Good automation framework = good framework for AI.

Pack adds an agentic AI layer to any service on the NetOrca platform. AI fails when inputs are unstructured and outcomes aren't verifiable. NetOrca changes are already schema-defined, context-backed, lifecycle-managed, auditable, and verifiable. Pack is what becomes possible once the structure exists.

Use when audience asks about AI or has an AI-first mandate

B.1 — The three Pack capabilities

Config. Verify. Execute. A self-healing loop.

01 — CONFIG

Reads the docs, writes the config.

An AI model reads API documentation and generates structured configuration against the service schema.

02 — VERIFY

A second model checks the first.

An independent AI model validates the output — the schema is the contract, and verification runs before any change touches infrastructure.

03 — EXECUTE

Deploy with a healing loop.

Automated deployment with error detection and correction built in — if the live state drifts from declared state, the loop reconciles it.

The self-healing loop, visualised

B.2 — Why Pack works on NetOrca specifically

AI fails on free-text tickets. It works on schema-defined changes.

What AI needs to work well

Any agent is only as reliable as its inputs and its ability to verify outputs. Without structure, you get plausible-sounding wrong answers.

Structured inputs, not free text
Context the model can ground on
Verifiable outputs before execution
Lifecycle — not just "create"
A full audit trail

What NetOrca already provides

Pack isn't bolted on — it becomes possible because the substrate already has the properties AI needs.

Schema-defined services
Git-native context (intent, history, reviewers)
Verification gates before deploy
CREATE / MODIFY / DELETE as first-class
Every change logged end-to-end

Universal Executor

If it has an API, Pack can automate it.

The Universal Executor extends Pack across any infrastructure with an API — removing the need to pre-build a bespoke integration for every vendor.

Pack is an add-on, not a replacement. Use it where it adds value. Use traditional automation everywhere else. Pack deliberately sits in the appendix — leading with AI is what breaks current decks.