Bringing AI to Community Management
Empowering communities to become more efficient!! We are building — HiveMind — an LLM-powered chatbot, that can answer complex questions, while harnessing the information spread across a community’s different platforms.
Get exclusive content and product news through our newsletter.
In the past few months, our team has been conducting user research seeking to understand the top pain points of community managers, and found that being overwhelmed and lacking time are key challenge they face every day.
The research also highlighted a second significant problem for communities per se — knowledge is often fragmented across several platforms and often poorly documented. This leads to a significant time loss when trying to recall informal conversations or locate the platforms where they occurred. For example, to find out “What did the marketing team do last week?” the only way is to ask them, taking their focus away from their work.
Our team got down to brainstorming on the best way to provide a solution to these problems, and zeroed down on building an AI powered bot, which we’ve named HiveMind. The tool is designed to ingest information from across the multiple platforms where a community’s information is spread, such as Discord, Discourse, GitHub, Google Drive, and Notion. And further, by operating as a Q&A chatbot, the tool provides an intuitive and easy to use interface to its users. With HiveMind, users can quickly get answers to complex questions, thereby fostering efficient collaboration among community members and facilitating decision-making, coordination, and inclusivity.
HiveMind functions as a multi-agent system, incorporating multiple LLM-powered agents that collaboratively address a broad range of questions. This architecture equips HiveMind with the ability to tackle complex queries effectively. For example, consider a question like: “What were the suggested solutions for [PROBLEM], and which of them yielded positive results?” Without HiveMind, users would have to navigate through multiple inquiries:
Query 1: find solutions
Query 2: what were results of solution 1
Query 3: what were results of solution 2
HiveMind on the other hand, can split the objective into separate queries, answer them individually and merge the results into a single final answer, providing speed and a superior user experience.
The architecture of Hivemind draws inspiration from the human brain, with multiple forms of memory, including episodic, semantic, and procedural. HiveMind constructs its episodic and semantic memory by exploring the information environment, utilizing vector stores containing data from different sources (Discord, discourse, etc) and levels of abstraction. This enables the system to effectively answer loosely structured questions like, “The other day, when we were talking about something related to the new partnership, what was the decision?”, drawing from content shared across different platforms of the community.
To maximize flexibility for user communities, HiveMind operates as both an API and a Discord chatbot. This will enable users to have their questions answered in the same place they’d normally ask them, thus reducing friction and promoting seamless user adoption.
Join our Community
- Talk with other community builders: TogetherCrew community
- Receive tips and follow along: TogetherCrew on Twitter
- Get exclusive content and product news through our newsletter
