I took a careful look at the layer around your practice. A few things surprised me.
Eight short observations, based on what is public. No pitch, no pricing, no shopping list. I want to see which of these you think are real, and which I got wrong.
This document discusses marketing infrastructure for a New York personal injury practice. Every recommendation below is structured around NY Rules of Professional Conduct 7.1 through 7.5 (including Rule 7.3 on direct contact with prospective clients) and UPL restrictions. Any past-results language quoted is treated with the standard "Prior results do not guarantee a similar outcome" disclaimer.
You have something most 28-year practices do not: a peer-attorney publicly endorsing you on Google.
Fifty-five Google reviews across Seaford, Garden City, and Queens at an aggregate rating around 4.75 stars. A $1.26M auto-settlement record. A trial partner who is fluent in Spanish and French. And an arc most lawyers don't have, an ABC News rights-and-clearances attorney who contributed to the coverage that earned a Peabody for September 11, 2001, before pivoting to negligence trial work.
"As a fellow attorney in Nassau County, Long Island focusing on criminal defense law, I have been noticing throughout the years that the overwhelming majority of personal injury attorneys either lack skill or empathy towards their clients…" Peer attorney, public Google review of The Selvin Law Firm, 3 months ago
A licensed Nassau County attorney put his name on a public endorsement of your practice. That is the single strongest piece of third-party validation in the entire recon. It currently lives on one Google Maps listing and nowhere else, not Avvo peer endorsements, not LinkedIn, not your own reviews page, not any email signature.
Case results referenced are drawn from the firm's website. Prior results do not guarantee a similar outcome.
Sabrina is trilingual. Your homepage is not.
Your own Instagram says "attorney Sabrina Taub is 100% fluent in Spanish." Your /about-sabrina page calls her "a trilingual litigation attorney who speaks both Spanish and French." Your dog-bite page for Massapequa tells the reader "hablamos español." Your own Facebook has a video captioned "Our Spanish-speaking clients are extremely loyal and always…" You're running Indeed postings for a Spanish-speaking paralegal and an immigration assistant.
A Spanish-speaking caller after a Northern Boulevard accident in Queens searching "abogado de accidentes de auto cerca de mí" has no way to find your Jamaica Avenue office. A bilingual firm with three offices, a trilingual co-founder, and a Facebook brag about loyal Spanish-speaking clients is posting English-only to every discovery channel they own.
This is not a Spanish gap. It is a Spanish inconsistency. You've half-committed. The play is to finish what you started.
- Morgan & Morgan maintains forthepeople.com/es: full Spanish site, dedicated Spanish intake, Spanish landing pages per state. Their Spanish traffic is a meaningful share of total US.
- Cellino Law (NY) runs side-by-side English/Spanish practice-area and county pages with bar-compliant disclaimers built into both.
Fifty-five reviews split across three GBPs read weaker than 192 on one.
Weinstock Law, PLLC, which appears as "people also search for" on your own Seaford profile, has 192 reviews at 5.0 on a single listing. Your 55 in aggregate is high-quality work, your Garden City 4.4 is worth looking at, but the way it surfaces to a prospective client comparing you side by side with Weinstock is a number about one third of theirs.
And that is just Google. Everywhere else:
Twenty-eight years of New York practice, $1.26M peak auto settlement, a published peer-attorney endorsement, and zero Avvo client reviews. That is not a service-quality problem. It is a distribution problem with a fix that does not cost much and does not touch Rule 7.1.
- Rule 7.1(e)(3) bars incentivized testimonials. Any post-matter review ask we build never offers anything in exchange, and never gates by star rating.
- No sentiment screening. Every satisfied client is pointed at a public hub, they pick Google / Yelp / Avvo themselves.
- Past results language carries the required disclaimer anywhere it appears in the flow.
You rank for your hometown. You don't rank for the county you practice in.
Search "personal injury lawyer Seaford NY" and you do appear on page one, both in the local pack and organically. I also saw one of your own pages, /wantagh, sitting at around position eight. That tells me the SEO infrastructure already exists. One city page proves the pattern works.
Now search "car accident lawyer Nassau County NY". You are not on page one. You are not in the local pack. You are not in the sponsored slots. Langella & Langella, Cohen & Jaffe, Friedman & Simon, and Richmond Vona occupy those results. And car accidents are roughly 25% of your Avvo case distribution, your largest segment. You have a Garden City office squarely inside Nassau County.
