Blog
What we learn building, shipping and running AI agents on Keel.
12 November 2026
Where AI actually helps a 50 person business, and where it wastes a quarterThe jobs AI does well in a small company today, the ones it does badly, and the order that gets a working system into production without a data team.
12 November 2026
What is RAG and why does it stop AI making things up?Retrieval-augmented generation in plain English: why chatbots invent answers, what retrieval adds, when RAG is the wrong tool, and how to test a system in 20 minutes.
12 November 2026
Why most AI pilots never reach real workMost AI pilots stall on sample data, missing connectors, no guardrails and no owner. How a 48 hour proof of concept on your own data changes the odds.
12 November 2026
What is Keel? The agentic AI platform behind InfrenicKeel is Infrenic's agentic AI platform: agents, knowledge with sources, connectors, approvals, hardware and monitoring in one core, so new AI apps ship in days.
12 November 2026
The AI agent production checklist: 10 things to have before launchA practical checklist for putting an AI agent into production: data access, sources, connectors, guardrails, approvals, evaluations, monitoring and cost.
12 November 2026
How Keel does RAG: hybrid search, local embeddings and answers with sourcesInside Keel's retrieval pipeline: recursive chunking, embeddings computed locally, pgvector plus full text, reciprocal rank fusion, and an agent that must cite or refuse.
12 November 2026
A help desk that answers from your own documentsA RAG use case on Keel: cited answers from your help docs, AI triage on everything it cannot solve, and a staff inbox that keeps a person in charge.
The Infrenic AI Dispatch
Practical architecture notes on AI agents, RAG, and production systems. No buzzwords, no spam. Sent when we have something real to share.
Read earlier editions at infrenic.beehiiv.com