Uploaded May 2026 | Updated September 2026, 2 weeks ago
Use Custom Agents to build a personalized meeting assistant for a weekly project sync.
It pulls from project docs, a decision log, and the previous week’s recap, bringing everything together in an agenda for review. After a meeting, it creates a recap with action items and shares it in Slack so the whole team knows what’s next. And once it’s set up, it just does this automatically, every week.
Notion’s Samuel Li walks through the full build process.
Learn more about Custom Agents in Notion’s help center: notion.com/help/category/custom-agents
And browse our Custom Agents use cases for more ways to automate your busywork: notion.com/product/ai/use-cases?page=0&feature=customAgent
Follow along with Notion:
X: https://x.com/NotionHQ
LinkedIn: linkedin.com/company/notionhq
Instagram: instagram.com/notionhq
TikTok: tiktok.com/@notionhq
Threads: threads.net/@notionhq
Chapters:
00:00 Intro (what the agent will do)
00:39 Create a new agent
00:54 Describe the meeting assistant in chat
01:35 Tools & access (what the agent needs)
02:04 Connect calendar + Slack access
02:34 Set up triggers (when the agent runs)
03:21 Review + refine instructions (add examples)
05:34 Test runs + activity log
07:28 Wrap-up / why it matters
Use Custom Agents to build a personalized meeting assistant for a weekly project sync.
It pulls from project docs, a decision log, and the previous week’s recap, bringing everything together in an agenda for review. After a meeting, it creates a recap with action items and shares it in Slack so the whole team knows what’s next. And once it’s set up, it just does this automatically, every week.
Notion’s Samuel Li walks through the full build process.
Learn more about Custom Agents in Notion’s help center: notion.com/help/category/custom-agents
And browse our Custom Agents use cases for more ways to automate your busywork: notion.com/product/ai/use-cases?page=0&feature=customAgent
Follow along with Notion:
X: https://x.com/NotionHQ
LinkedIn: linkedin.com/company/notionhq
Instagram: instagram.com/notionhq
TikTok: tiktok.com/@notionhq
Threads: threads.net/@notionhq
Chapters:
00:00 Intro (what the agent will do)
00:39 Create a new agent
00:54 Describe the meeting assistant in chat
01:35 Tools & access (what the agent needs)
02:04 Connect calendar + Slack access
02:34 Set up triggers (when the agent runs)
03:21 Review + refine instructions (add examples)
05:34 Test runs + activity log
07:28 Wrap-up / why it matters







![Inside Notions Multi-Agent System for Customer Feedback
Every month, 5 Custom Agents go through 1000s of hours of customer calls to identify themes in customer requests. The result is our GTM Top 10, a ranked report of feature requests tied to real quotes and potential business impact. In this video, Sandra breaks down the system agent by agent, from customer call to finished report. Let’s get started!
Chapters:
0:00 Inside Notions Multi-Agent System for Customer Feedback
0:27 Intake Agent
0:51 Extractor Agent
1:11 Matcher Agent
1:28 Logger Agent
1:44 Reporter Agent
Build your own multi-agent system for customer feedback:
- It takes time and experimentation to set up a system like this one successfully. The best place to start is with simple prompts to build each Custom Agent, and then to validate each step in turn. In this case, we use 5 agents with narrow scopes for this exact reason, so that we can easily debug and improve the quality of each agent. Here are the starter prompts we’ve found most helpful:
- Intake Agent → You are the Intake Agent— the first agent in the Gong FR Pipeline. Your sole job is to read a Gong call transcript and determine whether it contains an explicit, customer-voiced feature request. You do not extract quotes, search databases, or create entries. You make a binary decision: is a feature request detected or not.
- Extractor Agent → You are the Extractor Agent— the second agent in the Gong FR Pipeline. Your sole job is to locate the exact verbatim customer quote from the Gong transcript that grounds the identified feature request, and create a Snippet entry in the Snippets database. You do not decide whether the FR is valid (the Intake Agent already did that). You do not search for matches (the Matcher Agent does that next).
- Matcher Agent → You are the Matcher Agent— the third agent in the Gong FR Pipeline. Your job is to determine whether the identified feature request already exists in the Feature Requests Database, whether it has already been launched as a GA feature, or whether it is committed to an upcoming launch.
- Logger Agent → You are the Logger Agent— the fourth agent in the Gong FR Pipeline. Your sole job is to create a properly formatted entry in the Feature Request Triage Log database using all the information that has been assembled on the pipeline card by the previous agents.
- Reporter Agent → Your are the Reporter Agent— the fifth and final agent in the Gong FR Pipeline. Your sole job is to create a monthly “GTM Top 10” report using the same structure, headings, callouts, toggles, and tables as in the tagged report template [here]. Pull top feature requests from the Feature Request Triage Log [here].
- If you’d like professional support building a system like this one, contact sales to learn more → https://www.notion.com/contact-sales
Resources:
- How to build a Custom Agent → https://www.youtube.com/watch?v=ojAvnSdsc1I
- Explore Custom Agents in our “Hall of Fame” → https://notion.notion.site/Official-Getting-Started-with-Custom-Agents-312efdeead058059a812c8b9bca1d42e
- Build a Product Feedback Router → https://www.youtube.com/watch?v=w5cFKe2H3Qw
Follow Notion:
- X: https://x.com/NotionHQ
- LinkedIn: https://linkedin.com/company/notionhq/
- Instagram: https://instagram.com/notionhq/
- TikTok: https://tiktok.com/@notionhq
- Threads: https://www.threads.net/@notionhq Inside Notions Multi-Agent System for Customer Feedback](https://i.ytimg.com/vi/ndTwYKuWjlE/mqdefault.jpg)


