Hello, I'm Higashi from the marketing team EM at JAPAN AI.
Today I want to share the story of a Slack thread summarization bot we built during a one-day AI hackathon held by the JAPAN AI business division in August.
The hackathon's goal was "developing AI tools to improve our own work efficiency." About 30 people participated—engineers, product managers, customer success team members—working individually or in teams to build AI applications. Since time was limited, I chose to work alone rather than in a team.
I implemented a bot that summarizes Slack thread content. I often get mentioned in the middle of long threads and spend too much time trying to understand what's being discussed, so I wanted to solve this problem by building something myself.
But simple AI summarization and natural language search are already available as Slack's official AI add-on (though it's a bit pricey at ¥1,200 per user per month).
So instead of just replicating that functionality, I decided to implement some of those features while adding ideas for external integrations that Slack's AI add-on can't do.
When users mention the bot, it analyzes conversations in the thread and creates concise summaries, including decisions made and action items.
Incident Analysis
Analyzes content in service incident reporting channels, searches for similar past cases and solutions. Integrates with Confluence pages to suggest next actions based on historical cases.
GitHub Integration
Summarizes specification discussions that happen in Slack threads and should be linked to PRs. Posts summaries directly as comments on relevant PRs.
Just mention the app and it summarizes for you. (In practice, this assumes conversations between multiple accounts.)
・Slack Bot: TypeScript, Bolt for JavaScript
・AI Workflow: Dify