Context
Gaea Legal is a two-partner immigration advisory practice (Madrid/Murcia). Every WhatsApp message, email and Instagram post was handled manually — time that lawyers should spend on cases, not on repeating the same answers.
The problem
- ~10 client conversations/day arriving on WhatsApp, mostly recurring questions about immigration procedures.
- Email triage and drafting consumed partner time daily.
- Social content (3 videos/week) was fully manual: recording, editing, captions.
What I built
- WhatsApp agent (Evolution API): answers procedural immigration questions with precision, qualifies and converts leads, serves existing clients, and hands off anything outside its scope to the partners.
- Email automation: incoming mail triaged, responses drafted for review.
- Human-in-the-loop learning: every answer the partners correct feeds back into the bot's knowledge — it hardens itself with real usage instead of drifting.
- AI social pipeline: partner records a video and uploads it; the system edits, produces the reel and prepares the post. Server-side approval gates (CAS on PostgREST) freeze approved content against accidental modification.
- One admin dashboard operating all channels: WhatsApp, email, social.
Results
| Metric | Value |
|---|---|
| WhatsApp volume | ~10 conversations/day (~300/month) |
| Content production time | ~90% reduction |
| Partner time recovered | 7-10 hours/week |
| Engagement model | Retainer client (recurring revenue + growth partnership) |
Stack
Next.js 16 · TypeScript · Supabase · Evolution API · Mailgun · Vercel · Tailwind v4