Notes on AI Centers of Excellence from n8n In the Loop 2026

Notes on AI Center of Excellence from n8n In the Loop 2026 conference in Berlin

Yesterday I attended an insightful session on AI Center of Excellence (AI CoE) with Mick McGrath from Bitovi.

Here are a few takeaways relevant to almost any company exploring AI - not just those using n8n.

(Notes taken by an actual human: me. No AI note-taker involved 🙂)

Why to have AI CoE - To avoid duplication of efforts, waste of resources, inconsistent standards, no clear traceability

Use Case Intake - Each use case should have an owner - Every use case should ship with agreed KPIs defined at intake - Velocity of the feedback is important - Use cases are to be built ideally model-agnostic

AI Enablement Trainings - Org-wide training is required to make sure people can build simple things themselves - Make AI Adoption Workshops (where teams talk to each other, sometimes for the first time) - Concept of AI Captains - trained people from within the company who help others - Make workshops for cross-polination of ideas between the teams

Model Selection, AI Gateway - AI Gateway controlling usage of the models and token consumption needs to exist - There should be a list of approved models - Per-workflow token budgets to be there - Categorize automations into the Risk Tiers

Interesting (no category) Human ON The Loop (HOTL) concept is gaining traction, while Human In The Loop may slow down the automation and sometimes needs to be converted into HOTL