Search "medical malpractice lawyer Long Island NY". Same story. Dell & Dean, which is also in Garden City, occupies a top-three organic slot. You do not appear.
Eight practice areas times the municipalities your own site already lists, Seaford, Garden City, Jamaica, Massapequa, Bethpage, Wantagh, Hicksville, Levittown, Valley Stream, is roughly seventy pages. Each one drafted from the local facts a negligence case actually turns on, hospitals, courts, recent local case law and news, attorney-reviewed for Rule 7.1 compliance, schema-marked for LegalService + Review + LocalBusiness.
- Morgan & Morgan runs thousands of city × practice landing pages on this exact pattern. Example directory →
- Smaller firm case study: a single-market PI firm using AI-drafted city pages (attorney-reviewed) documented measurable local SERP gains within 3–6 months.
You've invested in TV creative. The rest of your funnel doesn't know.
Your own Facebook has a post telling clients to "check out our new TV ad currently airing on Long Island." Whether that flight is still running today or was last quarter, the creative exists and the commitment to broadcast has already been made. Broadcast is expensive and hard to measure on its own; the standard way operators close the measurement gap is to make every other channel compound the broadcast dollar.
Right now, someone sees your TV spot, searches you, and lands on your homepage. From there:
Your YouTube channel has three videos from 2014 with about 1,400 cumulative views and four subscribers. It has been dormant for eleven years. The same creative that is on TV right now could be on that channel tomorrow, as shorts, as Reels, as Facebook video, as the opening twenty seconds of every blog post. The production cost is already paid.
And because TV spending reads as brand and digital spending reads as performance, they compound. A viewer who saw you on TV at 8 PM Thursday and then sees a retargeted ad on Instagram on Saturday is dramatically more likely to convert than either touch alone.
Custom systems built around your firm, not a stack of SaaS on a monthly bill.
Two years ago, every one of the pieces above had exactly one commercial answer: sign up for a law-vertical SaaS (Clio Grow, Lawmatics, Birdeye, Podium, Smith.ai), pay per seat forever, and accept that the tool speaks the vendor's voice instead of Jared's.
There is a third path now. Modern reasoning AI (Claude, GPT-4) plus low-code automation (n8n, Make, Airtable) lets us build each of these pieces custom for your firm: fast, cheap, owned by you, tuned to your 55 reviews, your bilingual capability, your trial voice, your compliance posture. Not renting generic tools forever. Not a marketing agency on retainer.
Five plays follow. Each one lives in a specific place in the funnel. Each one exists in the wild today, either in-house at firms like Morgan & Morgan or on publicly documented open-source tooling. None are hypothetical.
Compliance is not bolted on at the end. Every play was scoped with Rule 7.1, Rule 7.3, and UPL in mind before it was written.
Bilingual AI intake, inbound only, 24/7.
Turn every after-hours form submission into a captured, qualified, ethics-compliant conversation before it goes cold. Closes the gap between your "FREE CONSULTATION 24/7" headline and your 9 AM Monday reply.
Today
A Spanish-speaking woman in Jamaica, Queens, is rear-ended at 10 PM Saturday on Hillside Avenue. She searches for an attorney on her phone from the ER. She lands on your site. The site says "FREE CONSULTATION 24/7." She fills out the contact form. Her submission sits in an inbox until Monday 9:15. By then she has also filled out forms at three competing firms, all of whom advertise "Open 24 hours." Whoever responded first tends to win.
What changes
A bilingual AI intake agent runs on your site (and optionally on Meta DM and your GBP chat). When a visitor engages:
- Detects English or Spanish from the first message automatically, no toggle
- Explains clearly that this is an intake conversation and does not create an attorney-client relationship or constitute legal advice
- Captures the facts a PI intake actually needs: incident date (critical for the statute of limitations clock and for Rule 7.3's 30-day direct-contact window on any downstream communication), injury type, whether 911 was called, whether medical treatment has started, other-party info, police-report status
- Surfaces your case-result context with the required "prior results do not guarantee a similar outcome" disclaimer baked into the response template
- Books directly onto Jared's or Sabrina's calendar for the next available attorney slot
- Sends you a neat summary by SMS, so Monday morning starts with a briefed lead, not an unsorted inbox
Compliance guardrails, built in
- Rule 7.3 inbound-only. The intake agent only responds to inquiries that initiate from the prospective client. We never build outbound SMS to lists of identified accident victims inside the 30-day direct-contact window. Period.
