An automation platform and a production AI agent
Self-hosted n8n in AWS, grown from 5 to 35+ workflows running 10,000+ executions a month and replacing roughly 600 to 750 hours a year of manual work. The same platform hosts a Claude agent that operations, care and marketing use every day.
Context
When engineering at Tembo Health consolidated to one engineer, the company still depended on custom code for client reporting, internal notifications and cross-system data sync. Clinical operations and sales were spending 12 to 15 hours a week on manual work, and every new request landed on engineering. Writing custom code for every workflow was the bottleneck, so the way the company operated had to change.
What I did
I deployed self-hosted n8n inside the HIPAA-compliant AWS environment, starting with about five workflows, and worked directly with clinical operations and sales to find the highest-value manual work and move it onto the platform.
On top of it I designed and shipped Victoria, a production AI agent that answers cross-source data questions through Slack. It runs on Claude via Amazon Bedrock inside the same HIPAA-compliant AWS environment, and non-technical staff use it the same way they talk to each other. I chose n8n’s agent capabilities over a dedicated agent framework on purpose: the agent stays inside infrastructure I already operated, debugged and monitored, instead of adding a new framework with its own operational model.
Result
The platform runs 35+ workflows and 10,000+ executions a month and has replaced roughly 600 to 750 hours a year of manual work, about a third of a full-time role. Data requests that took days became Slack answers that take minutes. Marketing sources growth and impact stories from the agent’s output for public communications. Engineering stopped being the bottleneck for every new request.
Stack
n8n · Claude via Amazon Bedrock · Slack · AWS · HIPAA-compliant deployment