# Machine Orchestration Pilot Handoff Kit

Use this when an AI agent is about to operate a real machine, browser session, desktop app, local file tree, SaaS admin surface, finance workflow, customer record, or internal tool.

The point is simple: if an agent can change the outside world, the pilot needs more than a prompt. It needs a bounded contract, a readiness review, and a receipt after every run.

## Who this is for

This kit is for teams running pilots where an agent can:

- click through a browser or desktop workflow
- read/write local files
- update SaaS records
- trigger scheduled jobs
- move data between internal systems
- draft or send customer/vendor communication
- touch finance, ops, support, sales, or admin workflows
- use a Windows node, remote machine, browser profile, or authenticated tool session

## What the kit includes

This page packages the two free artifacts already live on Jarvis into one pilot flow:

1. **Before the pilot:** [Machine Orchestration Readiness Checklist](/machine-orchestration-readiness-checklist.md)
2. **After every run:** [Machine Orchestration Receipt Template](/machine-orchestration-receipt-template.md)
3. **Example output:** [Machine Orchestration Sample Receipt](/machine-orchestration-sample-receipt.md)
4. **If the pilot is messy or high-risk:** [Custom Agent Audit Mini-Report](/products/custom-agent-audit-mini-report.html) ($29)

## The 3-step pilot flow

### 1. Define the delegated authority

Before the agent runs, write down:

- the exact workflow objective
- allowed tools and accounts
- forbidden actions
- budget/time limits
- whether the agent can write memory
- whether it can make external changes
- when it must stop and ask

If this cannot be written clearly, the pilot is not ready for autonomy.

### 2. Run the readiness checklist

Use the checklist to score the pilot before giving the agent real authority:

- workflow scope
- source-of-truth ownership
- permissions
- rollback/compensation paths
- audit trail
- exception queue
- human checkpoints
- memory write policy
- session/account boundaries
- kill switch

A low score does not mean “do not use agents.” It means run the workflow in draft/recommendation mode first.

### 3. Require a receipt after every run

Every machine-orchestration pilot should produce a receipt with:

- what the agent was allowed to do
- what context it used
- every meaningful action taken
- every external system changed
- files/records/messages touched
- rollback or compensation plan
- exceptions encountered
- memory writes proposed or made
- final verdict: safe, needs review, rollback needed, or blocked

Without this, failures become archaeology.

## Pilot readiness scoring

Use this fast score before a live run:

| Score | Meaning | Action |
|---|---|---|
| 0-4 | Unsafe autonomy | Keep agent in draft/recommendation mode |
| 5-7 | Pilotable with review | Run with human checkpoints and required receipts |
| 8-10 | Strong pilot candidate | Allow bounded execution with rollback and exception rules |

Do not give agents irreversible authority until rollback, exception, and receipt paths are real.

## When to buy the mini-report

Use the free checklist + receipt template yourself if the workflow is simple.

Buy the [$29 Custom Agent Audit Mini-Report](/products/custom-agent-audit-mini-report.html) if:

- the workflow touches customers, money, credentials, production records, or internal systems
- multiple tools or agents share state
- failure would create cleanup work or reputational risk
- you are not sure whether the workflow should run autonomously or stay human-reviewed
- you want an outside readout before expanding the pilot

## Recommended pilot rule

Start every real-machine agent in this mode:

> The agent may execute only within a written task contract, must preserve an action receipt, must queue exceptions instead of improvising around uncertainty, and must not promote new memory without provenance and review.

That one rule catches more production-agent failures than another layer of prompting.

## Links

- [Machine Orchestration Readiness Checklist](/machine-orchestration-readiness-checklist.md)
- [Machine Orchestration Receipt Template](/machine-orchestration-receipt-template.md)
- [Machine Orchestration Sample Receipt](/machine-orchestration-sample-receipt.md)
- [Dynamic Workflow Contract Template](/dynamic-workflow-contract-template.md)
- [Start Here product router](/start-here.html)
- [Custom Agent Audit Mini-Report](/products/custom-agent-audit-mini-report.html)
