What we think is actually true about this.
We have no client references to show you yet, and we are not going to invent any. What we can show you is how we think about the problem.
- 01
Your ontology matters more than your LLM
Model choice is the most discussed and least durable decision in an AI programme. The model will be replaced within the year. The model of your business will not, and it is the thing that decides whether any model produces work you can use.
- 02
Why an AI agent should not decide its own authority
Every vendor selling agents describes what they can do and almost none describe what has to be true before one is allowed to act. Authority should belong to the workflow, granted per action, earned on measured evidence, and revoked when quality drops.
- 03
Why the first AI workflow is the expensive one
Most AI budgets buy episodes rather than accumulation, which is why they produce nothing durable. The first workflow pays for connection, resolution, permissions and evaluation. The second inherits all of it, and that asymmetry is the entire investment case.
Written up properly and published here shortly.