Interview Agents

What is an Interview Agent

An AI moderator that talks to your users when specific conditions are met. Seena has two types — the general feedback agent and custom interview agents.

An interview agent is an AI moderator that runs a short conversation with one of your users when specific conditions are met. Seena uses the context it knows about your site, you as the admin of the system and your research needs, and its understanding about the current user based on their current journey with the application. Based on the context, the triggering conditions, and research goals, the AI interviewer asks a question, listens (voice or text), asks a thoughtful follow-up, and wraps up — usually in 30 to 90 seconds, sometimes more. The conversation is transcribed and analyzed automatically; you read the result in the dashboard.

Seena has two kinds of agents, and they're meaningfully different.

The general feedback agent

You can turn on the persistent feedback agent. It's the pill that sits at the edge (you can choose its placement) of your page showing a label like "Feedback" — your visitors can click it anytime to open a short conversation.

  • Always on. You enable or disable it with a single toggle in the dashboard. There's no trigger logic to configure.
  • Visitor-initiated. The general feedback agent doesn't pop up unprompted. Visitors open it when they have something to say.
  • Open-ended. It asks a broad question ("What's on your mind?") and based on what the user says, the AI is trained to probe like a UX researcher and probes the user further to uncover your users needs. No script.
  • Appearance-customizable. You can change the pill's position (bottom, left edge, right edge), its label, the theme, and the logo color.

This is the lowest-friction way to start getting qualitative signal: install Seena, leave the general feedback agent on, and you have a working "tell us what you think" channel on every page.

Custom interview agents

A custom interview agent is a conversation you design for a specific research need. You create as many as you need (based on your plan).

  • Conditional. Each custom agent has triggers — time on page, scroll depth, URL pattern, visited-pages sequence, exit intent, or (on Pro-coming soon) agentic triggers that fire when Seena detects frustration, engagement, confusion, or intent to leave.
  • Scripted or goal-based. You either write the questions explicitly (specific mode), or you write research goals and let Seena generate questions to match them at conversation time (goals mode — Pro-coming soon).
  • Prioritized. If multiple agents match the same visitor at the same time, Seena fires the one with the lowest priority number. First match wins.
  • Targeted. Each custom agent targets global URLs (everywhere), or specific URL patterns, or specific sequences of pages.

Examples of things you'd build a custom agent for:

  • "Why are people abandoning checkout after step 2?" — triggers on exit intent from /checkout/step-2.
  • "How is our new onboarding landing?" — triggers after 60 seconds on the onboarding route.
  • "What's confusing about the dashboard?" — agentic trigger on detected confusion, anywhere in /dashboard/*.

How the two types coexist

The general feedback agent and custom agents share the same visual pill in the edge of your page, but they behave differently in how they engage.

When no custom agent is actively inviting a visitor, the pill shows the general feedback agent's label (e.g. "Feedback") and is clickable. When a custom agent's triggers fire, the pill morphs to show that agent's invitation — subtitle, call to action — and opens on click. When the custom interview closes, the pill animates back to the general feedback state.

Both kinds of agents respect the same governance rules: per-session cap, cooldown between prompts, dismissal escalation, and visitor recurrence. See Agent governance for the exact rules.

What to read next