How do I use a Panel to run discussions and interviews?

The short answer: a Panel is a saved group of AI personas that you build once and use repeatedly. Create the Panel, then start research projects inside it whenever you have a question for that audience. Three steps from setup to results.

---

When to use a Panel instead of starting fresh

Use a Panel when:

• You want to study the same audience over time ("our core users")
• You'll ask this group multiple questions across months
• You need consistent feedback across studies (same personas, different questions)

Don't use a Panel when:

• It's a one-off question — just start a study directly, choose personas inline
• The audience for each question is different

A Panel is a saved research group. It's not a project. Projects run inside Panels.

---

Step 1 — Create the Panel

Go to the top navigation → Panel → Create New Panel.

Describe your target audience in plain language. Example:

"Budget-conscious young adults in tier 1–2 cities with strong interest in fitness, half male half female."

AI searches the persona library and shows you 10–15 candidates. Add or remove until the group feels right. Confirm. The Panel is saved.

The more specific your description, the better the match. Include age, city, spending habits, attitudes, life stage. "Young adults in China" gives worse results than "25–32 year-olds in tier 1 cities, gym members, spend ¥200–500/month on themselves."

---

Step 2 — Start a research project inside the Panel

Open the Panel → click "New Project." You choose the research type:

User Interview. One-on-one interviews with each persona in the Panel. Best for understanding individual motivations — why they do what they do, what they'd pay for, what would change their mind.

Expert Interview. Consultative format with expert personas. Best for industry insights, technical validation, professional decision-making.

Focus Group. All personas discuss together, AI moderator guides the conversation. Best for seeing where people agree and disagree quickly, comparing reactions to multiple options.

Then describe your research question:

"We're considering a subscription model. Test user reactions to monthly vs annual billing."

"Package design direction A vs B — which do users prefer, and why?"

"In what situations would users choose us over competitor X?"

---

Step 3 — View results

The project runs in the background. You can close the page; the research keeps going.

For Focus Group projects you get:

• Full conversation transcript
• AI-generated discussion summary
• Structured minutes (consensus, disagreements, key contention points)

For Interview projects you get:

• Individual transcript for each persona
• Conclusion summary per persona
• Overall interview summary

While running, the header shows "Agent Running." Completion notification arrives by email and in-product.

---

How big should a Panel be

Start with the lower end. You can always add more personas to a Panel later. Smaller, well-chosen groups usually beat large generic ones.

---

Managing a Panel over time

Add a persona. Panel detail page → Add → search library → select. Duplicates are filtered automatically.

Remove a persona. Hover the persona card → × in the top-right → confirm.

Edit the Panel description. Useful if your target audience shifts. The personas in the Panel don't change unless you manually swap them.

Permissions. Only the Panel creator can add or remove personas. Team members can launch projects using the Panel but can't change who's in it.

---

Multiple projects on one Panel

You can run as many projects as you want on a Panel. Each project is independent:

• Results don't interfere
• A new project doesn't "reset" anything from a previous one
• You can run a focus group and interviews back-to-back on the same Panel to get group-level and individual-level insights on the same audience

This is the main reason Panels exist — comparing how the same audience responds to different questions over time.

---

Mid-project adjustments

You can intervene while a project is running. Click "View Agent Chat" on the project page to open the conversation with the AI. Type adjustments like:

• "Add a question about pricing sensitivity"
• "Push back on the social persona's dismissal of premium features"
• "Skip the next two questions and move to recommendations"

The AI takes the steer mid-interview. You can also cancel and re-run if the direction is fundamentally off.

---

When a Panel stops being useful

A Panel ages out when your audience changes:

• You've expanded to a new segment (Gen Z users when your Panel was millennials)
• Your product has pivoted (B2C to B2B)
• Market has shifted significantly (pandemic-era → post-pandemic)

In these cases, build a new Panel rather than mutating the old one. Old research stays attached to the old Panel, so you don't lose anything.

---

Common patterns worth knowing

Build a Panel of your actual customers. Upload 10–20 customer interview transcripts as custom personas. Now you can test product changes against your real customer base without re-interviewing them.

Build a Panel of your competitor's users. Scout social media for users discussing your competitor, generate 5–10 personas from those discussions. Use this Panel to understand what attracts them to the alternative and what would make them switch.

Run a Panel through a quarterly research cycle. Same Panel, four questions per year. Track how the same personas' attitudes shift as you ship product changes or as the market evolves.

---

What Panels can't do (yet)

• Edit persona attributes. You can add or remove from a Panel, but you can't directly edit a persona's individual traits. To customize, build the persona from your own data first, then add it to the Panel.
• Cross-Panel comparison views. Reports from different Panels don't automatically compare. You'd need to read them side by side manually.
• Version history of a Panel. Currently, "savepoints" of Panel composition aren't tracked. If you remove a persona and add them back later, you can't restore previous versions.

---

Related: AI Persona Library, AI Research, Interview Chat, Discussion Chat