How work moves through IRBIS.
The governed workflow from first survey to finished manuscript — and what the platform guarantees at every step.
Collect — surveys, files, and provenance
Data enters IRBIS through structured collection, not loose uploads. Surveys carry consent and supervisor review; files live in governed project folders; provenance attaches the moment data arrives.
- Design surveys with respondent consent built in
- Review submissions before they reach analysis
- Keep datasets, files, and versions in project folders
- Trace every dataset back to its source
Research — the literature, summarized
The Research Hub brainstorms hypotheses with you and summarizes the sources you bring, so analysis starts from evidence rather than guesswork.
- Brainstorm hypotheses with the assistant
- Summarize papers and sources you provide
- Keep research notes beside the project they serve
Analyze — vetted engines, any AI agent
Choose the AI agent that suits the task. Whatever the model, it works through the same vetted statistical operations and never sees raw data.
- Run 90+ reviewed statistical methods
- Let AI propose; authoritative engines execute
- Check assumptions as you go
- Switch agents without changing the rules
Publish — manuscripts that rebuild
Reports and analytical papers are assembled from deterministic EDA artifacts. Rerun the analysis and IRBIS rebuilds the manuscript, reporting what changed and why.
- Draft reports from real, citable artifacts
- Rebuild the manuscript from scratch on demand
- Export in English, Russian, or Kazakh
Governance — what holds at every step
The guarantees are architectural, not promises:
The firewall
AI receives aliases and aggregates — never raw rows.
The audit log
Every operation is recorded with its policy result.
Coins
AI usage is metered visibly. It pauses; it never bills extra.
Languages
English, Russian, and Kazakh are first-class across reports.