AI for Product Management: Build Your Own AI Assistant

A chat assistant forgets you the moment you close the tab. The one you build today doesn’t. It knows your role, products, stakeholders and decisions, and it gets better the longer it runs: a personal assistant who becomes an expert in anything you need.
You design it first, with our guided template: what it should know, what it should do, what it may never do, and what stays local. Then you build it: a coding agent pointed at a folder of markdown files (Karpathy’s LLM wiki pattern, for example), with Obsidian as your window into it. You compile it from your own material, give it memory that lives in your files, and add three skills, one provided, two your own.
In the afternoon you connect it, read-only first, and put sensitive material where only a local model can read it. You leave with a running assistant you own, and a routine to run it.

Design first, then build.
The morning is the assistant’s brain. You design it with our guided template, then set up your agent, your folder and the schema file that tells the assistant who you are and how to behave. You compile your wiki from your own material, talk to it, and fix what it got wrong.
The afternoon is skills, connections and boundaries. You add the Value-First skill and build two of your own from a skeleton. You connect a search layer, then your tools, read-only. You put sensitive material in a folder only a local model can read. You learn what dreaming is and how to switch it on later, set your Monday routine, and demo the result.

Product Managers, Product Owners, Heads of Product, Scrum Masters and Agile Coaches at organisations of any size, freelancers included, with the 2-day training as foundation, or equivalent knowledge, who want an assistant that knows their world and stays theirs.
Bring a laptop you can install on, a coding agent, paid (Claude Code, Codex) or self-hosted (Hermes Agent), and material you’re allowed to use (personal notes suffice). A local model if your laptop can run one.

A rented assistant knows what its vendor (such as OpenAI or Anthropic) lets it know and keeps your context on their servers. This one knows your products, your people and your decisions, in plain text on your computer, with any model you choose. Switch models next year and your assistant comes with you.

A markdown wiki for knowledge (Karpathy’s pattern, for example), skills in the open standard, MCP for connections, a search layer for scale, dreaming for memory consolidation. That’s the architecture behind the strongest personal agents today, in a version you understand because you built it.

A private folder only your local model can read, so cloud models never see it. Connections read-only by default. Privacy you set up yourself and can explain to your security officer in two sentences.

After registration you’ll receive a confirmation with practical details and the pre-work: install your coding agent and Obsidian, gather the material you’re allowed to use, and optionally install LM Studio (or an equivalent) with a local model if your laptop can run one. Lunch and snacks are included. You’ll receive the digital materials, the design template, the skill skeletons, the Value-First skill and the dreaming how-to guide, plus access to the alumni community.

See you soon!