Case study : from scattered documents to a pilot
What an applied-AI team can build for a sales organization : a prototype enablement agent, what went wrong along the way, and a starter kit for the next team.
Posts
Longer writing, newest first. 39 in total, page 1 of 4.
2026
Putting an agent next to a sales rep --- agent architecture, data model, security/privacy, and cost structure
The harness, data model, security and privacy points, and cost of a real-time sales assistant built with Google ADK, and the break-even it has to clear.
Turning narrated demos into scalable agentic knowledge
Instead of asking a team to watch recordings they will forget tomorrow, I turned two narrated demos into knowledge an agent serves at the moment of need.
The knowledge nobody has at the moment they need it
Tips, documents, onboarding and on-the-job training each cover part of a task. Most enablement still misses the moment of work.
Vector, graph, or a curated wiki --- how a purpose-built agent can get its knowledge
I came to curated knowledge from vector search and direct documents. Here is how I would choose between them for a purpose-built agent.
Markdown files an agent can trust --- an introduction to LLM Wiki and the Open Knowledge Format
A plain-language introduction to the LLM Wiki pattern and the Open Knowledge Format, and how a small agent can use a folder of Markdown files as its knowledge.
What AI Builders Inside Large Orgs Actually Deal With When They Are Not a Part of Development, Engineering, or Technology Teams
Building AI prototypes inside a large enterprise is a political challenge as much as a technical one, and the patterns that sideline good ideas are predictable.
Code-as-Video for Learning Content
A running experiment in treating learning and product video like code: versioned, reproducible, and re-rendered on demand with an agent in the loop.
Local AI Lab --- The Full Stack of a Local Claude Cowork and Code Alternative I Have to Build Myself
An early, honest review of the finished local stack: how close it gets to Claude Cowork and Claude Code, and where it falls short.
The Proxmox Local AI Lab
The map of the Proxmox AI Lab: local inference on two GPUs, a coding agent, and a gateway reachable from my phone, with no cloud AI in the loop.