Forem: Cess Mbugua

Day 3 of 60 — and today's build is one I think a lot of freelancers and agencies will relate to.

Cess Mbugua — Fri, 01 May 2026 08:06:47 +0000

I built a fully automated client onboarding pipeline that kicks in the moment a new client submits an intake form. No manual steps. No welcome email to write. No folder to create. No team to brief. All of it — done in under 5 minutes.

The Problem I Wanted to Solve

Manual client onboarding is one of those hidden time drains. Writing a personalised welcome email, creating a Drive folder, generating a project brief, notifying the team, each step feels small but together they eat 45–90 minutes per client. And if you're busy, something always gets missed.

I wanted to learn how to wire together form submissions, AI writing, CRM updates, file creation, and team notifications into one seamless pipeline.

What Happens Automatically

Step 1 — Form Submission
Client submits a Typeform intake form. That single action triggers everything.

Step 2 — Claude Writes the Welcome Email
Claude API (claude-sonnet-4-5) reads the form data and writes a fully personalised welcome email — not a template, an actual tailored message.

Step 3 — Airtable Record Created
A new client record is created in Airtable with all intake details logged.

Step 4 — Welcome Email Sent
Gmail delivers the personalised welcome email to the client.

Step 5 — Google Drive Folder Created
A dedicated client folder is automatically created in Google Drive.

Step 6 — Claude Generates a Project Brief
Claude writes a structured project brief as a Google Doc inside the client folder — ready for the team to act on immediately.

Step 7 — Airtable Updated
The client record is updated with the Drive folder link for easy access.

Step 8 — Team Notified on Slack
The team receives a Slack notification with the client details, folder link, and project brief — everything they need in one message.

Results

Onboarding time: from ~60 minutes manual → under 5 minutes automated
Human effort after form submission: 0 minutes
Handles unlimited simultaneous submissions without breaking a sweat

Stack

n8n · Claude API (claude-sonnet-4-5) · Typeform · Airtable · Gmail · Google Drive · Slack

What I Learned on Day 3

The most interesting challenge today was getting Claude to write content that felt genuinely personalised and not templated — twice in the same workflow. The welcome email and the project brief need completely different tones: warm and human for the client, structured and professional for the team.

That means two very different prompts, each pulling specific fields from the Typeform submission. Getting the data mapping right between Typeform → n8n → Claude took more iteration than I expected.

Biggest lesson: one workflow, multiple Claude calls — each prompt needs its own clear context and purpose or the output bleeds into the wrong tone.

🔗 Full workflow on GitHub → https://github.com/mbuguacessy-glitch/automated-client-onboarding

Day 3 of my 60-day public AI automation build challenge. Learning by shipping — every workflow documented openly.

57 more to go.

ai #automation #n8n #claudeapi #buildinpublic #learninpublic #showdev

Day 2 of 60 — Automated Client Reporting System

Cess Mbugua — Thu, 30 Apr 2026 11:54:04 +0000

I'm a student learning AI integration and automation by building one real workflow every day for 60 days. Everything gets documented and pushed to GitHub.

**
Today's build: Automated Client Reporting System**

Built a workflow that writes and delivers professional weekly performance reports to every active client, every Monday at 7am, automatically, forever.

What I learned: Getting Claude to write consistent, structured narrative reports across different client datasets is a prompt engineering challenge, not just an API call. Tone, section structure, and handling flat or missing data without hallucinating trends all needed explicit instructions.

The loop logic hit different too — one mapping error and the wrong client gets the wrong report. Zero margin for error.

How it works:

Schedu
le trigger fires every Monday 7am

Airtable returns active client list

Loop pulls each client's metrics from Google Sheets

Claude writes a 5-section narrative report per client

Google Doc created → Gmail delivers to client → Airtable logs it

Biggest lesson: Test with dummy data obsessively before anything touches real client data.

Stack: n8n · Claude API · Google Sheets · Google Docs · Gmail · Airtable

🔗 https://github.com/mbuguacessy-glitch/automated-client-reporting

58 builds to go 💪

Day 1 of 60 — Learning AI Integration in Public

Cess Mbugua — Tue, 28 Apr 2026 20:19:18 +0000

I'm a student learning AI integration and automation by building one real workflow every day for 60 days. Everything gets documented and pushed to GitHub. Starting today.

What I learned: you can use an LLM to drive actual routing logic in a workflow — not just generate text. Claude reads each email and returns a full JSON decision object that n8n acts on instantly.
How it works:

Gmail triggers n8n on every new email
Claude API classifies it — urgency, category, sentiment, summary & suggested reply
Switch node routes it:

Urgent → instant Slack alert
Standard → Airtable + Slack
FAQ → auto-replied by Claude, zero human touch

Biggest lesson: getting Claude to return consistent, parseable JSON every time is harder than it sounds — and if it breaks, everything breaks. Prompt structure matters more than I expected.
Stack: n8n · Claude API · Gmail · Airtable · Slack
🔗 https://github.com/mbuguacessy-glitch/ai-support-ticket-router
59 builds to go!

learninpublic #buildinpublic #automation #ai