AI insights and analysis
How Polzy reads every answer, notices what changed, and turns it into a finding you can act on.
On this page
Polzy reads every answer
Switch on AI analysis in a survey's Scoring tab (Solo and up). From then on, Polzy reads each response as it arrives.
- A summary and a sentiment. What the person said, and whether it reads as positive, neutral, negative or mixed.
- The themes it touches. Each theme the answer talks about gets its own score. The keywords you give a theme help Polzy recognise it.
- What matters most. Low satisfaction, negative sentiment and detractors are flagged, so a shift stands out.
Your scores never depend on AI. Ratings and NPS are scored either way. If the month's AI allowance runs out, answers are still collected and scored; only the reading waits.
Insights: what changed
Open Insights (Solo and up, for admins and managers). Polzy compares the period with the one before and looks for real change, not noise.
- A group moving against the rest. One team's score falls while the workspace holds steady.
- A new topic. Something people suddenly mention, across more than one survey.
- Negatives in one place. One group holds a large share of the negative answers.
- A group gone quiet. People who answered last period and not this one.
- A topic that moves the score. Answers that mention it score well above or below the rest.
A finding is only made from groups of at least five people, and never from a group under your minimum group size.
Read a finding
Each finding says what kind it is — Risk, Needs attention, Win or Trend — and shows the numbers behind it. Most also carry one anonymous comment that illustrates it, and how confident Polzy is, from the number of people involved.
Polzy writes a headline and a Suggested next step for each. At the top, This period in one paragraph sums up the period. Switch between Everything, Risks, Wins and Trends.
- Create action. Creates an action from the finding at once, assigned to you. Open it in Actions to give it an owner and a date.
- Ask Polzy why. Opens Ask Polzy with the finding, to dig into it.
- Not useful. Hides it for you, and Polzy shows fewer like it.
Summaries and reports
Polzy also writes for you in two places. On Comments, the Polzy summary says what people are writing about (Solo and up).
On Business and up, Reports go out on a schedule — weekly, monthly or quarterly — or when a campaign closes. Each has a summary written by Polzy, the headline figures, the period's insights and the theme scores. Send it in the app and by email with Send now, or Export PDF.
Groups below your minimum group size are left out of every report before anyone reads it.
Ask Polzy about your results
Ask Polzy is on every plan and reads the same figures you do: scores, trends, eNPS, themes, comments and how groups compare. It can draw a chart in the conversation.
It only sees what your role can see, and never reports on a group below the minimum group size — it says so rather than guess. On Insights, Questions worth asking suggests where to start.
More on the assistant. Get the most from Ask Polzy covers its modes and what it can change.
Tokens, your own model and thinking
AI work uses tokens from your plan's monthly allowance. Organization settings → Usage shows how many each feature used this month.
On Enterprise, use your own Anthropic, OpenAI or Azure OpenAI key under Settings → General → Ask Polzy and AI — usage is then billed by them. Extended thinking gives Polzy more time to reason before it answers.
To turn an AI feature off for a workspace, switch it off under Settings → General → What this workspace does.
Signed-in members can ask Polzy or message the team from Support. How fast we reply depends on your plan.