Hue: AI that shows its work.
Ask questions of responses, transcripts, documents, tables, and coded excerpts. Hue drafts and checks analysis with the supporting records pinned alongside, and runs on AI hosted in the United States under commercial API terms, so your study data is never used to train models.
What Hue does
Three things, done with the source in hand.
Hue is a study-aware assistant, not a chatbot bolted onto a report tool. Every answer, draft, and verification ties back to the responses in your study.
Ask your study questions in plain language.
Hue answers from your responses, codes, themes, and quantitative variables. Every answer points back to the source so you can verify before you cite.
- Filter by site, demographic, theme, or code on the fly
- Answers cite the response IDs they came from
- Open the underlying transcript in one click
Draft and revise sections from real evidence.
Theme write-ups, finding paragraphs, executive summaries, Hue assembles drafts from coded evidence already in your study. You stay in charge of voice, scope, and final wording.
- Drafts grounded in coded segments and source records
- Plain-language rewrites for non-specialist audiences
- Tracks which evidence supports each paragraph
Check claims against the source before they are shared.
Paste a draft claim, ask Hue to compare it with the study record, and review the responses or documents that support, qualify, or challenge it.
- Highlights statements that may overreach the data
- Shows responses that qualify or challenge a draft finding
- Lists evaluation questions that still need evidence
How Hue is grounded
Grounded assistance with review points.
Scoped to your study
Hue uses the study materials available inside Heurista when it supports analysis and drafting.
Source-linked answers
Every finding points at the response, transcript, or document it came from. You can open the source from inside the conversation.
Editable by default
Everything Hue produces is a draft. You accept, edit, or reject, and Heurista records what you approved.
No training on your data
Hue conversations and study data are handled under commercial API terms and are not used to train public models.
Under the hood
Engineered like the research instrument it is.
The AI in Heurista is built to be examined: pipelines with review gates, scores you can audit, and outputs that carry their evidence with them.
Three-phase coding pipeline
AI proposes codes against your codebook, a reviewer accepts or rejects each assignment, and synthesis runs only on approved work. Every phase is logged.
Confidence-aware review
Each AI code assignment carries a confidence score, and the least certain ones are flagged for priority human review, so attention goes where the model is weakest.
Statistics are computed, never generated
Tests run in a deterministic statistics engine inside the platform. The AI can interpret a result when you ask; it cannot invent one.
Citation-integrity gates
Report drafts pass through gates that check quotes verbatim against the study record and strip any citation that cannot be traced, before the draft ever renders.
The right model for the job
High-volume passes run on fast models, synthesis on the strongest, with batch processing for large studies. The cost of an AI action is visible before you run it.
Provenance that survives synthesis
Sources keep their segment, session, and message identifiers through every AI step, so the final draft still knows exactly where each claim came from.
In the workflow
Where Hue helps.
Three stages where researchers reach for Hue most. Real prompts, the kind you’d type into the conversation while you’re mid-study.
During analysis
Sense-making across responses, codes, and themes.
While drafting
Pulling evidence into shape on the page.
Before submitting
Spot-checks before a report goes out the door.
AI guardrails
Built for reviewable research.
Hue is designed to support research work while keeping evidence, decisions, and changes visible to the team.
Drafts stay tied to study material.
Hue works from the material available in the study. When evidence is missing, partial, or outside the source base, that should remain visible for review.
Researchers decide what becomes a finding.
Hue can summarize patterns and suggest interpretations, but final analytical decisions stay with the research team.
Study changes remain reviewable.
Codebook changes, theme updates, and report edits go through researcher review before they become part of the study record.
Claims stay within the evidence base.
Hue can help check whether a statement is supported, partially supported, or unsupported by the available data and sources.
Test on real evidence
Try Hue on your next study.
Bring a real study. Ask Hue real questions. Check whether the answers, drafts, and source links hold up to your review process.