Supporting page

Methodology

API is intended to be practical. The goal is not mathematical precision; it is repeatable assessment. This page outlines one construction method that organizations can adapt.

Step 1 — Define the process boundary

• Name the process (e.g., “Vendor invoice intake → approval → payment”).
• Define start condition and completion condition.
• List dependencies (systems, data sources, humans, approvals).

Step 2 — Write the success definition

• What does “done” mean?
• What quality checks must pass?
• What constraints must always be respected (policy, security, compliance)?

Step 3 — Score each component

Use a simple 0–5 rubric for each component from Components.

• 0: not present / purely manual
• 1: basic support, mostly human-driven
• 2: partial capability with frequent intervention
• 3: works end-to-end in normal cases
• 4: handles common exceptions and verifies outcomes reliably
• 5: robust, continuous, governed autonomy

Step 4 — Derive an overall level

• Minimum rule: the overall API level cannot exceed the lowest critical component.
• Critical components: execution, verification, exception handling, and governance.
• Optional weighting: if you weight components, document the weights and keep them stable.

Step 5 — Record evidence

• Attach logs, test results, runbooks, and example runs.
• Document known failure modes and when humans must take over.
• Re-score on a cadence (e.g., quarterly) to track progress.