What are AI agents in Userorbit?
AI agents in Userorbit are intelligent assistants that help your team work faster and more consistently across every customer-facing workflow. They operate inside your existing Userorbit workspace, drawing on your product context, customer data, and team preferences to generate content, automate repetitive tasks, and surface insights you might otherwise miss.
Unlike generic AI chatbots, Userorbit agents are deeply integrated with your help center, feedback pipeline, announcements, and changelog. They understand your product vocabulary, your audience, and the way your team communicates.
What can agents do?
Agents cover three broad areas of capability:
- Content generation — Draft help center articles, changelog entries, announcement posts, and feedback responses. Agents use your existing content as a style reference so new drafts match your voice.
- Task automation — Handle routine work like categorizing incoming feedback, suggesting responses to common questions, or flagging duplicate feature requests. Agents propose actions and, depending on your confidence settings, either queue them for review or execute them automatically.
- Intelligent workflows — Chain multiple steps together. For example, an agent can watch for new feedback, summarize it, attach it to the relevant feature request, and draft a response — all without manual intervention.
Agent modes
Every agent operates in one of several modes that control how much autonomy it has:
- Assistant mode — The agent responds only when you ask. You interact through the chat interface, request drafts, ask questions, or run quick actions. Nothing happens without your explicit prompt.
- Supervised mode — The agent watches for triggers (new feedback, scheduled times, events) and proposes actions. Each proposal lands in the agent inbox where a team member reviews and approves or rejects it before anything is published or sent.
- Autonomous mode — The agent acts on its own when its confidence score exceeds the threshold you set. High-confidence actions execute immediately; lower-confidence actions still route to the inbox for human review.
How agents stay accurate
Agents rely on several safeguards to maintain quality:
- Confidence scoring — Every proposed action carries a confidence score. You decide the threshold at which actions need human approval.
- Guidance instructions — You provide agents with custom instructions about your product, tone, and policies so their output aligns with your standards.
- Quiet hours — You control when agents are active so they never send a notification or publish content outside approved windows.
Together, these controls let you start conservatively — reviewing every proposal — and gradually grant more autonomy as you build trust in the agent's output.
Where to go next
If you are new to agents, start with Getting started with AI agents for a hands-on walkthrough. To understand modes in detail, read Understanding agent modes and capabilities.