MEETU_CHARSIEWPAO

Inspiration

Advisors keep losing clients because they drop the thread. A client mentions a new baby, a job change, or an upcoming trip, and that detail vanishes into a chat log while the CRM sits empty and waiting for manual entry. We built MEETU so the remembering happens on its own.

What it does

MEETU gives independent financial advisors a workspace with six tabs: activity, leads, meetings, client profiles, a news radar, and a memory graph. It reads incoming WhatsApp messages, pulls out facts and suggested touchpoints, adds news and social context to each profile, and stores everything in a knowledge graph the advisor can browse.

How we built it

We built the frontend on Next.js 16 with the React 19 compiler and Tailwind v4, and ran the backend on Convex for the real-time database and AI actions. The Vercel AI SDK drives the analysis across OpenAI and Kimi, WhatsApp arrives through the Meta Cloud API and a Baileys bridge, and Apify and mem0 handle scraping and long-term memory.

Challenges we ran into

We worked on a bleeding-edge Next.js 16 and React 19 setup where the conventions broke our assumptions, so we checked the docs for almost every change. Wiring stable WhatsApp ingestion took real plumbing, and getting reliable structured output from the model without hallucination took careful schema work with Convex and Zod.

Accomplishments that we're proud of

We shipped a working end-to-end path: a real WhatsApp message flows in, gets analyzed, updates a client's memory, and shows up as an action the advisor can take. The product also looks the part, with a warm editorial interface and a live memory graph that makes the AI's knowledge visible.

What we learned

We learned that AI earns its place as a quiet layer handling tedious work, and that structured outputs with confidence scores beat raw model size for a real workflow. Convex cut out a large amount of glue code, and a calm interface did more for trust than any single feature.

What's next for MEETU_CHARSIEWPAO

Next we want two-way messaging with one-click AI-drafted follow-ups, plus email, Instagram, and LinkedIn feeding the same memory graph. After that we plan automatic meeting briefings, team workspaces with proper auth, and the compliance, privacy, and retention controls regulated advisors need.

Built With

• apify
• bun
• claudecode
• codex
• convex
• mem0
• next.js
• tailwindcss
• typescript