Cut missed client requests to near zero for a 15-person agency by logging every channel in one place
Client snapshot
A 15-person marketing agency with about 30 active clients. Account managers handle requests that come in by email, WhatsApp and phone. No ticketing system; clients would never use one.
The problem
Requests arrived everywhere and were tracked nowhere. A client asked for a change on WhatsApp on Friday night, an account manager saw it Monday, and it lived in their head until it didn't. When a client asked "where are we on the three things I sent last week?", the answer took fifteen minutes of scrolling through three apps. Every month a handful of promised changes slipped, and the agency usually heard about it from the client.
What we built
One request log, fed from every channel.
n8n watches the shared inboxes and the agency's WhatsApp Business number. Claude reads each message and decides whether it contains a request. If it does, it writes a row to Supabase: client, what was asked, who asked, the channel, the date promised if one was mentioned, and a link back to the original message. Call notes go in through a two-field form after each call.
Each morning the account managers get a short list: requests due today, requests overdue, and requests that came in overnight with no owner. One click assigns an owner.
The log is also exposed through an MCP server, so the team can ask questions from the AI chat tool they already use: "what's open for this client?", "what did they ask for last month that we haven't done?". Claude can draft a status reply from the log, and an account manager reads and approves it before it goes anywhere.
Results, 90 days in
- Requests logged without anyone typing them in: about 40% → over 95%.
- Requests that passed their promised date: 15 to 20 a month → 2 or 3.
- Answering "what's open for this client?": about 15 minutes → a few seconds.
- Status replies to clients drafted in seconds and approved by a person before sending.
Quote
"I used to dread the 'where are we on this' email. Now I ask the log and send the answer before my coffee is cold." Account manager
Build snapshot
- Timeline: 14 business days
- Stack: n8n, Gmail, WhatsApp Business API, Supabase, Claude, MCP server
- Investment: $5,500
- Who owned it: the head of client services reviews the overdue list daily and owns the request categories.
The pattern
This is the pattern behind the Operator's Dashboard engagement: read across the surfaces where work shows up, and put the things that need a person on one list. If your requests arrive in four apps, the build is the same, with your channels swapped in.
This is an illustrative composite drawn from operator engagements. Names, identifying details, and exact metrics are anonymised. The workflow, stack, and approach are real.
Looks like your situation?
A 30-minute call tells us whether the same pattern fits your operation, and what it would cost to ship it.