- UPL boundary. The agent is explicitly constrained: it does not give legal advice, does not opine on case strength, does not estimate dollar outcomes. It gathers facts, explains the next step, books a human conversation.
- Rule 7.1 language review. All system prompts, disclaimers, and the "past results" phrasing get attorney-reviewed before go-live, and the version is logged.
- Full audit log. Every exchange is stored with what was asked, what the model answered, and a timestamp. A bar complaint or E&O carrier inquiry is answerable from one query, not a vendor ticket.
vs. Smith.ai, Clio Grow intake, or a 24/7 call center
- Law-vertical SaaS intake: $300–$800/month per seat, forever. Call-center answering services: per-minute plus monthly. Custom: one build + pennies per conversation via enterprise API.
- SaaS intake reads the vendor's script. A custom agent reads the script you wrote, in both languages, with your compliance phrasing.
- Call-center staff are not lawyers and not fluent in your practice areas; AI trained on your intake guide and your bar-compliance rules is a better first-touch.
- When a regulatory interpretation or your intake workflow changes, we update a prompt that same day. SaaS requires a ticket and a release cycle.
What the firm gets
The Sunday-night ER-waiting-room lead stops vanishing. Spanish-speaking callers who previously bounced because no one answered in their language are in your pipeline by Monday. Your intake staff start Mondays already briefed, not sorting forms.
Where it's already happening
Legal-intake AI is an active, documented category. Smith.ai and Clio Grow charge monthly for SaaS versions. The raw building blocks, Claude + Twilio voice/SMS + a scheduler, are publicly documented and can run owned-by-you in about a week of engineering. Lemonade Insurance's AI intake closes 96% of first-notice claims without a human; the same plumbing applies upstream to law-firm intake, just scoped more narrowly.
Finish the Spanish.
You have a trilingual trial attorney, Spanish client testimonials, and a Queens office serving one of the densest Spanish-speaking markets in the country. Your homepage is still English.
Today
Some of your subpages say "hablamos español." Your about-Sabrina page says she is trilingual. Your Facebook has a video bragging that Spanish-speaking clients are loyal. Your homepage has none of that. Your practice-area pages have none of that. Your city pages have none of that. There is no language toggle. A new Spanish-speaking prospect cannot get a consistent experience on your site.
What changes
- Full Spanish version of the site at /es, header toggle on every page, hreflang tags correct for Google
- AI-drafted Spanish translation of every practice-area, city, and bio page, attorney-reviewed (Sabrina) before publish
- Bar-required disclaimers translated once, correctly, then propagated everywhere automatically
- Spanish landing pages for the highest-intent city × practice combinations, with local-Spanish phrasing (Dominican Queens reads different from Mexican Long Island; the agent can tune)
- Spanish-language GBP post cadence, especially for the Jamaica Avenue office
- Spanish content track on Instagram: Sabrina speaking directly to camera on fifteen-second clips answering questions in Spanish that English-speaking attorneys can't
vs. hiring a bilingual marketing consultant
- A bilingual marketing hire or agency: $4K–$8K/month, generic copy, attorney review is bolt-on. Custom: one build, Sabrina's voice, attorney review is first-class.
- Generic translators don't carry your disclaimers through correctly; an attorney-reviewed AI pipeline does, by design.
- When State Bar or NY DFS updates language requirements, one prompt updates propagate everywhere. A human translator redoes pages individually.
What the firm gets
A Spanish-speaking prospect in Queens, Bronx, Long Island, or Brooklyn who searches for help after an accident finds a Spanish-language experience end-to-end: landing page, intake chat, scheduled call, attorney meeting. Sabrina's fluency stops being an invisible asset and starts being a measurable channel.
Where it's already happening
Morgan & Morgan's Spanish site is the public reference. Cellino Law in NY runs bilingual practice-area content with Rule 7.1 disclaimers in both languages. The Latino plaintiffs' bar in Queens and the Bronx is a substantial, underserved market; several firms have built their entire growth around it. Selvin has the people already; the infrastructure catches up in weeks.
