same plan can be answered AI-first or human-first at identical price, $300–$2,100/month for 30–300 calls with per-call overages of $8.50–$11.50. It reports handling 400,000+ calls a month with roughly 80% from law firms, and it integrates natively with Clio, Clio Grow, Lawmatics, MyCase, PracticePanther and Filevine — which is the part that actually matters, because the agent writes a matter rather than leaving a voicemail.
• Lawmatics — an intake CRM rather than an answering service, from about $199/month, priced per firm rather than per user (cheaper than Clio Grow for larger teams, pricier for a solo). It owns the follow-up sequence after capture, which is where most firms actually lose leads.
• Clio Grow — the same job inside the Clio ecosystem, billed per user. The argument for it is not features but that the matter, billing and documents already live in Clio.
• QualifyAI and Josef — lighter automation for firms that can articulate a qualification rubric and want it applied consistently. They sit in the $200–$2,000/month band that covers most credible legal intake tooling.

Legal — the AI that moved past intake into the case

• Eve — a plaintiff-firm platform spanning case intake, medical overviews, drafting and discovery. It reached a $1B valuation after raising $103M, having taken a $47M Series A led by Andreessen Horowitz. Intake is the entry point; the retention comes from the case work behind it.
• Supio — personal injury specific, strongest at ingesting and structuring medical records — the single most labour-intensive artefact in a PI case.
• EvenUp — the largest of the cluster at a $2B valuation after a $150M Series E led by Bessemer (with participation from RELX's venture arm), $385M raised in total. It automates demand letter generation and medical record analysis rather than the intake call itself.
• Darrow — inverts the problem: rather than qualifying inbound, it finds viable cases proactively. Relevant here because it competes for the same budget as intake tooling.

Healthcare — intake as a pre-visit workflow

• Phreesia — the incumbent for the in-office variant: patients check in, sign consents, verify insurance and pay on their own device or a kiosk before reaching the desk, with data flowing into the chart through EHR integration. Its intake is a revenue-cycle product as much as a clinical one.
• Luma Health — approaches intake through communications, from roughly $250/month, combining it with scheduling, multilingual reminders, referral follow-up and waitlist backfill. Choose it when the problem spans the whole patient journey rather than the check-in desk.
• Assort Health — the voice-first path, where intake happens on the phone and lands in the EHR. For most specialty practices the phone is still the intake channel, whatever the patient portal suggests.

Building it yourself

• Retell or Vapi — when your qualification logic is genuinely idiosyncratic, building on voice infrastructure at $0.05–$0.31 per minute costs a fraction of per-call intake pricing. The catch is that the conversation is the easy half; the integration into Clio, Filevine or an EHR is the work.
• Full DIY — an LLM, structured output for the intake fields, and API writes into your practice system. Defensible at high volume with a stable rubric. Not defensible for a firm taking thirty calls a month, where a missed case costs more than a year of Smith.ai.

Part 3 — pros and cons

How to pick, in 30 seconds

• Plaintiff or PI firm, leads are the constraint → Smith.ai for capture, and accept the human fallback premium.
• You capture leads but lose them → Lawmatics. The problem is follow-up, not answering.
• Already on Clio → Clio Grow, unless team size makes per-user pricing painful.
• Medical records or demand letters are the bottleneck → Supio or EvenUp; neither is an intake tool.
• Clinic fixing the front desk and collections → Phreesia. Fixing the whole journey → Luma Health.
• Patients intake by phone → Assort Health, or build on Retell if your logic is odd.

The investor read

• Intake is a wedge, not a market. Every well-funded legal AI company entered near intake and moved into case work — Eve into discovery, EvenUp into demands, Supio into records. Intake gets you the account; document-heavy work retains it. Fund the second step, not the first.
• Value per interaction rewrites the pricing model. Where support agents fight over $0.90 per resolution, intake tools sustain $8.50–$11.50 per call, because the buyer is comparing against a lost case rather than a support ticket. Any vertical where one conversation is worth thousands can carry an order of magnitude more price.
• Plaintiff law is the most concentrated bet in vertical AI. Over $700M into a handful of companies serving one side of one practice area, on the logic that contingency fees make ROI immediate and provable. Watch whether the same concentration appears in specialty medicine.
• Integration depth is the moat, not conversation quality. Smith.ai's defensibility is that it writes into Clio, Filevine, MyCase and PracticePanther. A better-sounding agent that leaves staff re-keying data loses to a worse one that does not.
• Hybrid human fallback survives here. Everywhere else it looks like transitional scaffolding; in intake it is a permanent feature, because the cost of the human is negligible against the value of the captured case.

Frequently asked questions

What is an AI intake agent?

An agent that answers a prospective client or patient, qualifies them against your criteria, collects the structured facts a case file or chart needs, checks for conflicts, and books the next step — then writes a complete record into your practice management system or EHR. It differs from a receptionist agent mainly in what happens after the conversation.

How much does AI legal intake cost?

Credible legal intake tooling sits in a $200–$2,000 per month band. Smith.ai runs $300–$2,100 a month for 30–300 calls with $8.50–$11.50 per-call overages; Lawmatics starts around $199 a month priced per firm; Clio Grow bills per user. Case-work platforms like Eve, Supio and EvenUp are quote-only enterprise contracts.

Why is intake priced so much higher than AI support agents?

Because the value per conversation is orders of magnitude higher. A support vendor competes at $0.90–$2.00 per resolution against the cost of a support ticket. An intake tool is measured against a lost case worth tens of thousands in fees, so $10 a call is trivially worth paying — and human fallback is worth paying for too.

Can an AI agent do conflict checks?

It can trigger one, but it should never promise representation before the check clears. Treat conflict checking as a hard gate in the workflow rather than something the model reasons about, and make sure the agent is scripted to avoid anything that reads as legal advice or engagement.

What is the best AI intake software for a law firm?

For capture with human fallback, Smith.ai — largely because it writes natively into Clio, Clio Grow, Lawmatics, MyCase, PracticePanther and Filevine. For follow-up after capture, Lawmatics. If you are already standardised on Clio, Clio Grow removes the integration question entirely.

Is AI patient intake HIPAA compliant?

Vendor-dependent. Healthcare-specific platforms — Phreesia, Luma Health, Assort Health — build compliance in and sign BAAs as standard. If you build on general voice infrastructure instead, compliance is a paid add-on: Vapi, for example, charges $2,000 a month for HIPAA. Never assume a standard plan covers you.

Other AI agent verticals

Sources

Pricing and integrations read in August 2026: Smith.ai, legal intake cost comparison, Lawmatics, Luma Health, Phreesia vs Luma Health. Funding: Eve $1B, EvenUp $150M at $2B, plaintiff-law AI funding cluster. Call-abandonment and firm-volume figures are vendor-reported, not independently audited.