Help Center
AI & intelligence
Survey generation and review, configurable models (including on-prem inference), multilingual chat with data, narrative reports, translation, sentiment, themes, mood cues, and operational hints.
What this guide helps you do
Use assistive AI safely across survey design, quality review, open-text analysis, narrative reports, and conversational exploration of your dataset.
Before you start
- Know your data governance rules for AI and external models.
- Have a clear research question.
- Confirm which workspace owns the survey and responses.
What you will do
- Confirm governance
- Generate and review drafts
- Run structured analysis passes
- Anchor narratives to evidence
- Operationalize insights
You are done when
- AI outputs are reviewed by a human before publication or executive use.
- Model and provider choices match policy.
- Stakeholders know how to interpret AI summaries and chat answers.
Overview
Sentink layers assistive AI across the survey lifecycle. Generate questionnaires from prompts or uploads, audit logic with Quality Check (see Surveys), run core analysis against local models for sensitive datasets, optionally connect enterprise LLMs (OpenAI, Anthropic, Gemini, Ollama, compatible APIs), extract sentiment/themes from open-ended text with optional splits by gender and age when captured, analyze responses per open-ended field or jointly, surface response-level mood cues alongside polarity, assemble AI-assisted full reports anchored in your filtered dataset and active data layer, create two-sentence AI briefs, chat with the dataset in natural language, get question rewrites plus drop-off and tiered recommendations, and accelerate multilingual rollouts with AI-assisted translation—always reviewed by humans.
When to use this feature
Step-by-step guide
Confirm governance
Decide whether data must remain on local models, which vendors are approved, and who may change provider credentials.
Generate and review drafts
Start from prompts or documents, then reconcile AI output with stakeholder requirements.
Run structured analysis passes
Scope sentiment/themes to individual open ends or merged sets; add splits when demographics exist.
Anchor narratives to evidence
Filter responses, activate the analytical data layer when needed, and generate summaries or deck-ready stories from the same filtered slice.
Operationalize insights
Share AI brief cards, automate follow-ups via integrations, or continue exploration with dataset chat.
Human oversight
AI speeds synthesis; approvals, bias review, and policy checks remain yours.
Screens / UI explanation
Workspace AI settings inherit into generation and summarization workflows.
Narratives should cite the timeframe, filters, and sample context you applied.
Translations still need bilingual review before high-stakes publication.
Best practices
Document model choice, temperature, or safety guardrails alongside each regulated program.
Pair thematic AI output with verbatim quotes stakeholders can audit.
Refresh AI summaries after materially new responses arrive.
Tips
Use briefs for executives and full narratives for analysts working the same filtered slice.
When chat answers feel generic, tighten filters or provide a sharper question.
Common issues
Provider errors after rotation.
Re-test API credentials, quotas, firewall rules for custom endpoints.
Themes miss nuance.
Increase sample coverage, tighten screener wording, run per-question scopes instead of an all-open-ends rollup.
FAQ
Can teams keep data inside their network?
Yes—core analysis paths can leverage local inference; pair with integrations only when policy allows egress.