City × practice pages, attorney-reviewed.
The infrastructure already exists. One page proves it. Seventy pages is a budget, not a rebuild.
Today
You have /wantagh and a practice-area-plus-city page for dog bites in Massapequa. That's it. Eight practice areas, roughly nine municipalities in your stated service zone, means about seventy city × practice combinations. Sixty-eight of them don't have dedicated pages. Every one of those pages is a search-result your competitors either own or leave uncontested.
What changes
- AI reads your existing 11 website testimonials, 55 Google reviews, your case-result pages, and your practice-area descriptions to extract what a Selvin client actually faces in each city (hospitals, court venues, typical injury patterns, language profile)
- Drafts a first cut of each page: local facts, relevant statutes, linked case-result examples with the Rule 7.1 disclaimer, a clear CTA, schema markup for
LegalService+Review+LocalBusiness - You or Sabrina does a 10–15 minute attorney review pass per page before publish; the agent learns from corrections and does better on the next batch
- Pages refresh quarterly with updated local statistics and new case results as they come in, no touch from you
- Spanish version of each page lives at /es/... per Play Two
vs. a legal-SEO agency on retainer
- Legal SEO retainers: $3K–$8K/month, forever. 70-page custom build: a defined scope with a finite ceiling; quarterly refresh is light.
- SEO agencies draft in generic legal tone; attorney review feels like a chore. AI drafts in your voice pulled from your own published writing; attorney review feels like editing.
- Agencies answer to their book of business. A custom pipeline answers to one client: yours.
What the firm gets
Over six months: rankings on the searches that matter. Car accident lawyer Nassau County. Medical malpractice lawyer Garden City. Construction accident attorney Queens. Dog bite lawyer Bethpage. Pages that compound in Google instead of decaying in an SEO-agency retainer.
Where it's already happening
KrispCall built a page per U.S. area code; those now generate 82% of its U.S. organic traffic. The Search Initiative's local-SEO case study documents the same mechanics at small-local-service-firm scale. The pattern has been stable for years; AI just removes the bottleneck on first-draft content so the attorney review pass is no longer the only path to shipping volume.
Review systematization across all public channels, ethics-first.
Yelp at zero. Avvo at zero. Justia unclaimed. Google split three ways. Fixable in weeks, without a single incentivized review.
Today
Every time you resolve a matter, a grateful client walks out. Sometimes someone in the office remembers to ask for a Google review. Sometimes that review lands on Seaford when the client was served from Garden City. Yelp has been claimed for years and empty. Avvo is claimed but shows zero client reviews against your 28 years. Justia has a banner saying "click here to claim your profile for free." You have not claimed it.
What changes
- When your case management system marks a matter as resolved, a workflow fires. An AI agent drafts a bilingual thank-you that references the specific matter type in neutral, ethics-safe language (no compensation claims, no result-specific language, no incentive)
- The client receives one message with a choice of one public review link, pre-selected based on which channel most needs the signal (Avvo, then Yelp, then the Google profile matching the office they worked with)
- A gentle one-touch reminder five days later. No third message. Ever.
- No sentiment filtering. No star-gating. No incentives of any kind. Every satisfied client gets the same ask regardless of feedback they already gave privately
- Separate workflow for peer-attorney endorsements on Avvo: Jared manually selects ten opposing and co-counsel peers per quarter; the agent drafts the ask, Jared sends
- A one-time profile-completeness sprint: claim Justia, rebuild the LinkedIn company page (the old URL returns "page no longer exists"), reconcile the Facebook /theSelvinLawFirm page with the Instagram @selvinlawfirm handle so both search the same way, populate Avvo with practice percentages and case history
Compliance guardrails, explicit
- Rule 7.1(e)(3). No incentives, no gift cards, no discounts, no promises. The system cannot physically send an incentive because there is nothing in the code path that offers one.
- No sentiment gating. Technically enforced. The system does not have a branch that suppresses an ask based on a prior negative interaction. Every satisfied client, same ask.
- Attorney-client privilege. No matter details surface in the message beyond practice-area category.
- Rule 7.3 timing. Reviews are asked for after case resolution, to existing clients with an established relationship, never to anyone inside the 30-day direct-contact window or during active intake.
