# Recommend whether a release is ready to ship

Vessamoor · Product-management work sample
All company, customer and product information is original synthetic material.

## Situation and outcome
Vessamoor is a fictional company that sells Desk, a support workspace where AI drafts replies for support agents. Desk 4.2 is its October release: seven items, a planned release on Tue 20 Oct and a go/no-go review with the Head of Product on Fri 16 Oct at 10:00. The Head of Product decides whether it ships. You took over the release on 7 Oct when the previous PM moved teams, and the company's release agent has written a first recommendation that says everything is ready.

Your outcome is a **readiness recommendation** for the Head of Product: the status of each release criterion on the candidate build, the blockers, which items you propose to ship and which to hold back, and a proposed verdict with options and what each does to two customer commitments. At associate level, give the PM a disposition for each item and the questions only the PM or the Head of Product can settle, in place of a proposed verdict. For any item you propose to hold back, say how it is withheld, what depends on it and what QA must verify. A well-argued hold, or "this cannot be settled until these named people answer", is a valid result. The verdict itself, any waiver, deploying the release, writing release notes, fixing defects, redesigning the evals and talking to customers are not part of this work.

The evidence period is 2026-07-13 through 2026-10-15 inclusive. Examine 60 source records from plans and criteria, defect and test records, assurance notes, readiness exports, beta customer records and release-channel messages. The records, with the correction note, stand in for the people; do not contact fictional organisations or seek private data. The AI first recommendation is not evidence authority. Work with the supplied source IDs and item codes.

## Your role
You are joining the team as Product Manager.

Own the readiness recommendation for Friday's go/no-go review: the status of each criterion, the blockers, a proposed verdict with options and, for any item you propose to withhold, how it is withheld, what it does to customer commitments and what QA must verify. The Head of Product decides the verdict, any waiver and any change to the criteria; account owners speak to customers.

## About the team
An AI-native support-software team preparing a go/no-go readiness recommendation for its Head of Product's Friday review, ahead of a release planned for the following Tuesday.

A release agent compiles the readiness digest and the first recommendation you audit, and a model grades the draft evals with human spot checks. A person checks every criterion status against its evidence before the Head of Product sees it.

The Head of Product decides the verdict, any waiver and any change to the criteria; release managers deploy only on her recorded decision; account owners speak to customers; engineering fixes defects.

## Work to complete
1. Read sources/coverage-and-access.md before making claims about what a record covers. The release scope and the request to you are the first two records; the criteria, the severity policy and the calendar follow.
2. Audit the AI first recommendation against the original records, row by row. Keep what the records support, and change only what they do not, naming the record that shows it. A row that still stands is kept, with its reason.
3. Apply the criteria as they were approved. Say which build each result comes from, and count by release item, not by record. A repeated or relayed record is not a second observation.
4. Give each item a proposed disposition (ship, withhold, or needs evidence) tied to the criterion it meets or fails. For anything withheld, state the method, what depends on it, the effect on the two customer commitments and what QA must verify.
5. Take a rare but serious event seriously whatever its frequency, keep contrary evidence and real benefits visible without letting either set the verdict, and say what the evidence cannot settle.
6. Read the correction note (U001) in your download like any other source and apply it. Carry each change through the recommendation, the criteria status, the evidence table and the revision log, and keep the original wording of a record the note disputes. In evidence.csv, revise the affected row in place, cite U001 in its source or rationale, and keep the original wording and item code in that row's rationale (for example "was [old item code] in the table; changed to [new item code] by U001"). U001 does not get a row of its own.
7. Keep authority where it sits. Propose, do not decide: put the verdict, waivers, criteria changes and anything for customers to the people who own them, as requests (an associate hands the PM dispositions and decision questions instead of proposing the verdict or asking for waivers), and route open items to one owner each with a date.

## Time
You have 90 minutes of work time. The clock starts when you press Start.

After that comes a 10-minute break, then a written interview: up to eight typed questions about your own files, for about 15 minutes. If you submit early, your break starts straight away.

