Design proposal · not implemented
Prepared 28 Sep 2026
Synthetic sample data
Build with AI for written assessments
A recruiter links a requisition, chooses how much control to take, and gets a draft made of existing library questions to review. The flow extends today's Create written assessment modal and hands off to the existing editor. It never writes new questions and never publishes on its own.
Flow map
Every entry point opens the same modal. The proposed work sits between the existing entry points and the existing editor. Select any step to open it in the prototype.
Entry pointsExisting
Build with AIProposed
1 · Job detailsRequisition, job description preview, title
2 · Question setupConfirm the recommended mix, or adjust it in place
2 · Question setup, adjustedCounts and difficulty changed · reset available
Selecting questionsIndeterminate progress · can't be dismissed
Saved draftPopulated, status Draft
EditorExisting, with additions
Review questionsSummary, expand, Matches label
ReplaceOne slot, original kept until saved
RemoveThis assessment only
Add from libraryExcludes selected questions
Save and previewNo answers or hidden tests
PublishExisting, separate action. Outside this design.
RecoveryEach returns to the flow
Requisitions fail to loadTry again · create new
Role mapping unclearEdit requisition · add manually
Recommendation unavailableTry again · set up manually
Partial matchUse these · edit setup · browse
No matching questionsEdit setup · browse library
Selection unavailableTry again · browse library
Request timed outChecking draft status, no duplicate
Draft save failedSave again without re-selecting
Question became unavailableChoose a replacement
Requisition archived mid-flowChoose another, fields kept
Replacement failedOriginal kept · try again
Proposed Existing behavior Boundary of this design Recovery state
Deliverables
01Connected flow mapEntry points to the saved draft and the publish boundary. On this page.
02Desktop screensJob details, recommendation, Customize, selecting, draft, expanded question, replace drawer.
03Partial, empty and failure statesPartial and no results in full, compact variants for the rest.
04Mobile and dark mode390px setup and review, plus representative dark states.
05Clickable prototypeRuns in memory with a separate scenario picker for review.
06Component mappingRepository components, how each was checked, and explicit gaps.
07Developer handoffExisting behavior, proposed additions and open dependencies.
Scope
In this design
- Selecting existing multiple-choice and coding questions from the permitted libraries.
- Two compact steps in the existing modal, then the existing editor. Step two is a confirmation of the recommended mix.
- Review, remove, replace and add from library on the draft.
- Every empty, partial and failure state, each with a next action.
Not in this design
- Writing new questions or rewriting answers.
- Open-ended questions, until they are supported end to end.
- Publishing, candidate delivery, scoring, proctoring and video interviews.
- Credit estimates, speed claims, confidence scores or the model's name.
Sources and caveats
- Figma reference. Node 326:1972 in E2E-v2, read on 28 Sep 2026. A copy is saved at reference/figma-written-assessment-reference.png.
- Repository. The ATS folder could not be opened from this workspace. Component files were confirmed through the code index, and six source snapshots were read. The mapping is provisional until it is checked against current source.
- Product screenshot. The screenshot attached to the original brief did not reach this workspace. The Figma node above shows the same modal and was used instead.
- Tracking. Frontend and design handoff 86d2cmcv8, related backend 86d2cmctr.
- Sample data. REQ-1042 Backend Engineer, REQ-1051 Customer Support Lead and every question are invented for this design. Counts and difficulty mixes are illustrations, not library results.