Supervisor & Examiner
Trust Pack
QualIntel OS · Updated June 2026
This document explains how QualIntel OS supports researcher-led qualitative analysis, what the platform does and does not do, how the audit trail works, and what you should look for when reviewing student or researcher submissions generated with QualIntel OS.
In brief
- ✓QualIntel OS surfaces candidate evidence. The researcher confirms or rejects every one.
- ✓Every AI suggestion accepted or rejected is logged with a timestamp and researcher attribution.
- ✓The platform does not interpret data, generate conclusions, or produce findings — the researcher does all of that.
- ✓Submissions include an AI Disclosure Statement and full audit trail, designed for examiner review.
- ✓The tool is methodology-aware: researchers select RTA, IPA, Grounded Theory, or other frameworks at the outset.
1. What QualIntel OS is — and what it is not
QualIntel OS is an AI-assisted qualitative research platform designed for postgraduate researchers, PhD candidates, and academic staff. Its role is cognitive scaffolding — holding the structural and organisational burden of qualitative analysis so the researcher can focus on interpretation.
The platform does not produce findings. It does not interpret data. It does not generate conclusions. These remain entirely the responsibility of the researcher.
QualIntel OS does
- ·Extract candidate codes from a research design (not from data)
- ·Surface semantically matched transcript segments for researcher review
- ·Organise accepted evidence against researcher-defined themes
- ·Draft thematic synthesis scaffolds using accepted evidence only
- ·Log every decision with timestamp and researcher attribution
- ·Generate a submission package with full AI disclosure
QualIntel OS does not
- ·Interpret what evidence means
- ·Generate conclusions or findings
- ·Accept evidence automatically — every suggestion requires researcher confirmation
- ·Train on or retain your research data after deletion
- ·Replace the researcher's analytical judgement
- ·Guarantee academic quality of the final submission
2. How the analysis workflow operates
The platform guides researchers through six structured steps. Each step requires active researcher input — there is no automated progression.
Research design upload
The researcher uploads their proposal, interview guide, or rubric. The AI reads this document to extract candidate codes — themes and subcodes anchored to the researcher's stated research design, not inferred from data.
Codebook review
The researcher reviews all candidate codes. They may accept, reject, rename, merge, or add codes manually. The final codebook is theirs — the AI's initial extraction is a starting point only.
Qualitative data upload
The researcher uploads interview transcripts, field notes, or other qualitative data sources. The platform embeds these as searchable segments — no analysis occurs automatically.
Evidence review
For each code, the platform surfaces semantically matched transcript segments as candidates. The researcher confirms or rejects each one individually. Accepted evidence is logged with the decision timestamp. Rejected evidence is retained in the audit trail.
Thematic synthesis
The researcher writes their thematic synthesis in a structured editor. An AI scaffold may be generated as a starting draft — but this uses only accepted evidence and the researcher's codebook. The researcher edits, revises, and owns every word.
Submission package export
The platform generates a ZIP containing: Evidence Pack (DOCX), Codebook, AI Disclosure Statement, Reflexivity Template, and an audit log. The AI Disclosure Statement explicitly names what the platform did and did not do.
3. The audit trail — what it records and why
Every action in QualIntel OS that involves an AI suggestion is logged to an immutable audit trail. The trail is included in the submission package and available to the researcher at any time during the project.
| What is logged | Why it matters for examiners |
|---|---|
| Every AI-suggested code at extraction | Shows the AI's starting point vs the researcher's final codebook — any divergence is visible |
| Every code the researcher kept, renamed, or deleted | Confirms the codebook reflects researcher judgement, not default AI output |
| Every evidence suggestion (accepted or rejected) | Confirms no evidence was auto-accepted; shows the researcher's selection rationale |
| Rejection decisions with timestamps | Documents that the researcher critically evaluated AI suggestions, not just accepted them |
| Synthesis generation inputs | Records which evidence set the synthesis draft was built from — fully traceable |
| AI model version and methodology mode | Reproducibility: the exact AI configuration used is on record |
4. Methodology support
QualIntel OS is not a generic text analysis tool. Researchers select a qualitative methodology at project setup, and the platform adapts its prompts, synthesis guidance, and quality checks accordingly.
