assay · audience · for the people deciding whether to adopt it

What it changes about control, and where the evidence lives.

A decision to run agent delivery under Assay is usually made by a small group reading at different altitudes: one owns the transformation story, one owns the attestation that reported work happened, and one has to make the whole thing survive an examiner in the codebase. Assay by Medici is an agent-operated delivery pipeline with a human-governance layer, and that governance layer is one control story told to all three. This page routes each reading to the page that answers it, and hands the technical evaluation to the surfaces built for it.

Exhibit · the control story

You own the AI transformation story

Your exposure is that one defensibility failure in agent-produced work undoes the story, and that the transformation stalls because oversight of agent output does not scale past human-paced review. Assay is the pipeline that scales it, and its governance layer is what holds: verification run by a party that did not implement the work, on a record that is append-only and shows tampering. What earns this reader is the precise claim rather than an overclaim — a claim that survives a hostile reading is the one worth telling a board.

Your reading

  1. The methodology

    What the control is and refuses to be.

  2. The live board

    The operated evidence you can point to.

The control story →

Exhibit · the attestation

You are accountable for attesting that reported work happened

You must be able to say that what the board shows actually occurred. Assay's governance layer is segregation of duties for agent work, in the vocabulary you already use: the implementer never verifies, verification is executable, and the registers are append-only with tampering made visible. The claim is stated at its exact limit — the board is derived from agent-authored artifacts with consistency linting, not measured from ground truth — which is the form an attestation can stand behind, because a reader can check it and find it true.

Your reading

  1. The methodology

    The control model and its residual gap.

  2. Tiering

    How the pilot starts at no license cost.

The control model →

Exhibit · the examination

You have to make agent delivery survive an examiner

You are the technical evaluator the sponsor routes the decision to, and the quiet veto if it smells like ceremony, lock-in, or a claim an auditor would shred. Assay extends the change-control and separation-of-duties regime you already run to work an agent authored, rather than standing up a parallel one: it is documents in your own repositories plus one binary in your CI, sitting under whichever code host and CI you keep. The limit is stated in the same breath: it is attribution enforced by the host, not proof that a review was diligent. The pipeline runs from intake to verified on main; release, deploy, and rollback are not in it today, and they remain yours.

Your reading

  1. The methodology

    The control, before the setup.

  2. Install and GitHub Apps

    The setup as it stands.

  3. The working detail

    On the operator path: who reviews, who verifies, and what installing costs.

The control → The working detail