wiki/decisions/2026-06-27-persona-journey-ui-architecture.md
ADR: Persona-Shaped UI, Three Doors Into One Intel Engine
⚠️ SUPERSEDED 2026-06-28 by 2026-06-28-prism-is-chowmes-prism-hermes-instance. This ADR designed a custom web app (three role-shaped UIs). PRISM is no longer a custom-built app, it is the Chowmes-PRISM Hermes instance, reached via Telegram (and a future in-SPA chat), not a bespoke dashboard. Kept for history; the persona insight (AE/BDR/Leader do different jobs) may still inform the chat UX, but the tabbed-dashboard/web-UI architecture here is not being built.
Date: 2026-06-27 Status: Superseded (2026-06-28)
Context
The first PRISM frontend build put a 6-tab dashboard (Overview / Research / Search Audit / Business Case / Competitive / Sales Actions) in the center panel and crammed the aRRIe chat into a narrow right rail. On review, the founder rejected it: directionally "like" the right reference apps (Claude Desktop, Codex Desktop) but functionally wrong.
The root problem: the build assumed one UI for all users. But PRISM's three target users do fundamentally different jobs, on different time horizons, with different hero surfaces. A single layout cannot serve all three. The AE wants depth on one account; the BDR wants volume triage across a list; the sales leader wants aggregate portfolio view. The tabbed dashboard that was wrong for the AE is actually right for the leader.
Decision
PRISM is three role-shaped entry surfaces over one shared intel/audit engine + artifact system. Build them in sequence: AE → BDR → Sales Leader.
| Role | Job (horizon) | Hero surface | Chat's role |
|---|---|---|---|
| AE | Walk into a call expert; leave with a quantified reason to buy (per-meeting/deal) | 60-sec pre-call brief + business-case artifact | Drill co-pilot |
| BDR | Find sharpest hook across 100+ accounts and send outreach (per-day volume) | Prioritized account queue + personalized outreach package | Triage command line |
| Sales Leader | See total opportunity, allocate effort, coach, forecast (weekly/quarterly) | Portfolio dashboard (heatmap, ranked accounts, $ opportunity) | Analytical query |
Design language confirmed: Claude Desktop calm as the resting aesthetic (warm-neutral, low chrome, content-first, generous whitespace, artifact panel slides in on demand) + Codex Desktop working-state for the audit-running moment (streaming step receipts, collapsible). Conversation-as-hero; artifacts slide in beside it rather than living in always-on tabs.
Rationale
- AE first, the deepest journey; it proves the underlying intelligence is actually good before breadth or aggregation is layered on. If the single-account intel isn't compelling, nothing downstream matters.
- BDR second, reuses the same intel engine, adding a queue (scoring/ranking) and an outreach package (ABX). Net-new surface, same brain.
- Leader last, aggregates what the AE and BDR journeys already produce; nothing to aggregate until those exist.
- The 6-tab dashboard isn't deleted, it's relocated to the leader, where an aggregate dashboard is the correct hero.
Alternatives Considered
| Option | Why rejected |
|---|---|
| One universal UI for all roles | The original build. Different jobs/horizons/hero surfaces; one layout serves none well. |
| Dashboard-first (tabs as default center) | Right for the leader only. For AE/BDR it buries the conversation and the deliverable behind navigation. |
| Build BDR or Leader first | Both depend on the intel engine being proven. AE is the depth test that de-risks the rest. |
Consequences
- AE journey is the next design target: 60-sec pre-call brief + drill-to-artifact flow, in calm Claude-desktop language.
- The current
frontend/center-panel tabbed dashboard is superseded for the AE surface (relocates to the future leader surface). - Frontend predates the v2 backend refactor, rewiring tool calls to v2 module contracts is prerequisite to any live AE mockup with real data.
- Research the real AE pre-call workflow before rendering (research scope to be approved separately).