General questions
What is UserVoice's agent?
The agent lets you ask questions about your feedback data in plain English — ideas, accounts, users — and answers using your actual UserVoice data in real time.
What can the UserVoice agent do that I can't already do in UserVoice?
The agent lets you query your feedback corpus conversationally — cross-cutting questions like "what are churned accounts that wanted Salesforce integration asking for?" would require multiple manual filter steps today. It also reads the content of comments and internal notes to answer "why" questions, not just "how many."
Does the UserVoice agent use my real data?
Yes. The agent queries your live UserVoice instance in real time. It doesn't use anonymized or sample data — every answer is grounded in your actual ideas, accounts, and users.
Which AI model does the UserVoice agent use?
The UserVoice agent is powered by OpenAI models. We update the model as better ones become available.
Is the UserVoice agent available to all UserVoice admins?
If your plan includes it, an account owner turns it on in Settings → Integrations → AI Assistants, and then the admins on your account can use it. If AI Assistants is greyed out, your plan doesn't include it. Use the link there to schedule a call about upgrading. See Turn the UserVoice agent and MCP on or off.
Can I use the UserVoice agent's tools from Claude or ChatGPT?
Yes, through UserVoice MCP. An account owner first turns on MCP access in Settings → Integrations → AI Assistants. Your AI tool can then search your ideas, accounts, and users and, if you have write access, make the same changes the agent can. See Connect UserVoice to Claude, ChatGPT, or Codex.
What the UserVoice agent can and can't do
Can the UserVoice agent make changes in my UserVoice instance?
Yes, with your confirmation. The agent can create ideas, record a customer's feedback or votes on an idea, flag ideas as important or remove flags, add ideas to a list or remove them, save ideas as a new list, update labels, set public or internal status, assign owners, edit one idea's title, description, or category, link an idea to an existing Jira issue or Azure DevOps work item, merge duplicate ideas into one, delete ideas, and mark ideas as spam. Each change shows a confirmation card first, and nothing happens until you click. You need feedback write access. The agent can't create or edit Jira or Azure DevOps issues, and it doesn't send messages beyond a status message or the notifications a change sends.
Does the UserVoice agent email anyone when it makes changes?
Changes send the same notifications as doing them by hand. New ideas in public forums go live right away and notify admins who get new-idea emails. Flagging an idea notifies admins who follow flags, and removing a flag emails the person who created it. Adding feedback or voters doesn't email the customer. When the agent merges ideas, supporters are emailed only if you ask it to notify them, and the confirmation card shows how many people will get an email. A public status change posts its message on your portal and emails the idea's contributors; supporters are emailed only if you ask. An internal status change emails your team only if you ask. Status changes can't be undone.
Can the UserVoice agent search across all my feedback, including closed ideas?
By default, the agent searches open ideas (published, approved, and under-review). If you want to include closed or merged ideas, ask explicitly — e.g., "Show me closed ideas about search from the last year."
Can the UserVoice agent read internal admin notes on an idea?
Yes. When you ask about a specific idea in depth, the agent can read its body, public comments, and internal admin notes. As an admin, you have full access to all of this data within the agent.
Can the UserVoice agent access my Salesforce or HubSpot data?
The agent can surface CRM-linked fields that are synced into your UserVoice accounts — fields like plan, market segment, industry, and revenue that were pulled in via your CRM integration, plus any custom account fields you sync. You can filter on custom fields in your questions, including relative dates like "renewing in the next 90 days." It cannot query Salesforce or HubSpot directly.
Can the UserVoice agent access Gainsight health scores?
If your UserVoice instance has a Gainsight integration configured, the agent can surface health scores, stage, and renewal dates for accounts. If the integration isn't set up, those fields won't be available.
Does the UserVoice agent see revenue data?
Revenue and LTV fields, and sorting ideas by revenue, need the revenue permission on your admin account. If you're not seeing revenue in the agent's answers, your login may not have this permission. The agent shows revenue exactly as your CRM stores it. It doesn't know whether that's a monthly or annual figure, so it never labels it MRR or ARR and never converts it.
Can I ask the UserVoice agent to write a PRD or opportunity doc?
Not yet. The agent is currently scoped to answering questions about your UserVoice data. If you'd like the agent to help you draft documents, submit feedback in the Beta Forum.
Can the UserVoice agent generate charts or graphs?
Not yet. The agent is currently scoped to answering questions about your UserVoice data. If you'd like the agent to visualize your data, submit feedback in the Beta Forum.
Can the UserVoice agent count or break down ideas?
