Business problem
A welcome email is a small part of onboarding, but it sets the tone for what comes next. Writing one for every new client means repeatedly gathering account details, adapting the message, and recording the follow-up. I wanted to see how much of that repetitive work I could connect into one reliable workflow.
Existing manual process
The manual process this prototype addresses is straightforward: review the client's name, company, products, and onboarding notes; write a relevant welcome email; send it; then update the client record. Each handoff takes attention and creates another opportunity for details or status updates to be missed.
Solution architecture
I built a working prototype using Airtable, Zapier, the Claude API, and Gmail. A new Airtable client record starts the workflow. Zapier passes the account context to Claude, which generates a personalized welcome email. Gmail sends the message, and the workflow updates the Airtable record.
- AirtableNew client record
- ZapierCoordinate workflow
- Claude APIGenerate email
- GmailSend email
- AirtableUpdate record
I mapped the fields, configured the connected steps, refined the prompt, and worked through response parsing so the generated content could move between tools.
Business impact
Expected operational impact, not measured production results.
The workflow is designed to reduce repetitive writing and record updates, make welcome communications more consistent, and shorten the gap between client intake and initial outreach. That could give implementation teams more time for discovery, kickoff conversations, and resolving customer blockers.
To validate the benefit, I'd compare manual effort per email and time to first outreach, while tracking corrections and failed runs. This was an independent portfolio build, with no claimed customer volume or measured time savings.
Lessons learned
The details between tools matter. Raw API responses needed parsing before downstream steps could use them, and even trailing spaces in JSON keys caused problems. Prompt quality also shaped how useful and relevant the email became.
My biggest takeaway: a good automation needs dependable inputs and outputs as well as useful writing. Before broader use, I'd add required-field checks, duplicate-send protection, failure alerts, and an explicit review step for customer-facing messages.