AI Automation Stack: Zero to Production

SwissAware runs cybercrime prevention programs across Switzerland with a lean team. Every week, that team spent hours on the same recurring tasks: sorting and routing inbound messages, drafting standard responses, generating reports, and updating task lists manually.

There was no AI in place. No automation. Staff time was being consumed by work that required no judgment, leaving less time for the high-value work that actually required human expertise.

The additional constraint: the team was not technical. Any solution had to be simple enough for non-developers to operate and maintain independently after handover.

I scoped the automation in layers, starting with the highest-volume pain points and building outward. Rather than buying a single tool and fitting the org around it, I mapped every recurring workflow first, then matched tools to each workflow specifically.

Layer 1, Conversational agents: Deployed Gemini Enterprise-powered agents to handle standard inbound queries and route complex ones to the right person. Trained on internal FAQs and response templates.

Layer 2, Email triage: Built n8n workflows that classify inbound email by type, urgency, and sender, then route automatically. Urgent items get flagged and escalated without human review of the inbox.

Layer 3, Document generation: Connected Notion AI to existing templates. Staff trigger generation via a simple form input. First drafts of standard documents now take under 2 minutes instead of 30.

Layer 4, Task management: Zapier workflows sync task status updates across tools and notify the right person. No more manual status updates or cross-checking.

Training was built into the rollout. I ran hands-on sessions with every team member, including citizen developers with no technical background, until each person could modify and extend their own workflows independently.