Yes. Ask "how many ideas did we get by month?" or "break down open ideas by category" and the agent returns a breakdown table. To count ideas that mention a term ("how many ideas mention SSO?"), it counts only ideas that contain that term.
Asking questions
How specific do I need to be?
You don't need to be precise — the agent will ask a clarifying question if it needs more context. That said, more specific questions return better answers. "Top 10 ideas by revenue in the Integrations category" is more useful than "what should I work on?" Not sure where to start? Start a new chat and pick a starter prompt, then fill in the blanks.
Can I use my own terminology — category names, forum names, label names?
Yes. The agent knows the names of all your categories, forums, and labels. You can use them directly in your questions without knowing their internal IDs.
What if I use a term that could mean multiple things?
The agent will ask you one short clarifying question and stop. For example, if "Integration" matches both a category and a label in your instance, it will ask which one you mean. Answer in your next message and it proceeds.
How many ideas or comments can the agent look at?
For "show me all…" requests, the agent pages through results and lists up to 200 ideas per answer. When you ask it to read everything on an idea, it reads up to 100 comments, notes, or feedback items in full and tells you how many it read.
Idea tables and lists
Why is the data in an idea table changed?
Idea tables always show live data. The agent saves which ideas were in the table and re-pulls their current details (votes, status, owner, etc.) each time you reopen a conversation. The written analysis, by contrast, is saved exactly as the agent wrote it. So the table reflects your data as it is now, while the commentary reflects it as it was then — treat the table as the source of truth for current numbers.
Can I export what the agent finds?
Yes. Every table the agent shows has Download CSV and Copy table buttons, and idea tables have Save as List. Exports include every row, not only the page you're viewing.
Conversations and history
Are my conversations saved?
Yes. The UserVoice agent saves your conversations automatically. They appear in the sidebar and can be resumed, renamed, or deleted at any time.
Can I resume a conversation from a previous session?
Yes. Past conversations are available in the left sidebar. The UserVoice agent will retain the context from that thread when you resume.
Is there a limit to how many conversations are stored?
There is no conversation limit yet, but during the beta we may bulk remove conversations at certain points as we make larger changes.
Can other admins in my organization see my conversations?
There is no conversation sharing available.
Themes and labels
How are you detecting themes?
When you ask an analytical question, the UserVoice agent reads the actual ideas in scope — their titles, bodies, and customer comments — and analyzes them on the spot to find recurring pain points. It's not pulling from a pre-built report; the analysis is generated fresh each time you ask. You'll see live progress while it runs.
The work is split in two:
- The agent clusters and names. It groups related ideas into themes, writes a specific pain statement for each, judges its intensity, and pulls a real, verbatim quote from a customer comment as evidence.
- The agent does the math. Every number you see — supporters, accounts, revenue, dates, admin activity — is calculated directly from your database, not estimated by the AI.
Two things worth knowing: a theme only forms when at least two genuinely related ideas support it, and ideas that don't fit a clear pattern are intentionally left out as "uncategorized" rather than forced into a theme. Because the analysis is generated live, results can vary slightly between runs.
Why aren't these themes the same as my Themes & Trends report?
The UserVoice agent themes are generated live, scoped to the specific question you just asked, and shaped to drive a decision — so instead of a category label you get a pointed pain statement, enriched with impact (supporters, accounts, revenue, admin activity) and a representative quote.
So differences are expected: the UserVoice agent analyzes the slice of ideas relevant to your question at the moment you ask, using its own method and a more specific naming style, rather than reading back your saved report. The agent's themes are also ephemeral — they aren't saved and won't appear in Themes & Trends, and your T&T reports won't change because of them. Your Themes & Trends data stays exactly as it is.
Beta access and data questions
Is this a beta release? What does that mean?
Yes, the UserVoice agent is in beta. That means that the core functionality is stable and running on live data, but it's actively in development — we're shipping updates regularly and collecting feedback to shape what comes next.
How can I be sure that other accounts can't see my data?
The UserVoice agent is fully tenant-isolated — it can only query data from your UserVoice instance. It cannot access other customers' data.
Feedback
What's the best way to provide feedback?
We wouldn't be UserVoice if we didn't ask you to provide feedback!
- To provide ideas for new functionality, use the invite-only forum specific to the UserVoice agent.
- Within the UserVoice agent interface, you can give agent responses a thumbs up or thumbs down with a comment. Sending us feedback through this method tells us if the agent responded in a helpful or incorrect way tied to the specific response.
For more detail on what the UserVoice agent can do, see the UserVoice agent Knowledge Base article.