The Four Questions
The gap between what an agent can do and what an institution will accept is itself a finding.
One test for every note
No experiment runs in this Lab for curiosity alone. Each one must move at least one of four questions, and a note that moves none of them is not published. The questions are few on purpose: a programme that asks everything answers nothing.
The delegation boundary
Which acts of each practice can an agent perform to a standard the practitioner would sign — and which collapse without the human? Not in general, but act by act: triage, drafting, assurance, the preparation of decisions, the reading of rooms. The answer will differ across Transformation, Portfolio, Programme, and Leadership, and the differences matter as much as the verdicts.
The direction discipline
Building and directing agents on real work: is it a repeatable discipline with methods and failure modes of its own, or one person’s knack? Every instrument built in this Lab — every pipeline, every skill, every orchestration that works or breaks — is evidence for one side of that question.
The atrophy question
When drafting, triage, or assurance is delegated, some judgment goes unexercised. Does it degrade, or does it move somewhere higher? Each note in this Lab records what the practitioner did not have to think about that run. At intervals, the Lab stops and asks the harder cumulative version: after all these cycles, what am I no longer exercising — and would I notice?
The trust boundary
There is a second boundary behind the first. An output can be signable and still unacceptable: a sponsor, a board, an SRO may refuse an agent-prepared basis for a call regardless of its quality. The gap between what an agent can do and what an institution will accept is itself a finding — and it may prove the wider of the two.
Why these four
Together they cover the work, the worker, and the world the work answers to: what can be handed over, what handing it over does to the one who directs, and what the institutions around the work will bear. The grid is scoped to the four practices; the questions are not — an agent built for any real work the practitioner does is admissible evidence, and the discipline either transfers across domains or it was never a discipline. Every verdict on the grid traces back to at least one of them.