AI intake that answers 100% of calls for firms paying $50–$200+ per click

Every unanswered intake call is a three-figure write-off

Personal-injury keywords are among the most expensive clicks money can buy: common PI terms run $50–$200+ per click, and mesothelioma terms reach $935. Legal services advertising passed $2.5B in 2024, with PI television alone at roughly $1.2B a year (ATRA). At those prices, the cheapest growth lever is not more ads, it is answering every call the ads already generate.

Entropy & Co builds AI intake and nurture systems for law firms: voice agents that take after-hours and overflow calls, SMS follow-up inside a minute, conflict pre-screens against your Clio contacts, and sequences that keep working an unsigned consult for 12 months. No new ad spend, the system signs more cases from the clicks you already pay for.

Agencies serving legal charge $3K–$15K/mo retainers because one signed case can be worth $100K+ in fees (NetPartners, 2026). When a single case can cover a year of automation spend, the system only has to recover what your intake already leaks.

The intake math on a $100 click

The model, with sourced inputs and stated assumptions, substitute your own numbers:

• Mid-range PI cost per click: $100 (sourced range: $50–$200+)

• Assume one intake call per 10 clicks, pull your actual ratio from CallRail

• Each intake call therefore cost about $1,000 before anyone picked up

• If 30% of calls land after hours or on busy lines (check your own logs) and hit voicemail, roughly $300 of every $1,000 in click spend produces calls nobody works

A claimant who reaches voicemail does not wait. They call the next firm on the same results page, and contingency work is winner-take-all: second place signs nothing. The fix is a system that answers every call, qualifies within the first two minutes, and books the consult while the caller is still on the line. Our AI receptionist for law firms breakdown covers the answering layer; this page covers intake to retainer.

Workflows we build for law firms

Each item names the tool it runs on:

1. After-hours and overflow voice intake, a Vapi/Telnyx voice agent that answers, screens for practice-area fit, captures incident date, location, injuries, and insurance details, then books a consult on the attorney's calendar

2. Missed-call text-back, SMS within 60 seconds of any unanswered ring, wired into the same intake record

3. Conflict pre-screen, adverse-party names checked against Clio (or Filevine/Litify) contacts before a consult is booked; the system flags, a human clears

4. Web form speed-to-lead, form fill to first SMS or call attempt in under a minute

5. Unsigned-consult nurture, 12-month sequences for consults that didn't sign, with statute-of-limitations-aware timing per matter type

6. Retainer chase, engagement letter out via Clio or DocuSign with automatic follow-up until signed or declined

7. Referral-out routing, out-of-scope matters routed to referral partners with fee-agreement tracking

8. Closed-case review and referral requests, timed asks after disbursement, feeding the referral pipeline

9. Intake analytics, answer rate, consult rate, and signed-case rate per ad campaign, reported weekly from CallRail and Clio Grow data

Explicitly out of scope: legal advice or case-merit opinions of any kind (that is the unauthorized-practice line, and we engineer hard against it), and running your ad campaigns, that is paid ads supervision, relevant now that Google is force-migrating DSA campaigns into AI Max during 2026.

What stays human

Three guardrails are non-negotiable in every legal build. First, the agent never gives legal advice, evaluates case merit, or predicts outcomes, its scope is scripted, and any advice request escalates to a human with the full transcript. Second, prospective-client information is handled to the standard Model Rules 1.6 and 1.18 demand: encrypted, access-scoped, audit-logged, never used for model training. Third, every transcript is reviewable, managing partners read real conversations during shadow mode, not summaries.

The failure modes are public record: 75% of enterprises rolled back customer-facing AI agents, citing data exposure (31%), hallucination (22%), and missing diagnostics (16%) (Sinch, n=2,500+, May 2026). In a law firm those are not support tickets, they are bar complaints. We wrote up what the rollbacks actually teach; the builds that survive have scoped permissions, escalation paths, and audit trails from day one. Not every intake problem needs an agent, either, agents vs automation maps the line, and our scoping call tells you which side yours is on.

