Users run desirability/feasibility/viability conversations with evidence, dispute, live concept reflection, and final accept/override with audit. The three pillars operate as checkpoint gates: each feature must receive a terminal verdict (pass / fail / needs revisit) on all three before it can exit the filter into release allocation. See PRD → Three-Pillar Checkpoint Gates for gate definitions and pass criteria.
As a user, I want the left rail to show Desirability → Feasibility → Viability as my path through the filter, with a way to learn why those pillars matter, a way to choose Collaborate vs Let-AI-handle-it mode upfront, and a three-column board in the team room, So that the UI matches Quorum's methodology—not a generic "filter" black box—and I can pick how much I want to participate in the analysis.
Maps to: FR21, UX-DR4, UX-DR12, UX-DR13, UX-DR14 (ux-design-specification.md §4.2.1–§4.2.2).
Acceptance Criteria:
Given the workspace is in step 3 (Three-pillar analysis)
When I view the app shell
Then the full pipeline remains visible in the left rail (§4.2.1), and step 3 is expanded to show Desirability → Feasibility → Viability as nested items (each with Not started / In progress / Complete, optional Needs revisit)
And the current pillar is visually highlighted and drives the session context (header / orchestration hook)
And no invented top-level methodology steps appear (not "Filter" as the only label for step 3, not "Conflict resolution" or "Sync" as rail spine)
Given entering step 3 for the first time
When I am prompted to choose a mode
Then I see two options: Collaborate (work alongside agents, discuss each evaluation, argue back) and Let AI handle it (agents run full analysis independently, I review output and override)
And mode selection is persisted and visible throughout the step
And mode can be changed mid-step if I choose to, with clear confirmation
Given I am in the filter
When I open “Why these pillars?” (or equivalent ⓘ)
Then I see concise copy explaining Desirability (user/value), Feasibility (build/operate), Viability (should we / business & strategy) and why all three together matter
And optional per-pillar short blurbs are available without blocking progress
Given filter step active
When I view the team room
Then three columns render — Desirability | Feasibility | Viability — with groupings from Epic 3
And layout matches updated Figma / figma-handoff-brief.md Frame 4
As a user, I want evidence, sources, and rationale visible, So that nothing feels like a black box.
Maps to: FR22, NFR-P1 (expand/collapse performance).
Acceptance Criteria:
Given pillar results
When I expand an evidence card
Then I see source titles, links, snippets, confidence
And keyboard-accessible expand (NFR-A1)
As a user, I want to dispute a conclusion and get deeper analysis, So that I can argue with the team.
Maps to: FR23.
Acceptance Criteria:
Given a pillar item
When I choose dispute with comment
Then orchestration runs follow-up and threads results
And agent backbone tone respected (non-sycophantic copy templates)
As a user, I want concept views to reflect filter-driven scope changes, So that visuals match decisions.
Maps to: FR24.
Acceptance Criteria:
Given scope changes from filter
When items are cut or merged
Then concept artifact or preview updates or shows diff callout
And fidelity badge remains Concept until refinement epic
As a user, I want to finalize recommendations or override with reason, So that I stay accountable.
Maps to: FR25, FR25d, UX-DR5, NFR-S6.
Acceptance Criteria:
Given finalization action
When user overrides a gate verdict (pass → fail, fail → pass, or needs-revisit → forced resolution)
Then modal collects reason and shows audit notice
And decision stored append-only for audit (FR45 alignment, FR25d)
And both the original verdict and the override remain visible on the record — the override does not rewrite history
And override remains visible in filter context, not buried
And downstream artifacts (release allocation, PRD narrative) display an override marker against affected features
As a user, I want per-pillar rankings, an overall combined ranking, and MVP/V2/V3 release recommendations with reasoning, So that I leave the filter with a defensible, release-allocated feature set — not just a sorted list.
Maps to: FR25a, FR25c (allocation consumes only gate-cleared features).
Acceptance Criteria:
Given step 3 has completed (mode: Collaborate or Let-AI-handle-it) and every feature has a terminal verdict on all three gates (pass, fail, or user-override-with-audit)
When the output view loads
Then I see three per-pillar rankings (features ranked within Desirability, Feasibility, Viability individually, most-important to least)
And I see an overall combined ranking across all three pillars
And I see release allocation recommendations: MVP / V2 / V3 / deferred, with reasoning shown per allocation (what data was weighted, which gate verdicts drove the call)
And features that failed any gate without override are listed in a "did not clear" group with the failing gate(s) named — not hidden
And features that were overridden carry an explicit override marker in the output
And concept visuals reflect the final scope (features not in MVP are visually marked or filtered)
And the output is exportable alongside other planning artifacts
And the user can override any allocation with audit trail (FR25 alignment)
Given any feature still sits in needs revisit on any gate
When the user attempts to advance to release allocation
Then the system blocks advancement and lists the unresolved gates, prompting the user to either resolve each gate (via dispute loop, more evidence, or accept-as-is) or override (FR25c enforcement)
As a user, I want each pillar evaluation to produce a terminal verdict — pass, fail, or needs revisit — against documented pass criteria, So that I know whether a feature cleared the gate, not just where it ranked.
Maps to: FR25b, FR25c. Informs FR23 (dispute keeps a feature in needs-revisit).
Acceptance Criteria:
Given a feature under pillar evaluation
When the agent(s) complete the assessment
Then the feature receives a terminal verdict (pass / fail / needs revisit) on that pillar
And the verdict is paired with a confidence level (low / moderate / high)
And the evidence snippets, sources, and rationale that drove the verdict are attached and viewable (Story 4.2)
And the verdict is stored with the feature record (not just displayed)
Given a pillar card in the three-column board
When I view a feature on that card
Then I see the verdict badge (pass / fail / needs revisit) on the feature
And color and copy are distinct enough to read at a glance — pass, fail, and needs-revisit are never ambiguous (NFR-A4 contrast applies)
And hovering / focusing reveals the pass criteria tested for that pillar and how this feature scored against them
Given a feature in needs revisit on any pillar
When I attempt to finalize step 3 / advance to allocation
Then the system surfaces every unresolved feature-gate pair and blocks advancement (Story 4.6 alignment)
And I can resolve by: (a) running a dispute / deeper analysis (Story 4.3), (b) accepting as-is (forces pass), or (c) overriding with rationale (Story 4.5)
Given I am reviewing pillar verdicts
When I open a "Why this verdict?" affordance on any pillar card
Then I see the pass criteria for that pillar (what the agent tested for), the evidence collected, and the call — transparent not a black box (FR22 alignment)