Dashboard/PRISM/Wiki/Decisions-2026-05-08-Phase-1-Product-Direction.md
PRISM

Wiki/Decisions-2026-05-08-Phase-1-Product-Direction.md

Decisions — Phase 1 Product Direction

Date: 2026-05-08 Status: Accepted

Context

PRISM is being shaped as a chat-first SaaS application that feels like Claude Desktop or ChatGPT, while delegating real execution to deterministic backend APIs, Temporal workflows, typed modules, and durable storage.

The open questions were about the first usable milestone, compatibility with the existing skill-generated report package, module sequencing, pilot persona, and whether artifacts should be editable.

Decisions

  1. Phase 1 supports both quick audit and full audit. The user should be able to run a quick audit/lookup or launch a full audit from the same chat interface. Full audit progress must be visible in chat.

  2. P0 optimizes for a better chat-native artifact package. Exact compatibility with the current report package is P1, not P0. P0 should focus on creating and presenting artifacts naturally in chat, with the right panel opening dynamically.

  3. Artifact interaction model follows Codex/Claude Desktop. Chat creates or references artifacts; the right panel opens to show the relevant artifact, such as an evidence board, screenshot gallery, business case, sales artifact, or report draft.

  4. Build audit-browser before intel-competitors. After intel-queries, rebuild browser audit next. Then build competitor intelligence.

  5. First pilot users are AEs and BDRs. Initial workflows should optimize for AE/BDR account research, prep, outreach, and editable sales artifacts.

  6. Editable artifacts are part of v1. v1 should not be read-only. Sales artifacts should be editable, versioned, and grounded back to evidence.

Consequences

  • Phase 1 needs two execution paths: quick audit and full audit, both chat-native.
  • The artifact system becomes a P0 product surface, not a later reporting feature.
  • The right panel should be treated as a dynamic artifact canvas, not a static dashboard drawer.
  • intel-queries and audit-browser move ahead of intel-competitors in the core audit build sequence.
  • AE/BDR workflows should drive IA, permissions, language, and artifact editing requirements.
  • Exact legacy report export compatibility should be preserved as a known P1 requirement but should not constrain P0 artifact exploration.