How AI is used in AnalyZ Solutions, and what it never sees

01Three questions to ask of any tool that uses AI

Analysis software now often includes an AI feature, and the label alone tells you very little. Whether the feature is safe and useful for your work depends on three things, and you can ask them of any tool.

• What does the AI receive? Your raw records, a summary of your results, or only a few words you typed?
• What does it actually do? Does it calculate, decide, or only describe?
• Who checks the output? Is it shown as a draft you review, or presented as a finished answer?

The rest of this guide answers those three questions for AnalyZ Solutions. For a broader view of why these questions matter, see why you should not rely on AI to analyze your data.

02Where AI appears

AI is used in three places, and in each it works on the results you have already produced.

Chart interpretation is described in how to build charts and share a dashboard, Sense Making in how to interpret your analysis with Sense Making, and Data Canvas in how to tell a story using your data.

03What is sent, and what is not

The information sent to the AI model is the same information you can already see on your screen: numbers in a table, the shape of a chart, a theme name. It is sent only to produce the interpretation text.

The following never leave your browser for AI:

• Your dataset, including every row and every respondent.
• The files you uploaded.
• Interview transcripts and other qualitative source text.

This follows from how AnalyZ Solutions is built. Your data is held and processed in your own browser, as explained in the privacy model. Only a few features make a network request at all, and the AI ones send aggregated outputs rather than records.

04What AI does not do

Equally important is what the AI is not asked to do.

• It does not run your statistics. Frequencies, tests, regressions and charts are calculated in your browser using documented methods. The AI only describes the output after it exists.
• It does not choose your test. You pick the method that suits your question and data. The guide for choosing the right analysis is supported by the built-in guided analysis module.
• It does not code your qualitative data. You build and apply the codes. AI is used only after the analysis, to help summarize it.
• It does not build your personas. Persona profiles come from the patterns in your coded data, not from a language model.

The order is the point. The numbers and themes are produced first by methods you can inspect. The AI comes afterwards and turns finished results into readable text.

05How you stay in control

Every AI output in AnalyZ Solutions is a draft for you to review. You decide what goes into it and what comes out.

• You choose the inputs. Only the results you add to a project or a canvas are used. Anything you leave out is not read.
• You set the frame. In Sense Making you choose one of four writing styles and a length, and you write the context. In Data Canvas you answer questions about the audience and write the headline of each finding. The writing styles guide explains the options.
• You can refine. A Refine panel accepts a plain instruction such as "make the conclusion more direct", and the text is rewritten. In Data Canvas you can refine a single slide or regenerate the whole story.
• You can go back. Sense Making keeps up to three earlier versions, and you can restore any of them.
• You can ask questions. After a synthesis, a chat panel lets you ask follow-up questions about the findings.
• You can edit directly. You can change any word yourself, which is better than a refine instruction when only a small fix is needed.

Interpretations do not update themselves. If you add or remove a result after generating, run the synthesis again so the text matches your current results.

06Monthly limits

AI features are limited per month. On the free plan you can generate 1 Sense Making interpretation, 1 dashboard interpretation and 2 Data Canvas stories, and use 5 Data Canvas slide refinements. The Starter and Pro plans have higher limits and also let you publish dashboards and stories as public links. Everything else in AnalyZ Solutions, including all analysis, is free and not counted against these limits.

Because the limits are small on the free plan, it helps to prepare before you generate: check your numbers, write a clear context, and write headlines that state what each finding means.

07Using AI-written text responsibly

• Check every number against the output it came from. The text should agree with your tables. If it does not, the table is right.
• Read for claims the data cannot support. A draft can sound more certain or more causal than your design allows. Edit those sentences.
• Keep the interpretation yours. Treat the draft as a starting point that saves time on wording, not as a conclusion.
• Follow your reporting rules. Many journals, funders and institutions ask you to disclose AI assistance. Check the rules that apply to you and describe how the text was produced.