Data and security
Automation needs clear boundaries. Data handling, quality checks and traceability are therefore part of the product itself. Client data storage and AI processing use highly secure Microsoft Azure within the EU; deterministic calculations run in your own browser.
AI never sees all of your data
The analysis is deterministic: findings, potential and every number are produced by versioned code — without AI. AI only assists in a few steps, under strict limits.
What AI can see
- Column and sheet names from your file.
- A handful of example values, with customer and product names replaced by codes (KUND-0147, ART-02318).
- Completed findings from the deterministic engine when the presentation text is written.
- What the respondent writes in an AI interview.
What AI never sees
- The whole file or whole columns — never a complete dataset.
- Customer lists, product registers or contact details in plain text.
- Rows you have not chosen to share — no transaction rows are sent for analysis.
- Your data for model training: Azure OpenAI does not use customer data to train its models. Microsoft may retain data for abuse monitoring under its applicable terms.
You set the level
Standard (recommended)
AI receives coded example values. Best quality and least manual work — and names are still never readable.
Limited (default)
Column and sheet names only. Examples are replaced by pattern descriptions such as "number, 2 decimal places, comma".
No AI on data (more manual work for you)
AI assistance in the data steps and AI-generated presentation text are switched off. Quality checks and deterministic analysis work as usual, but more choices must be made manually. AI interviews remain available: they use participants’ own answers, not the data file, and can add qualitative context to the findings.
You can see exactly what is sent
Before every AI call, you can open the complete text, word for word. Each call is logged per analysis with the time, step, mode and whether values were included — and the log can be downloaded.
You can switch mode whenever you like — for the whole account or for a single analysis. We never change the level you have chosen afterwards.
Deterministic core
- Pricing calculations are performed by versioned code — not a language model.
- Quality controls stop analysis when serious errors are found.
- Every finding traces back to its method and evidence.
- Assumptions and limitations are shown in the deliverables.
AI with clear boundaries
- AI proposes column mappings that the customer confirms.
- The AI interviewer gathers business context one question at a time.
- AI interprets only results calculated by the engine.
- Customer data is not used to train language models.
Data is processed within the EU
- Client files are stored encrypted in Microsoft Azure Blob Storage in Sweden; transcripts, analysis results and reports are stored in Azure Database for PostgreSQL in Ireland — all within the EU with a high level of security. Account, login and credit information is stored with Supabase.
- Deterministic calculations run locally in a Web Worker in your own browser, using versioned code and traceable steps — not on our servers.
- AI support runs in highly secure Microsoft Azure OpenAI EU Data Zone. Customer data is not used to train the models. Microsoft may retain data for abuse monitoring under its applicable terms; we do not promise zero retention.
- Data is segregated by account and analysis.