Recording in, structured minutes out: decisions, owners, deadlines — filed to your document space and turned into tracked tasks. Personal remarks are not mined; conduct rules bar gossip relay.
Hedy turns a meeting into governed output, not a transcript. It extracts decisions, owners, and deadlines, archives them into your own cloud docs, and pushes each action item into a trackable task. Because Hedy is a private AI employee self-hosted on your infrastructure, minutes never leave your network and no sub-processors touch them. It stays factual: no subjective judgments about people, and every claim is cited to its source under an evidence gate.
Last updated: 20 July 2026
Drop the transcript or notes in chat; Hedy extracts what was decided, by whom, due when.
Decisions, actions and open questions, in your template, filed to Feishu or Lark docs with a link back.
Each action item is tracked with owner and due date; Hedy follows up before it slips.
Decisions land in company memory — next week’s "what did we decide?" has a cited answer.
A note-taking assistant stops at a transcript. Hedy does the job that comes after the meeting. It reads the discussion and produces a structured record: what was decided, who owns each item, and when it is due. That record is written into your cloud docs, Feishu/Lark or Slack-linked storage you already run, formatted so a human can act on it without re-reading an hour of talk. One Hedy replaces the manual work a team used to split between a scribe, a project coordinator, and a follow-up chaser.
Action items do not die in a document. Hedy converts each one into a trackable task with an owner and a deadline, so the meeting output flows straight into the systems where work actually gets tracked. Nothing is invented: Hedy only records what was said. If a decision, owner, or date was not stated, it is left open rather than guessed, and the answer names exactly where each item came from.
Governance is the product, not a feature bolted on. Minutes and task creation run under the same controls as everything else: an append-only audit log, tiered approvals (L0/L1/L2), and a deny-by-default whitelist for write operations, so pushing tasks into a downstream tool is an authorized action, not an open door. Retrieval respects the asker's ACL, so a summary only surfaces what that person is cleared to see.
Two lines Hedy will not cross. It does not broadcast subjective evaluations of participants, no tone-policing, no unrequested judgments about who performed well. And it does not leave your walls: Hedy is self-hosted via ~30-minute Docker/Compose deployment, brings your own model key including local vLLM or Ollama, runs on an offline Ed25519 license, and uses no sub-processors. The meeting record is generated inside your infrastructure and stays there.