AI for product managers
Draft PRDs with testable acceptance criteria, synthesize interview transcripts, size experiments and write the metrics readout. All of it draws on the files already in your product folder.
Ask it for
“Draft a PRD for multi-currency invoicing from the discovery notes and the invoicing spec. Include testable acceptance criteria, non-goals and open questions.”
Multi-currency-invoicing-PRD.docx
“Code the 14 onboarding interview transcripts against our codebook. Count participants per theme, and pull a timestamped quote for each.”
Onboarding-interview-synthesis.xlsx
“Design the experiment comparing one-page and two-page checkout, powered to detect a 0.5-point lift on the 4.5% baseline, and set guardrail metrics.”
Checkout-experiment-plan.docx
“Write the readout for the onboarding checklist test that closed October 10, covering the sample ratio check, guardrail breaches and the ship or no-ship memo.”
Onboarding-test-readout.docx
“Run Van Westendorp on the 187 responses in pricing-survey.csv, then Gabor-Granger on the $29 to $79 ladder, and find the price points the results support.”
Pricing-analysis-VW-GG.xlsx
“Build the Q4 board operating update from the analytics and billing exports, following last quarter’s deck, and cover activation, cohort retention and NRR.”
Board-operating-update-Q4.pptx
How it works
- 1
Assemble the product folder
Interview transcripts, survey exports and the Amplitude and billing CSVs go into the folder. Past PRDs and specs, the research codebook and last quarter’s board deck round it out.
- 2
Ask for the deliverable
You get a coded interview synthesis with participant counts, a PRD draft, an experiment plan or readout, the pricing analysis, the weekly metrics narrative, release notes or a launch brief. Each is a finished file that cites its sources.
- 3
Verify every claim, then decide
Every quote and number names the file it came from, so checking a theme means rereading the cited lines, not the whole transcript. You confirm the criteria with engineering, make the ship call and move the tickets into Jira yourself.
Inside the Product Management mode
Pricing and packaging
Runs Van Westendorp and Gabor-Granger studies as each method requires, and reads the acceptable range, demand curve and price points off your responses.
Ab test design and readout
Plans checkout and landing-page tests with a hypothesis, randomization unit, sample size and guardrails, and writes the ship or no-ship memo when the test ends.
Operating metrics and reviews
Builds the weekly product review and the board operating update from your exports, with activation, cohort retention, churn split and NRR broken out.
Technical requirements and specifications
Turns feature notes into PRDs with numbered, testable acceptance criteria, plus the edge cases, non-goals and open questions engineering will ask about.
Qualitative research synthesis
Codes interviews, tickets and survey free-text against a codebook, counts people per theme rather than mentions, and traces every quote to its transcript.
- 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
Product managers and product owners in software, fintech, health tech, hardware and digital government, up to group PMs, product directors and chief product officers.
Questions
Will it invent customer quotes or statistics?
No, it quotes the transcripts in your folder and cites the file each quote came from. When it cannot reconcile something, it flags the gap rather than filling it with plausible text. None of that is a guarantee, though, so spot-check quotes against the recordings before the themes reach your deck. The final check before anything goes to stakeholders stays with you.
Can it decide whether to ship the experiment?
No, the ship decision stays with you and your team. The readout assembles the sample ratio check, power against the minimum detectable effect, the primary metric, guardrail movement and segment effects, then puts the recommendation in front of you with the numbers behind it. You decide what it means and write up the call in the decision log.
Which pricing research methods does it support?
It supports Van Westendorp, Gabor-Granger and conjoint, each handled the way the method dictates. For a plan or bundle review it drafts the four Van Westendorp price questions, runs the Gabor-Granger ladder over your candidate prices, and returns the acceptable range, the demand curve and the tiers the responses support. Use conjoint when features matter to the package as much as price does.
Will the PRD read like a generic template?
Not when the draft starts from your files. Add the old spec, the PRD format your engineers expect and the interview synthesis to the folder, and the requirements, edge cases and open questions come from those documents. Anything the sources do not support gets marked as unresolved rather than invented, so you see the gaps instead of a confident guess.
What happens to our roadmap and interview data?
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.