- Audit log. Every ask logged with text sent, target channel, response received, to answer any future inquiry in minutes.
vs. Birdeye, Podium, or Clio Grow
- Birdeye & Podium: $299–$599 per month per location, forever; three offices means three seats. Custom: one build.
- SaaS tools default to review-gating behavior the NY Bar has explicitly flagged as a vector for discipline; opting out sometimes requires contacting support. Custom is compliant by construction.
- Avvo peer endorsements are not a feature of Birdeye or Podium. A custom pipeline handles them.
What the firm gets
Over six months: Yelp fills up, Avvo fills up, the Garden City 4.4 rises as fresh service-level reviews come in, Justia gets claimed. That peer-attorney Google endorsement becomes the first of many Avvo peer endorsements. Your review signal stops being a single-point-of-failure on one Seaford listing.
Where it's already happening
Birdeye's public case studies show what review distribution can produce at scale. Same plumbing is publicly documented in the n8n workflow library as an open-source alternative, without the monthly bill. Firms that do the ethics-compliant version well see review volume on under-stocked platforms roughly match their Google volume within 9–12 months, without a single Rule 7.1 issue.
TV amplification + dormant-YouTube revival.
You already paid to shoot the creative. Spread it across the channels the same viewers are on three hours later.
Today
You've shot and paid for TV creative. Nobody is pixeling the traffic when it airs. Your YouTube channel is frozen at three 2014 videos, four subscribers, and ~1,400 total views. Your Instagram is at 480 followers under one handle while Facebook runs on a different handle, nothing cross-posted, nothing reels-cut. The measurement between TV dollar and form fill is manual, if it exists.
What changes
- Dedicated landing pages for each TV flight (e.g. /news12 or /tv-long-island), tagged with UTM, bilingual, with a Meta + Google pixel firing
- The TV creative is cut into twelve short variants: 15s for Instagram Reels, 15s for YouTube Shorts, 30s for Facebook, 60s for LinkedIn, vertical and horizontal, with on-screen captions in English and Spanish
- A retargeting audience is built from TV-landing-page visitors; Meta and Google Display ads follow the same viewer for 30 days with "see why they called us" content
- The YouTube channel gets reactivated with a short-form-first strategy: weekly 60-second clips of Jared or Sabrina answering one narrow PI question, AI captions in EN/ES, repurposed to Reels and Shorts and Facebook
- Monthly dashboard: TV flights vs. retargeted impressions vs. form fills vs. calls. One-page, one-eye.
Compliance notes
- All short-form cuts carry the same past-results-disclaimer language the original TV spot needed
- Case-specific numbers ($1.26M, $600K, $250K) stay paired with the disclaimer in any asset that surfaces them
- Retargeting audiences are site-visitors only, not matched against external accident-victim lists (clean separation from Rule 7.3's targeted-solicitation concerns)
vs. a media-buying agency or a social-content agency
- Full-service TV+digital agency: $5K–$15K/month + ad spend. Custom: the ad spend stays the same; the media buying and measurement layer becomes a fixed-scope engagement, not a retainer.
- Generic content agencies produce Canva-looking Reels that could belong to any firm. Cuts from your TV spot keep your production value.
- Agencies report on vanity metrics (impressions, reach). A custom dashboard reports on form fills and calls, which is the thing Jared actually wants to know.
What the firm gets
The TV dollar starts producing measurable lift in calls and form fills. The YouTube channel starts compounding again. The handle mismatch between FB and IG gets fixed. The same creative works five times instead of once.
Why custom is the safer play for a law firm specifically.
The kind of detail your E&O carrier, the Grievance Committee, and any future regulator quietly care about.
Attorney-client privilege and confidentiality
A law firm's first-touch intake contains the seeds of privilege. The moment a prospective client describes an accident in detail, confidentiality attaches (Rule 1.18). With Smith.ai, Birdeye, or any shared-tenant SaaS, those exchanges sit in their database, co-mingled with thousands of other firms' traffic. With a custom build, every intake conversation lives on infrastructure you control (your cloud account, your database, your API keys). When a client asks "what do you have on file about me," you can answer from your own system, not a vendor's export tool.
