AI for UX research and design
Ultramarine Work reads the session notes, exported screens and survey spreadsheets already in your study folder. You get back the test plan, the coded transcripts, the readout and the WCAG audit.
Research-readout.pptxExampleAsk it for
“Draft the test plan for the unmoderated check deposit study, 10 tasks, 8 freelancers and 8 admins, and strip every interface word from the task text.”
Deposit-test-plan.docx“I need SUS and task success scored for the deposit sessions, with small-sample confidence intervals, and each finding stated as counts of participants.”
Deposit-usability-report.docx“Code these 14 interview transcripts against our old codebook, add codes for the themes we never named, and pull every verbatim with its timestamp.”
Interview-codebook.xlsx“Turn the coded notes into a readout for the product trio, one finding per slide, evidence first, with single-participant observations marked separately.”
Research-readout.pptx“Compare wave 3 and wave 4 of the tracker by plan type, with margins of error on every cell. Flag any wave-to-wave move that sits outside them.”
Tracker-wave-crosstabs.xlsx“Audit the 12 exported checkout screens against WCAG 2.2 AA and draft the ACR wording, with issues listed by success criterion, each with a severity and a fix.”
Checkout-ACR-draft.docx
How it works
- 1
Drop the study in a folder
Transcripts and session notes, exported screens, raw survey tabs, your codebook and last quarter’s readout all go in one folder. Keep it organized the way your team already saves files.
- 2
Ask for the deliverable
The test plan comes back with tasks that never name the button, the crosstabs carry margins of error, and the audit is sorted by success criterion. It cites the file behind every claim and marks what it could not reconcile.
- 3
Check quotes, make the call
Re-read the quotes against the timestamps, rework the codes it clustered too quickly, and choose which themes survive. You present the readout and decide what ships, because it never files, posts or signs anything.
Inside the UX Research & Product Design mode
User research and usability
Turns study goals into screeners, guides and task text without interface words, and reports SUS and task success with intervals that fit an 8-person study.
Qualitative research synthesis
One codebook covers all the transcripts. Each theme carries the count of participants who said it, and every verbatim points back to its file and timestamp.
Survey design and analysis
It drafts the questionnaire with skip logic and balanced scales, and toplines and crosstabs show which wave-to-wave moves outrun their margins of error.
Accessibility audit
WCAG 2.2 issues come with a success criterion, a severity and a fix, and the VPAT or ACR wording arrives as a draft you verify before procurement sees it.
- Office skills
- Excel, Word, PowerPoint, PDF
- Tools
- Read, edit and write files, Search files, Run code, Search and read the web, View images
- Connectors
- SharePoint, Google Drive, Gmail, and additional custom connectors
- House style
- Your fonts, layouts and rules, on every file
Who it’s for
UX researchers, product, interaction and service designers, content designers and UX writers, through design and research leadership.
Questions
Will it invent quotes from participants?
No, every verbatim it quotes is tied to the transcript file and timestamp it came from, and anything it could not find or confirm is marked as a gap rather than rounded into a theme. Misattributed quotes are a known failure of AI synthesis, so the checking stays with you. Read every file it touched before a theme reaches a deck.
Do we need participant consent before AI works on transcripts?
That turns on your consent forms and privacy notice, and the decision belongs to your legal or data protection team. UXPA’s code requires consent to cover how participant data is stored and used, and ethics reviewers increasingly expect a plain-language disclosure of AI processing. Content in the folder is sent to Paper Instruments’ servers to be processed, so weigh that against what your participants agreed to.
Can it run the screen reader tests for our VPAT?
No, keyboard-only navigation and screen reader testing with NVDA, JAWS or VoiceOver stay manual, and a conformance claim without them is a defect. What the mode does is check the screens you export against WCAG 2.2 success criteria, log every issue with a severity and a fix, and draft the VPAT or ACR wording. You run the assistive technology, verify each criterion and sign it.
Where does participant data actually go?
Ultramarine Work changes files only in the project’s folder. When a request needs it, it can read files elsewhere on your computer, never passwords, keys or app data, and you can turn that off in settings. The model runs on Paper Instruments’ servers, so the content it works on is sent there to be processed. Paper Instruments is SOC 2 Type II, and our security and compliance details are on our compliance page.
Can it moderate interviews for us?
No, sessions stay in the hands of the researcher, because moderation requires judgment calls you are accountable for. It drafts the plan and the guide, and it codes the transcripts. You moderate, interpret and decide what stakeholders see. The open question in the field is whether AI compromises research quality, so treat its themes as a first pass and check them against the raw data.