Notes on making
AI useful.
The ideas behind Ortelian: how companies become AI-native, how agents understand the work and what we learn from building.
- Map first, tools second Agents get the tools we built for people and rebuild a map of the company on every run. A world model for knowledge work gives them the map to query, and the tools feed it.
- Is a world model for knowledge work a knowledge graph? Partly. A knowledge graph is its structure. A world model for knowledge work adds time, sources and a line between accepted facts and guesses.
- What is a context graph? A context graph connects what an agent needs for a task: the things involved, their current state and past decisions. It only holds up on a world model for knowledge work.
- What the EU AI Act asks of B2B companies The Act asks little of the AI work that creates most of the value and a lot of AI that decides about people. Here are the dates, the tiers and what we would do now.
- One week at one company A week after the first call, a sales team works from a list that refreshes itself, a meeting-prep agent they check before every call and a territory model a new rep can ask for a list from.
- The case for GraphRAG Some questions have no single document that answers them. GraphRAG builds a connected view of the evidence once and reuses it, for broader answers than plain retrieval without rereading everything.
- Agree the standard before you build The deployments that last agreed quality, cost, time and human effort on real examples before anything was built. The pilots that die never agreed what good meant.
- Redesign the work before you automate it An agent on the current process makes the current process faster, waste included. So the first two weeks of a deployment redesign the work, and only then do we build.
- More context makes agents worse Past a point, adding context degrades an agent. It needs the smallest sufficient view of the work, and only a model of the company can decide what that view is.
- Why coding agents work and company agents don't Coding agents took off first because a repository is already a world an agent can stand in. A company is not, so you have to build that world before asking the agent to work.
- The model can be rented. The world it works in cannot. Every company rents the same intelligence at a falling price. The advantage that lasts is the maintained model of your own business that the intelligence works in.
- What sight means Connected facts show what no single source contains. That is the real payoff of a world model, and the larger bet we are making.
- What is a world model for knowledge work? A world model for knowledge work is a map of what shapes a company's work, inside and outside the company, that agents query instead of rebuilding it from tools. It has three layers: data, graph and knowledge.
- What is an ontology? An ontology says what exists in a business, what each thing means, how things relate and what can be done with them. The graph then holds the particular companies and people.
- Access is not a map Connecting an agent to the CRM, Slack and email gives it territory. The map, what those things are and how they relate, still has to be built.
No notes in this topic yet.
Follow the work.
Notes from Ortelian, by email.
Subscribe to receive new notes. Unsubscribe at any time.
Tell us where the work gets stuck.
You don’t need to know yet what AI should do. We’ll work out with you where to start and whether a two-week audit is worth it.