Enterprise-grade AI privacy, by default
Both Claude (Anthropic) and GPT-4 (OpenAI) offer enterprise API tiers with explicit guarantees: your data is not used to train their models, retention is zero or near-zero, and a proper Business Associate-class Data Processing Agreement is available. A generic SaaS chatbot typically sits on lower-tier API access and pools your traffic with every other customer. A custom build puts your firm directly on the enterprise tier with your own API key and your own isolated traffic, the same privacy posture a hospital network or a bank uses. For a PI practice that routinely handles medical records, that posture matters.
No vendor lock-in, ever
Every law-vertical SaaS ships with the same pattern: once your intake data, templates, referral history, and workflow logic are inside their system, migration is painful on purpose. With custom, every piece is yours. Intake data in formats you own (CSV, Postgres, Airtable). Workflow logic in code any engineer can read. Prompt templates in version-controlled files. If our engagement ever ends, you keep the entire system, and can hire any competent engineer to maintain it for a fraction of the SaaS renewal bill.
Audit-grade transparency
NY Grievance Committees, your E&O carrier, and any future regulator care about one thing: can you show what happened. A custom build writes a complete audit log for every AI interaction: who asked what, what exact text was sent to the model, what came back, what decision was made, and when. If a client ever contests that a non-lawyer gave them legal advice in an intake chat, the log settles it in minutes. With SaaS, you get a compliance brochure and a promise, and the actual logs sit in their tools, not yours.
Economics that improve over time
SaaS pricing moves in one direction: up. More seats, more tiers, annual increases. LLM API pricing has moved the opposite direction, the same-quality call dropped roughly 10× in 18 months, and the curve keeps falling. A custom build locks in today's costs as a ceiling, not a floor, and rides the drop. Meanwhile you pay no per-seat fees, no renewal cycle, no "we're raising prices 20% this quarter" emails from three vendors at once.
The roadmap is yours
When the NY Bar tightens language around attorney advertising next quarter, we update one disclaimer template that same afternoon. When you open a fourth office, we add a row. When an appeals decision changes how a practice area gets described, we tune. SaaS vendors ship on their own quarterly roadmap and lobby you for upsells; your custom stack ships when you need it to.
Where the industry is going
The shift away from heavy SaaS toward LLM-native, customer-owned automation is underway industry-wide. Andreessen Horowitz frames it as "AI eating application software." In legal tech specifically, the discovery of attorney-client privilege issues around shared-tenant AI has accelerated adoption of self-hosted and own-tenant setups. Going custom in 2026 is not a contrarian bet; it's skating where the puck already went.
To be plain about my side of this.
I'm not pitching you a shopping list of subscriptions. I build and run the custom systems, using modern AI agents and low-code tooling. Concretely:
-
• Pick the right one or two plays firstNot all five at once. The one with the shortest path to a measurable win for your mix of auto / med-mal / construction / Spanish-speaking clients and your current TV schedule.
-
• Scope the compliance pass up frontRule 7.1 language, Rule 7.3 direct-contact boundaries, UPL constraints, privilege handling, disclaimer templates. Agreed on before a single line of code ships. I'll want Sabrina's eyes on the Spanish ones.
-
• Build and wire them inCustom AI agents, site widgets, SMS flows, bilingual content pipelines, retargeting glue, dashboards. All wired into your existing Wix site, your case management, and your phone. You own the output; no per-seat SaaS attached.
-
• Train them on your firmYour 55 reviews, your Peabody-to-PI arc, Sabrina's trilingual voice, your $1.26M and $600K and $250K case frames (disclaimed correctly), your trial cadence. Off-the-shelf AI sounds like a call center; yours should sound like The Selvin Law Firm.
-
• Run the weekly cadenceIntake, content, review follow-ups, SEO pages, retargeting, handled in the background. You approve anything that touches a client communication; everything else runs itself.
-
• Report what's actually movingMonthly: inbound-leads triaged, consultations booked, reviews gained per platform, TV-driven form fills, SERP positions. When a play is earning, we build the next; not before.
If any of this lands, or if I got something wrong and the picture is different from where you actually sit, I'd rather hear it than keep a polished read I can't defend. You've been practicing 28 years; you'll see any handwaving immediately.
Drop me a line whenever.
No deck to sit through, no slide count. Thirty minutes, on your clock.
dslabakov@me.com