HiveGive your first AI employee a job
Turn an outcome into a task you can review.
Start a conversation
Choose New AI employee in the sidebar. Describe the work, the information available, and what a good result looks like. Select an available harness and model before asking it to run.
Try a request like this:
Review these anonymized maintenance requests. Group the recurring problems, identify missing information, and draft a response checklist. Do not contact residents or change records. Show which requests support each finding.
This gives the AI employee an outcome, source material, boundaries, and an output you can check.
Provide enough context
Explain the customer, terminology, process, and exceptions. If there is an existing checklist or standard operating procedure, attach it. Tell the AI employee which source is authoritative when documents disagree.
Ask questions in the same conversation when you need a correction. Give concrete feedback: “A water leak must be escalated immediately” is clearer than “Be more careful.”
Review a real result
Check whether the AI employee used the right evidence, missed an exception, or made up information. Open referenced files and compare samples with the original records.
An AI employee saying it completed a task is not proof. For work in another system, verify the actual result there.
Make the next step explicit
Once a draft is reliable, decide whether the AI employee may perform the next action. Reading a mailbox, drafting a reply, and sending a reply are different responsibilities.
Reusable Skills and recurring company workflows are part of Hive's direction. Do not assume that a chat instruction alone creates an enforced permission rule or a running schedule.