The AM-OS Playbook
Immediate execution guide for new members. Use this playbook to prioritize high-return automation, deploy your first agent roles, and keep quality high with a repeatable weekly cadence.
1. The 5 Business Functions to Automate First (Ranked by ROI)
Use this order unless you have a specific bottleneck forcing a different sequence.
| Rank | Function | Why It Pays First | Primary KPI | First Automation Target |
|---|---|---|---|---|
| #1 | Customer Support Triage | Removes constant founder interruptions and protects response SLAs. | Median first-response time | Ticket routing + drafted response with escalation tags |
| #2 | Content Production Pipeline | Creates consistent top-of-funnel output without daily manual writing. | Published assets per week | Research summary to first draft plus QA checklist |
| #3 | Revenue Reporting and Reconciliation | Gives weekly decision clarity and catches leaks before they compound. | Weekly net cash delta | Stripe export summary + variance alerts |
| #4 | Lead Qualification and Outreach | Increases pipeline quality and frees sales energy for high-fit conversations. | Qualified conversations booked | Inbound scoring + personalized first-touch draft |
| #5 | Client Onboarding and Fulfillment | Standardizes delivery and lowers churn from early-stage confusion. | Time-to-first-value | Kickoff packet + milestone reminders + handoff log |
2. RICE Prompt Framework
Every production prompt in AM-OS follows this structure:
- R - Role: Define identity, scope, and standard of expertise.
- I - Instructions: Give explicit, ordered actions and boundaries.
- C - Context: Provide business rules, audience, tools, and constraints.
- E - Examples: Show target output format and quality bar.
R: You are the Reporting Agent for a bootstrapped service business. I: Every Monday at 8:00 AM, summarize last week's revenue, refunds, top channels, and any anomalies over 12% week-over-week. End with three actions. C: Use Stripe exports + CRM pipeline report. Keep total under 250 words. E: Format as: KPI Snapshot, Trend Flags, Recommended Actions.
RICE Quality Gate
- Could a new team member run this prompt without follow-up questions?
- Is the output format explicit enough to automate downstream actions?
- Are "do not" constraints included for sensitive tasks?
3. Agent Role Templates
Support Agent Template
Mission: Resolve common support requests quickly while escalating risk cases. Inputs: customer message, order ID, policy doc, order history. Process: classify -> draft response -> apply policy -> escalate if risk trigger. Output: reply draft + confidence + escalation flag. Escalate when: refund above threshold, legal threat, repeated billing failure.
Content Agent Template
Mission: Convert weekly research notes into publish-ready multi-channel content. Inputs: source notes, offer angle, audience profile, banned claims list. Process: angle map -> draft -> channel adaptation -> QA pass. Output: one long-form draft + three short-form variants + CTA options. Escalate when: claim cannot be verified or compliance rule is unclear.
Reporting Agent Template
Mission: Produce weekly executive summary across revenue and operations. Inputs: Stripe export, traffic data, support volumes, target KPIs. Process: aggregate -> compare to prior period -> flag variance -> recommend actions. Output: one-page summary with metrics, deltas, and owner-assigned next steps. Escalate when: data mismatch exceeds tolerance or KPI drops below floor.
4. Weekly Review Cadence Template
Run this every week in a single 60-minute operator review.
| Minutes | Segment | Owner Prompt | Output |
|---|---|---|---|
| 0-10 | KPI Snapshot | "What moved most this week and why?" | Top 3 signal changes |
| 10-25 | Failure and Drift Review | "Where did agents fail policy, tone, or timing?" | Issue log with severity |
| 25-40 | Prompt and Workflow Updates | "Which one SOP upgrade gives highest return next week?" | One approved change ticket |
| 40-55 | Execution Commitments | "What 3 actions must ship before next review?" | Owner + due date assignment |
| 55-60 | Risk Check | "Any unresolved risk that can block revenue or reputation?" | Escalation decision |
5. QA Audit Checklist
Prompt Integrity
- Every production prompt has complete RICE fields.
- Hard constraints are explicit for legal, compliance, and pricing claims.
- Output schema is stable for downstream automations.
Workflow Reliability
- Retries and fallback paths exist for API or webhook failures.
- Critical events are logged with timestamp, input hash, and owner.
- Escalation notifications route to a human within SLA.
Business Safety
- No autonomous action can issue refunds or discounts above policy cap without approval.
- Outbound messaging respects channel rules and suppression lists.
- Weekly review includes one random transcript audit per core agent.
Launch Day Go/No-Go
- Pass rate >= 95% on staging test set.
- At least one dry run completed end-to-end with production-like data.
- Rollback plan documented and tested.