Content that reaches people
Organic posts and paid ads were producing hundreds of comments per post. Every comment is a potential client with a real question.
Client work · Bilingual AI intake
Firma Broadway: viral posts, hundreds of comments, one AI intake that answers in Spanish
Firma Broadway's organic posts went viral — and brought hundreds of comments and DMs from Spanish-speaking accident victims. We built the AI DM assistant that answers immediately, every hour, and captures the one thing the firm needs: a phone number for a bilingual case manager.
Spanish-first
24/7 DM answering
No legal advice, ever
At a glance
- Client Firma Broadway — Abogados de Accidentes Latinos, personal injury, California.
- Channels Instagram and Facebook DMs.
- Goal Answer questions warmly and collect phone numbers for human follow-up.
- Guardrails No qualification, no legal advice, no settlement values, no timelines.
The client
A Spanish-first firm the community already trusts
Firma Broadway handles accident and personal injury cases for California's Hispanic community, with offices across the state and a fully bilingual team. Its organic content performs — which is exactly what created the intake problem.
A human-first process
Every conversation is meant to end with a bilingual case manager calling the person. The intake layer's only job is to make that call possible.
Trust-sensitive work
Legal enquiries carry real consequences. Anything an assistant says must be accurate, compliant and free of promises the firm cannot keep.
The problem
Attention that arrives faster than a team can answer
- A viral post brings hundreds of comments; DMs then arrive around the clock, including nights and weekends.
- People in pain do not wait. An unanswered DM often means they message the next firm.
- Questions are repetitive — does this count as a case, what does it cost, what happens next — but each answer must respect legal boundaries.
- Manual replies in two languages, at volume, are impossible to keep consistent and free of accidental promises.
- Leads who are not ready to talk immediately still need a follow-up that stays inside Meta's 24-hour messaging window.
Why it happens
Organic reach is bursty. The busiest moments — right after a post takes off — are exactly when the inbox is least manageable, and legal sensitivity rules out improvising.
This describes the gap the build was scoped to close.
What we built
An assistant with one job and strict limits
The system is deliberately narrow: answer basic questions warmly, ask for the phone number, and hand the conversation to a human. Everything else is forbidden by design.
Answers, not evaluation
The assistant does not qualify or disqualify anyone. It answers questions about the firm and its process and asks for a phone number; the human team evaluates the case. If someone believes they have a case, they can leave their number — that is the entire conversion.
Language control that always follows the client
An explicit language request wins immediately. Otherwise the assistant replies in the language of the current message, and continues in the last established language through neutral fragments like phone numbers. The default is Spanish, and a clear language switch is honored mid-conversation.
Legal guardrails written into the prompt
No legal advice. No settlement values or timelines. No promises about compensation. No claiming clients can reach the attorney's cell. No blanket pre-settlement advance offers. The assistant routes everything substantive to a human.
Available around the clock, honestly
The assistant never says the firm is closed or asks someone to come back later. It answers whenever the message arrives, and a human follows up once contact details are captured.
Follow-up inside the platform's rules
A four-message follow-up sequence is scheduled through the same system, designed around Meta's 24-hour messaging window so follow-ups that cannot legally send are never attempted.
Clean handoff to the case team
Captured conversations are summarized into the firm's lead system — with the internal summary always written in Spanish for the case managers — so the human call starts with full context.
Evidence
What the delivered system specifies
The delivered knowledge base and system prompt define the assistant's behavior. These are the structural facts of that build.
2
languages handled with explicit switching rules — Spanish-first, English on request
13+
case types the assistant can describe, from car accidents to wrongful death
4
follow-up messages scheduled with platform-window compliance built in
1
job: get a phone number to a bilingual case manager, without ever giving legal advice
Verified build facts
These facts come from the delivered knowledge base and prompt. No lead-volume, case or revenue claims are made — the firm's results are the firm's own. Firm statistics referenced in the knowledge base are the firm's claims, not independently verified by us.
ManyChatOpenAILead DocketInstagram DMFacebook Messenger
How a conversation runs
Answer, collect, hand over
1
DM arrives
From a post comment, a story reply, an ad, or directly on Instagram or Facebook.
2
Language set
The assistant follows the client's language, switching immediately on request.
3
Answer + ask
Basic questions answered in 1–3 sentences; the phone number is requested.
4
Hand off
The conversation is summarized for the case team; a bilingual case manager calls.
Honest limits
What this case study does not claim
No case or revenue claims
We built the intake layer. We do not claim signed cases, call volumes or revenue produced — those depend on the firm's case team.
The AI is not a lawyer
By design the assistant gives no legal advice and makes no promises. It answers, collects contact details and hands over. That boundary is the point of the build, not a limitation of it.
Firm facts are the firm's
Credentials, ratings and case counts in the knowledge base come from the firm. We label them as client-provided rather than presenting them as our own verification.
FAQ
Questions about this build
Why doesn't the AI screen or qualify legal cases?
Because that decision belongs to a human, and getting it wrong has real consequences for the person asking. The assistant's job is to be warm, answer general questions, and get a phone number to a case manager. Anyone who believes they have a case can get a human call — no AI judgment in between.
How does it handle someone who wants to speak in English?
An explicit language request is honored immediately, in the same response — not just acknowledged. The assistant also follows the language of each message naturally and keeps the established language through neutral messages such as a phone number. The default for an ambiguous first message is Spanish, matching how the firm positions itself.
What happens to a lead who goes quiet after messaging?
A four-message follow-up sequence keeps the conversation alive inside the platform's messaging window, and the case team receives a summary with the captured details. Everything the assistant sends stays within the approved claims and legal guardrails.
Get your plan
Getting more DMs than your team can answer?
Tell us where your enquiries arrive and what they ask. On a free strategy call we will map the fastest intake automation for your firm.
- A free 15-minute AI strategy call — no obligation.
- A scoped blueprint: the biggest leak, and the first automation to fix it.
- Honest fit check — we will tell you if automation is not your fastest win.
The fastest way in: answer six short questions and we will map your lead journey, find the biggest leak and show the fastest automation win — before we ever get on a call.
Free strategy call · No commitment · We reply by email
Keep going
Related pages
More client work
Relevant AI services
Industry pages
Ready to answer every enquiry within minutes?
Book a free AI strategy call and we will walk your intake flow against this build.