Notion-Native Account Plan Agent, "Yes Chef" (L3)
5 chained prompts in Notion AI generate a full account plan, stakeholder map, tiered contacts, history, recommended actions, in 90 seconds. Killed the 6-week change-management drag of rolling out account planning. By Justin Dries, Director of Sales Enablement at Legora, on the GTM AI Podcast.
The steps
- 01
Use Notion as the RAG, not a separate vector DB
Tool: Notion
If your team already lives in Notion, your RAG already exists. All product marketing, use cases, ICP docs, internal playbooks → the agent reads them natively. No separate retrieval system to build.
- 02
Use ONE agent with named instructions, not many
Tool: Notion AI
Notion's unlock: one agent + multiple instruction pages. Justin runs "Account Planner," "Pat GPT" (how the CRO would read your plan), "Max GPT" (how the CEO would read it). Each instruction is just a Notion page, humans can edit it.
- 03
Build prompts by talking to Notion AI itself
Tool: Notion AI
Don't prompt-engineer in a vacuum. Tell Notion AI what you want, let it draft the prompts, iterate. 5 chained prompts is what Legora landed on, stakeholder mapping → tiering → history → recommended action → consolidation.
- 04
Make the trigger phrase non-negotiable
Tool: Manual
Legora's kickoff theme was "Yes Chef." The agent only fires when the rep types "yes chef" on a new account plan template with the account name in the title. Forces rep adoption to be one keystroke, not a checklist.
- 05
Be twice as clear about what you DON'T want
Tool: Prompt design
Justin's rule: be very clear about what you want, twice as clear about what you don't want. Lists stale Slack channels to exclude, deprecated product docs to ignore, etc. Raises the floor of output dramatically.
Tools in this playbook
- Notion
- Notion AI
- Manual
- Prompt design
Next playbooks
Unfamiliar terms are defined in the AI and Revenue Dictionary. Related frameworks live in the framework library.
