> ## Documentation Index > Fetch the complete documentation index at: https://docs.pollandroll.com/llms.txt > Use this file to discover all available pages before exploring further. # Querying your results with AI > Asking questions in natural language about your survey results and letting the AI assistant go and find them. The **AI assistant** lets you explore a survey's results in **natural language**: you ask a question, it goes and finds the figures in the survey and answers you, without having to build a table or a crosstab by hand. It is not limited to pulling out figures: it is an **analysis assistant**. Beyond simply retrieving a result, it can **interpret the data, put it in perspective and bring out findings**. ## Accessing the assistant The assistant has its own **Assistant** tab in the survey's navigation bar. In the sidebar, click **New conversation** to start one, or pick up an existing conversation from the list just below — grouped by period (Today, This week, Older). Enter your question in the **Enter your question…** field, then send it. Within the same conversation, the assistant **keeps the context of the previous messages**: you can chain follow-up questions ("and among men?", "compare with brand B") without rephrasing everything. Start a new conversation when you change subject, to start again from a blank context. As for the rest of the results, what the assistant can consult depends on your **access level** on the survey. ## What the assistant can do ### Fetching the data To answer, the assistant starts by retrieving the right data in the survey: | Action | What it does | | - | - | | **Survey search** | Finds, by title or by identifier, the surveys you have access to. | | **Survey outline** | Lists the questions of the questionnaire to spot the one you are interested in. | | **Question results** | Retrieves the results of a given question. | | **Crosstab** | Crosses two questions with each other. | | **Questions comparison** | Compares side by side several questions that share the same answer items (e.g. several concepts tested with the same scale, or one and the same question tracked over several waves of a tracker). | | **Importance-performance matrix** | Lists, consults or computes an importance-performance matrix to identify the levers of an overall score. | You can also ask it to **restrict the results to a subpopulation** — for example to the respondents who gave a specific answer to a question. The assistant then builds the corresponding filter, as you would by hand. It can also **create a subtotal** on a question — it then groups the items you indicate under a single label, as you would by hand. ### Querying another survey By default, the assistant answers on the **current survey**. You can also ask it to find **another survey** you have access to (by its title, its identifier or its status) and to go and fetch results there, or to **compare a question across several surveys** (for example between the waves of a tracker), without leaving the conversation. As everywhere else, it only sees the surveys your access level allows you to consult. ### Working with your PowerPoint reports The assistant can also act directly on the reports in the **Reports** tab: * **Consult**: list the existing reports of a survey, consult the composition of a report (its slides, their charts and their filters). * **Create**: create a new empty report, add an empty slide to it — choosing its layout: free-form slide, predefined layout (two charts side by side, 2x2 grid…) or divider —, then add a chart to it — Standard (single question), Crossed, questions comparison, or importance-performance matrix. In a Crossed chart, the rows can be a subtotal of a grid question (for example a "Top 2 boxes"). * **Edit**: rename a report, move a slide, change its layout or its significance test, or change how an existing chart presents its results (display type, title, order, hidden rows or columns…) without having to recreate it. * **Check**: render a slide as it appears in the report builder, to spot an unreadable chart, a truncated legend or overlapping elements after a change. * **Filter**: apply a respondent filter to a report, a slide or a chart, or remove it. * **Write**: write or rewrite the summary of a slide or of a report from the results it has just consulted. * **Delete**: delete a report, a slide or a chart — and everything it contains. For example: "Create a new report and add a slide to it", "Add a chart of the awareness question to this slide", "Switch slide 4 to two charts side by side", "Hide the 'No answer' row of the chart on slide 2", "Filter this report to women under 50", "Write the summary of slide 3", "Delete the pie chart on slide 2". Deletions are permanent: a report, slide or chart deleted by the assistant cannot be restored. ### Analyzing and interpreting Once it has the figures in hand, the assistant can **put them to work for you**. In particular, it can: * **Compare** two items, two brands or two profiles, and say which one stands out — for example comparing two brands from the same question, or checking whether a result is stronger among women than among men. * **Characterize** a subject by crossing several questions — for example building a brand's image from the survey's various indicators. * **Explain a variation** from one profile to another (age, gender, subpopulation…) and point out the notable gaps. * **Summarize**: bring out the main findings of a question or of a set of results. Ask one question at a time and name the questions or the items as they appear in the survey: the assistant finds the right data more easily. ## Example questions You can make requests for **raw figures** as well as requests for **analysis**. A few examples: To obtain a result: * "Give me the results of the awareness question." * "Which brand is mentioned most?" * "Cross consideration with age." * "Among the 18–34s, which brand is the best known?" * "Compare the results of the questions on the three concepts." * "Find my surveys on the coffee market." For an analysis or an interpretation: * "What is this brand's image?" * "Compare brand A and brand B." * "Is consideration stronger among women than among men?" * "What are the main findings of this survey?" ## Following its reasoning Under each answer, the assistant displays its **Actions**: the steps it followed and the data it consulted. Expand them to check where a figure comes from. AI can make mistakes. Remember to check important information, cross-checking it if needed against the question's table in the results. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.