## Mid-task correction
There is no mid-task correction in this take-home. Everything you need is in the files you download at Start, including the correction note **updates/U001-release-manager-reply-v1.md**. Read it with the other sources, apply it where it changes your work, and cite it as U001.

## Tools and help
Use any tools and help you like: AI tools and agents, search, your own notes, friends or colleagues. You don't need a paid tool, an API key or an account.

Nothing you do in your own tools is watched. Only your files and your interview answers are reviewed. When you submit, you can say which tools or help you used. That is optional, and leaving it empty is never held against you.

## Submission
Hand in one ZIP with these files at its top level. Start from the copies in the templates folder.
- README.md: list every file you hand in.
- recommendation.md: your readiness recommendation, about 1,000 words at most. Give the proposed verdict labelled as a proposal (an associate gives per-item dispositions and questions for the PM instead), the status of each criterion on the candidate build, the blockers and which of them cannot be waived, each item's proposed disposition with the method, dependencies and verification for anything withheld, the effect on the two commitments, and the decisions that are the Head of Product's.
- criteria-status.csv: required columns criterion,item_id,status,blocking,waivable,proposed,method,verify,owner,next_check,source_ids. At most 30 rows, one owner each and the next check as a calendar date. Use one row for each criterion on the candidate build, with item_id F00 when it covers the whole release and the item code when it covers one item (add a row for each item that fails a criterion). Add one row for each item's disposition, writing "disposition" in the criterion column. A criterion that does not apply gets a row with the status "not applicable" and the reason in method. Example rows (format only): `C2,F00,[status],[yes/no],[yes/no],[proposed],[method],[what QA checks],[one owner],YYYY-MM-DD,[record IDs]` and `disposition,F03,[status],[yes/no],[yes/no],[ship/withhold/needs evidence],[how it is withheld],[what QA checks],[one owner],YYYY-MM-DD,[record IDs]`.
- evidence.csv: your template already has one row for each of the 60 records (R001–R060), with its source ID and item code filled in. Fill in the criterion each record bears on and a status such as supports, blocks, contrary, context, stale or excluded, and write one sentence of rationale on each row. Leave no record out. Add a row only when a record bears on a second criterion. Keep the item codes traceable and revise a row in place if the correction changes it. Cite the correction note as U001.
- revision-log.md: each error you found in the AI first recommendation with the record check that shows it, the rows you kept, and what the correction changed, about 600 words at most. No private prompts or unrelated files.
- questions.md: at most 10 open questions, each with one owner and a date.

Words are counted in the sentences you write in each document; headings, table rows, lists of record IDs and the README do not count.

Use plain text files only: .md, .txt, .csv or .json. Put your work in the files themselves; a link to a document somewhere else does not count. Your ZIP can be up to 5 MB, or 20 MB unzipped, with at most 39 files of up to 2 MB each. Programs, macros, password-protected ZIPs and ZIPs inside your ZIP are not accepted.

Upload your ZIP on your take-home page and wait for **Received and checked**. It normally appears within a few seconds of the upload finishing. If it has not appeared after a minute, upload the ZIP again; if it still does not appear, use **Help and support**. You can replace your ZIP until you submit. Submitting is final and starts your break.

If time runs out before you submit, the last ZIP marked **Received and checked** is handed in for you. If none was received, the team is told that you need support.

If something goes wrong, use **Help and support** on your take-home page. A technical problem is never counted against you. Writing to us does not pause the clock.

## What happens next
The written interview covers these parts of the job:
- Checks status against evidence
- Weighs risk and gaps
- Checks and repairs the AI draft
- Hands the decision to its owner
- Explains their choices
- Proposes a release that can be verified

An AI then drafts a review from your files and your answers. Each point quotes what it rests on. A person at the company reads the review and makes any decision. No one is hired, rejected or ranked automatically.
