Hypotheses you can inspect and test.
A case shows the architecture a problem needed; a decision record says why one choice holds under its premises. A lab isolates one pattern, runs it on synthetic data and exposes how it behaves, so the hypothesis can be checked rather than taken on trust.
01Experiments02 labs
- The question
- What can we sell to this customer, under which conditions, and why?
- Hypothesis
- Keep eligibility, conditions, fulfilment and authority as separate decisions, each with its owner, and one offer can be resolved deterministically—and explained.
- What you can inspect
- The premises and the tensions between them
- Plausible designs, and where each one breaks
- A resolver with the trace behind every answer
- What-if changes, and who owns each exception
- The question
- How far can AI take an unstructured commercial request before a person has to decide?
- Hypothesis
- AI should assemble the decision surface—not make every decision.
- What you can inspect
- The path from an incoming email to a reviewed record draft
- What is automated, what is reviewed and what needs authority
- What the experiment has built so far
Both experiments run on synthetic data and describe no real organization.