AI Automation
Tool choices, failure modes, and the tradeoffs between n8n, Make, custom agents, and plain code — and how to build automation that survives production.
MCP Servers: The Missing Glue in Your AI Automation Stack
Most AI automations wire tools together at the application layer — custom code, webhooks, platform connectors. Model Context Protocol offers a different approach: a single, standardised way to connect AI models to any tool or data source. Here's what that actually changes for teams building automations in 2026.
Matti Ilvonen·30 Aug 2026·4 min readWhen to Build an AI Agent (and When a Simple Automation Is Enough)
Every workflow tool vendor now has an "agent" mode. Every consultant is pitching agents for processes a basic n8n flow would handle in thirty minutes. Here is how to tell the difference before you build the wrong thing.
Henri Parkkonen·16 Aug 2026·5 min readWhat Nobody Tells You About Running AI Agents in Production
Your AI agent works flawlessly in the demo. Six weeks after deployment it's producing inconsistent outputs, burning through your token budget, and nobody can tell you why. Here's what production reliability for AI automations actually requires.
Henri Parkkonen·2 Aug 2026·5 min readWhy Most AI Automations Break in Production — and How to Build Ones That Don't
The demo runs flawlessly. Three weeks later someone is back to doing it by hand. The root cause is almost never the AI — it's the architecture around it. Here's how to build automation that survives production.
Matti Ilvonen·17 May 2026·5 min readAI Automation That Doesn't Break: How to Choose the Right Tool for the Job
The demo looked great. Three weeks later it stopped working and nobody could fix it. The standard AI automation failure has almost nothing to do with the technology — and everything to do with tool choice.
Matti Ilvonen·10 May 2026·5 min readOne letter a month. Nothing more.
What we actually shipped. What we did not. No marketing.