How an engagement runs

1. Intake audit, week 1. We pull 90 days of CallRail and phone logs, measure answer rate by hour, and count unsigned consults in Clio Grow. Output: a leak report priced against your actual cost per click.

2. Build, weeks 2–3. Voice agent on Vapi/Telnyx, SMS flows with TCPA consent capture, n8n plumbing into your practice management. An attorney at your firm approves every script before it touches a caller.

3. Shadow mode, week 4. The agent handles after-hours calls only; a partner reviews every transcript. Checkpoint: you approve expansion or we revise.

4. Staged rollout, weeks 5–6. Overflow next, then front-line assist. Rollback to human-only answering is one toggle, tested before launch.

5. Day-30 report. Answer rate, consults booked, retainers signed, and cost per signed case by campaign, against the week-1 baseline.

Get a scope and quote.

What this work costs in the market

Market anchors, not our rates, your quote is built against your scope. Agencies running intake and nurture automation for law firms charge $3K–$15K/mo, the widest retainer band of any vertical in that dataset. What moves the number: call volume, number of practice areas (each needs its own qualification script), practice-management stack (Clio's API is friendly, Litify means Salesforce work), recording-consent rules in two-party-consent states, Spanish-language intake, and whether nurture spans voice, SMS, and email. For component math, how much an AI agent costs itemizes model fees through monitoring, and AI receptionist cost covers the answering layer. Bring your call volume and cost per click to /contact for a number against your scope, not a rate card.

Why Entropy & Co

Three claims you can check:

1. Article 50 disclosure is built in. Every voice agent we ship discloses it is AI at call start, the EU AI Act transparency obligation lands August 2, 2026.

2. You own the stack. Telnyx numbers, the n8n instance, Clio API keys, transcripts, all in your accounts from day one. Fire us and intake keeps running.

3. The UPL guardrail is testable. Ask our demo agent for legal advice on a live call and watch it decline, log the request, and escalate.

FAQ

Will the AI give callers legal advice?

No. The agent's scope is scripted to fact-gathering and scheduling: incident details, dates, parties, insurance. Any request for advice, merit assessment, or fee prediction triggers escalation to your staff with the full transcript attached. That boundary is enforced in the system prompt, tested before launch, and logged on every call.

How is prospective-client information protected?

Intake data is encrypted in transit and at rest, access-scoped to named staff, and audit-logged on every read. Nothing is used for model training. We design to the duties Model Rules 1.6 and 1.18 impose around prospective-client information, and your firm reviews the data-flow documentation before anything goes live.

Does it integrate with our practice management system?

Clio (Grow and Manage), Filevine, MyCase, Lawmatics, and Lead Docket connect through native APIs via n8n. Litify means Salesforce-side work, which we scope separately. Where no API exists, intake records arrive as structured email or spreadsheet rows, uglier, but nothing gets lost.

Is the SMS follow-up TCPA-compliant?

Consent is captured and time-stamped at intake, opt-outs are honored immediately and propagated to every sequence, and sends respect quiet hours by recipient time zone. For two-party-consent states, call-recording disclosures are configured per jurisdiction. Your compliance counsel reviews the consent flow before launch.

We tried an answering service or chatbot and dropped it. Why is this different?

Most failures trace to missing guardrails, not bad models, data exposure and hallucination led the causes when 75% of enterprises rolled back AI agents (Sinch, May 2026). This build starts in shadow mode with partner-reviewed transcripts, scoped permissions, and a one-toggle rollback. Governance first, expansion second.

Get a scope and quote

One call gets you an intake-leak report against your own numbers and a fixed quote within five business days. Get a scope and quote.

Related: AI automation is the parent service, paid ads supervision protects the click spend this page assumes, and software development covers custom intake tooling like conflict-check databases.