RTA
Reflexive Thematic Analysis (Braun & Clarke)
IPA
Interpretative Phenomenological Analysis
GT
Grounded Theory (Charmaz constructivist)
CTA
Codebook Thematic Analysis
CA
Content Analysis
TA
Template Analysis
Gioia
Gioia Methodology
Methodology selection affects synthesis prompts, theme naming conventions, saturation criteria, and quality check thresholds — not just labelling.
5. The AI Disclosure Statement
Every QualIntel OS submission package includes an AI Disclosure Statement — a non-editable document generated by the platform that states exactly what AI assistance was used. Researchers cannot alter this document.
Example disclosure — extract
“This research used QualIntel OS (qualintel.io) for AI-assisted qualitative analysis support. The AI extracted 14 candidate codes from the uploaded research design document on [date]. The researcher accepted 11 codes, rejected 2, and added 3 manual codes. The final codebook of 14 codes was confirmed by the researcher before analysis began.
For each theme, the AI surfaced semantically matched transcript segments as candidate evidence. The researcher reviewed 187 candidate segments, accepting 94 and rejecting 93. No evidence was accepted automatically. The researcher wrote the thematic synthesis in the platform editor; an AI-generated draft scaffold was used as a starting point for [Theme 2] only and substantially revised.
AI assistance was limited to: code extraction from the research design, semantic evidence retrieval, and synthesis scaffolding. Interpretation, analytical judgement, and all conclusions are the researcher’s own.”
The disclosure statement is generated from the audit log — it reflects actual platform activity, not researcher self-report.
6. What examiners should look for
When reviewing a submission produced with QualIntel OS, the following indicators demonstrate methodological rigour:
Codebook divergence from initial AI extraction
Strong submission
Researcher modified, renamed, or rejected AI-suggested codes — shows critical engagement with the tool
Worth querying
Codebook is identical to initial AI extraction — may indicate passive acceptance
Evidence rejection rate
Strong submission
Researcher rejected a meaningful proportion of AI-surfaced evidence — demonstrates selectivity
Worth querying
Near-zero rejections across all themes — may indicate uncritical acceptance
Synthesis authorship
Strong submission
Thematic narrative is clearly researcher-voiced and references specific accepted evidence
Worth querying
Synthesis reads as AI-generated text without researcher interpretation
Reflexivity statement
Strong submission
Researcher reflects on how the tool affected their analysis process
Worth querying
Reflexivity statement absent or superficial
7. Data privacy and security
Research data uploaded to QualIntel OS is handled under the following conditions, which supervisors and ethics committees may wish to note:
- Research data is stored encrypted at rest and in transit (TLS 1.2+).
- Data is used solely to provide the QualIntel OS service — it is not used to train AI models.
- Anthropic's API (which powers AI features) does not train on API inputs under standard commercial agreements.
- Researchers can export all their data at any time and permanently delete their account and all associated data.
- The platform is GDPR-compliant. A Data Processing Agreement (DPA) is available for institutional use.
- Infrastructure is hosted in the United States (Railway, Qdrant). A full data residency statement is available at qualintel.io/privacy.
8. Academic integrity
QualIntel OS is designed to support, not replace, researcher analytical judgement. It is positioned as a methodological tool comparable to NVivo, MAXQDA, or Atlas.ti — tools that organise and support qualitative analysis without doing the analysis itself.
The AI disclosure statement included in every submission package is intended to support transparent reporting of AI assistance in line with institutional academic integrity policies. Researchers are responsible for disclosing AI use in accordance with their institution's requirements.
If your institution has specific policies on AI-assisted qualitative research tools, the audit trail and disclosure documentation produced by QualIntel OS is designed to satisfy those requirements. Contact us if your institution requires a specific disclosure format.
9. Contact and institutional enquiries
For institutional enquiries, pilot access, or questions about how QualIntel OS works, contact us at hello@qualintel.io.
A Data Processing Agreement (DPA) for institutional use is available at qualintel.io/dpa.
© 2026 QualIntel OS Limited. All rights reserved. Hamilton, New Zealand. · qualintel.io