Inspiration
In many teams, meetings generate long and messy notes, and important action items are often forgotten or unclear. People spend extra time searching past emails, documents, or chat messages to understand who was responsible and what decision was taken earlier. I wanted to build a practical AI solution that automatically converts unstructured meeting notes into clear, actionable tasks with historical context. The Elasticsearch Agent Builder Hackathon was the perfect opportunity to build a real-world productivity tool instead of just another chatbot.
What it does
Smart Meeting Action Agent is an AI agent that reads meeting notes and automatically:
Extracts action items
Identifies the responsible person (owner)
Detects due dates
Searches past tasks and meetings stored in Elasticsearch
Provides structured output with context and similar past actions
Instead of manually reviewing notes, teams instantly receive a clean list of tasks with references to previous related work. This reduces follow-up time and prevents missed responsibilities.
How we built it
The project was built using Elasticsearch Serverless and Elastic Agent Builder.
Steps involved:
Created Elasticsearch indexes such as tasks, meetings, and team
Inserted sample datasets through Kibana Console
Configured a custom AI agent using Agent Builder
Attached built-in Search tools to allow the agent to retrieve historical context
Wrote structured agent instructions for task extraction and reasoning
Tested the agent using realistic meeting notes in Agent Chat
Demonstrated real-time retrieval and reasoning capabilities
The solution combines LLM reasoning + Elasticsearch search + structured indexing to deliver accurate and contextual outputs.
Challenges we ran into
Understanding the correct _bulk data insertion format in Elasticsearch
Configuring tools without overwhelming the agent
Structuring instructions so the agent extracts tasks consistently
Balancing simplicity with real-world usefulness
Making the demo clear and concise within a 3-minute limit
These challenges helped refine both the technical setup and the presentation flow.
Accomplishments that we're proud of
Built a fully functional multi-step AI agent from scratch
Successfully integrated Elasticsearch search for contextual retrieval
Demonstrated real-time reasoning with historical references
Delivered a practical business-oriented solution rather than a generic chatbot
Created a clean demo showing problem → solution → impact
What we learned
How Elasticsearch can power AI applications beyond traditional search
Importance of prompt and instruction design in agent behavior
Tool orchestration within Agent Builder
Structuring data for better AI retrieval accuracy
The value of concise and clear demo storytelling
This project deepened understanding of context-driven AI agents and real-world workflow automation.
What's next for Smart Meeting Action Agent
Automatic task creation in project management tools
Slack / Microsoft Teams integration
Deadline reminders and notifications
Multi-agent collaboration (planner + verifier agents)
Real company data integration for enterprise productivity
The long-term vision is to evolve Smart Meeting Action Agent into a full meeting productivity assistant that not only extracts tasks but also tracks progress and ensures accountability across teams.
Log in or sign up for Devpost to join the conversation.