# Weekly board pack for a live AI agent

Companion to **How to Show the Board What an AI Agent Did**.

Fill **your** counts. The dollar line below is a published platform silhouette, not your bill and not money saved. We have no published agent case study that claims a containment rate.

## One row per week

| Line | What to paste | Fail the pack if |
| ---- | ------------- | ---------------- |
| Agent and owner | Family (support, sales, ops, inventory, knowledge) and the named human | "The AI team" with no name |
| Sessions | Turns the host actually ran | A chatbot vendor's "conversations" with no tool trace |
| Handoffs | Turns that stopped and opened the human queue, with session id | A containment % with no handoff count |
| Writes denied | Gateway or Policy DENY on refund, cancel, PO, quote issue, price, credit release | Zero denies because the write tool was never attached — say "no write tool", do not print 0 as safety |
| One wrong answer | The worst turn: what the tool returned, what the model said | A green dashboard and no example |
| Platform cost | Your AgentCore plus model invoice line | Pasting **~$791/mo** as savings |

## The only published cost figure to compare against

Support-style AgentCore at **50,000 sessions/month ~$791/month** platform plus model. Source: the [AgentCore vs Amazon Q decision guide](https://www.factualminds.com/blog/aws-bedrock-agentcore-vs-amazon-q-enterprise-decision-guide-2026/). Model your mix on the [AgentCore pricing calculator](https://www.factualminds.com/tools/amazon-bedrock-agentcore-pricing-calculator/). If your volume is not 50,000 sessions, do not quote $791 as your cost.

## Do not paste these as your result

| Published figure | What it is | What it is not |
| ---------------- | ---------- | -------------- |
| Gorgias via Redo, WISMO **~18%** of tickets | A public mix for where-is-my-order | Your containment rate |
| McKinsey State of AI 2025: **62%** experimenting, **23%** scaling | Industry survey | Proof your agent is in the 23% |
| Gateway median tool RTT **~180 ms → ~95 ms** | A B2B CRM canary, 12 tools, ~8k turns/day | Your OMS latency |
| Browser left on: about **3×** platform spend | A published support-shaped counter-case | A reason to hide the invoice |

## Related

- Ship gate: [`monday-checklist.md`](monday-checklist.md)
- Pre-build scoring (not this pack): [`ai-automation-opportunity-score.md`](ai-automation-opportunity-score.md)