Exercise: Building a Meeting Notes Analyzer
Explore how to build a hybrid multi-agent system that processes unstructured meeting notes. Learn to design an orchestration strategy that summarizes discussions, extracts action items, categorizes key decisions and risks, and compiles a polished briefing document.
It is time to apply the knowledge to a practical, hands-on challenge. In this exercise, we will build a complete, hybrid multi-agent system from the ground up. This task will require us to analyze a problem, design an appropriate orchestration strategy, and implement a full pipeline that combines both parallel and sequential execution to produce a useful result.
The scenario
A common challenge in any professional environment is managing the output of meetings. Notes are often taken in a hurry, resulting in unstructured text that mixes important decisions, action items, and general discussion points. Manually analyzing these notes to create a clean, shareable summary is a time-consuming and error-prone task.
Here is an example of the kind of messy input:
Product Sync – Mobile CheckoutDate: Oct 3Attendees: Sara (PM), Leo (Eng), Priya (Design), Omar (QA)Notes:- Sara: We’re seeing a 12% drop-off at the shipping step on mobile.- Priya: Proposed change: collapse address fields by default.- Leo: We can also auto-detect country and adjust keyboard types.- Omar: We need regression tests for iOS Safari.- Decision: Ship collapsed address fields for mobile only.- Action: Priya to update designs by Oct 7.- Action: Leo to implement changes by Oct 14.- Action: Omar to run smoke tests by Oct 16.- Open question: Should returning users see collapsed fields?- Risk: Keyboard behavior may vary across browsers.
Our mission is to ...