I am Johan Wallquist, a Partner Technology Strategist in Stockholm. I work mostly with AI and AI-first ways of working with partners. This is where I share what I learn and the things I build.
People who learn to work with AI will get an edge. You need exposure to get there: see how other people use it, try it yourself, then build something.
Case contents 17 published projects
Three ways into the work. Each one does a different job, so each one is packed differently.
Workshop
A consulting firm leaves with decisions about how AI changes the way they sell, deliver, and package what they know.
Live series + course
Start with AI Operator Intro and choose your next step: setup, the course or a talk. Operator Hour adds live sessions and public recaps.
Open to browse. Built for partners.
An almost completely autonomous resource library. Agents find the material and route it for review. People still decide what gets published.
Find learning material, skills and agents you can borrow, or talks and workshop material you can reuse.
Showing 19 of 19 projects
Open workshop material for consulting firms. The programs and exercises are made for Microsoft partners working on AI-first delivery.
An introduction to AI-first work with Copilot CLI. Choose how to get started, or browse the public recaps from Operator Hour.
A 20-minute introduction for partner consultants who want to turn messy notes and repetitive work into usable customer material with Copilot CLI. No coding required.
An almost completely autonomous library built for partners. Agents find and maintain the material, while people decide what gets published.
Watches one window you pick and tells you what changed. For example, use it to move through many pitch decks with a team and discuss each one without recording the rest of your desktop.
Researches a public guest, keeps a source ledger, and builds the question list. Every claim gets checked before you record.
Learns your voice from writing you already own, then flags the sentences that sound machine written. The private source text stays out of the repository.
Draws a map of your own agent setup: skills, agents, connected servers, and projects. Output is a single HTML page you can read in a browser.
A starter for a knowledge library on any topic. Agents research new material, maintain the collection, and run quality checks. People only approve major changes.
Build is Microsoft's annual developer event. This kit uses five prompts and a small skill to find the sessions and announcements that matter to you. The same approach works for any large event.
Frontier Consultancy. Repeatable work is how your practice scales.
You can make a useful first attempt earlier.
Consulting delivery changes when agents work across the assignment. The volume moves to the machine. Judgement, evidence quality, and the recommendation stay with the consultant.
A field report on how people and agents work together, what changes inside the organization, and what the audience can try immediately.
Seven levels of agentic work, from familiar assistance to systems that keep operating. The hard part is deciding when trust has been earned.
The deck workflow I use, from brief and research through outline, build, visual review, and publish. Ships with themes, checks, and a live demo.
A blueprint for a local assistant with layered memory, scheduled work, and hard limits on what it may send. Includes synthetic templates and leak checks.
Scouts your sources, scores ideas against the topics you write about, and prepares LinkedIn drafts in your voice. You still press publish.
A personal ChatGPT coach that combines Oura recovery, Polar training, and a simple strength log. Built to turn health data into a useful decision.
Nothing packed under that label yet. Pick another filter, or open the full list on GitHub.
Field notes
Agent tools change names often. Here is what the three I use are, without jargon, so you can tell which one you are looking at.
A shared workspace where you hand a task to an agent and it works on documents with you. You can upload a skill file so it follows your way of doing things.
An agent that picks up skills from a folder you point it at. Good when you want the same instructions available every time, without setting them up again per task.
The agent in a terminal window, sitting next to your files. Most of what you see on this page was built here, because the agent can read the project and run things itself.
A plain text file that tells the agent how you want something done. The file itself moves between all three. Scripts and connections usually need small edits per tool.
The source is on GitHub. Use it to learn, borrow the skills and agents, or adapt the patterns for your own work. Tell me if you make something useful.
Public does not always mean open source. Projects marked with a license are yours to reuse under those terms. The rest are there to read, learn from, or ask about.
Published repositories have been scanned for private links, tenant identifiers, local paths, and common credential patterns. If you want to reuse something without a license, ask first.