# eCommerce AI Agent Readiness Checklist

Companion to **The eCommerce AI Agent Readiness Assessment**.

This is an **org-wide** score, not the shopping-agent PDP/API score in [`shopping-agent-readiness-checklist.md`](shopping-agent-readiness-checklist.md) and not the external-agent channel score in [`agentic-commerce-merchant-readiness.md`](agentic-commerce-merchant-readiness.md). Score each row **0 / 1 / 2**. Total **30**. Below **16**: do not staff a write-capable agent this quarter.

Platform cost to model against (not savings): support-style AgentCore **~$791/mo at 50K sessions**.

```text
AI Agent Readiness =
  Data Readiness
  + Integration Readiness
  + Process Readiness
  + Governance Readiness
  + Business Priority
```

## Score card (fill yours)

| Dimension | 0 | 1 | 2 | Your score |
| --------- | - | - | - | ---------- |
| **Data — identity** | Three IDs for one customer / SKU / order | Mapping table exists; not enforced | Canonical IDs in the systems the agent will read | |
| **Data — completeness** | Required fields missing on >20% of live SKUs or open orders | Known gaps; enrichment backlog | Required fields present for the first workflow | |
| **Data — freshness** | Nightly dump; nobody can name `asOf` | SLA named; breached weekly | Named freshness SLA that ops trusts | |
| **Integration — APIs** | HTML / CSV / email only | Read APIs for one domain | Named OpenAPI/MCP tools for the first workflow | |
| **Integration — writes** | Admin keys in a chatbot | Writes exist; no Policy | Cedar path sketched (`LOG_ONLY` → `ENFORCE`) | |
| **Process — owner** | “The AI team” | Named sponsor, no RACI | Named business owner + engineering owner | |
| **Process — exception path** | Chatbot guesses | Escalate to a shared inbox | HITL queue with session id + tool trace | |
| **Governance — evals** | Demo prompts | A few goldens | Golden + must-not-write cases; owner | |
| **Governance — approval** | Prompt says “be careful” | Cap in the prompt | Cap in Policy; over-cap is a ticket | |
| **Governance — security** | Storefront Admin token | Scoped token; Browser on | Identity JWT, Browser off, no unrestricted SQL | |
| **Priority — problem** | “We need AI” | A theme (support, ops) | One sentence problem + first workflow | |
| **Priority — blast radius** | Writes on money/ATP/price week one | Reads first, writes later | Veto list signed (payment, ATP, live price, account) | |
| **Priority — sequence** | Fifteen agents | Three in parallel | One read-heavy workflow this month | |
| **Human approval** | None | Informal Slack | Queue; timeout fails closed | |
| **Business data layer** | Dashboards only | Replica / warehouse copilot | Named tools over orders, customers, products, inventory | |

**Total / 30:** ___

| Band | Meaning | This quarter |
| ---- | ------- | ------------ |
| **0–15** | Not ready | Fix IDs, APIs, owner. Chat widget is a liability. |
| **16–22** | Ready for a **read** agent | WISMO, policy lookup, daily brief. No refund/ATP/price writes. |
| **23–30** | Ready to add **governed writes** | One write behind Cedar + HITL. Not a multi-agent program. |

## Not this checklist

- Shopping-agent catalog/inventory/price score → [`shopping-agent-readiness-checklist.md`](shopping-agent-readiness-checklist.md)
- Which workflow to staff first → [`ai-automation-opportunity-score.md`](ai-automation-opportunity-score.md)
- Per-action Execute vs HITL → [`autonomy-spectrum-by-action.md`](autonomy-spectrum-by-action.md)

## Related

- Monday gate: [`monday-checklist.md`](monday-checklist.md)
- Security: [`ecommerce-agent-security-checklist.md`](ecommerce-agent-security-checklist.md)