# Enterprise Brain Milestones — Q2 2026 **Source:** [MOM 20260326](../phase-2/conversations/20260326-mom.md) **Planning date:** March 26, 2026 **Team:** Naveen Puttagunta, Yeshwanth Reddy Yerraguntla, Satyasri Prabhakar Mantripragada, Vara Kumar Jagarapu --- ## Overview Three milestones define the path from reactive copilot to full autonomous experience center. Each builds on the prior, with clear priority assignments and delivery dates. | Milestone | Priority | Target Date | Core Deliverable | | --------- | ------------ | -------------------- | ----------------------------------------------------- | | **M1** | P0 | ~April 1 (Wednesday) | Reactive Copilot with Invictus/Tata Steel integration | | **M2** | P1 | ~April 10 | Dynamic Workspace + Agent Training on the Fly | | **M3** | P2 → P3 → P4 | End of April | Full Experience Center (multi-source orchestration) | --- ## Strategic Context The reactive copilot (M1) is the foundation — a user-triggered Q&A mode that orchestrates multiple agents to synthesize answers with visual outputs. This is not a single-source text chatbot; it coordinates data source agents and delivers results through interactive charts and drill-downs. The dynamic workspace is the output delivery surface — the canvas through which results are communicated. It can be driven by either: - **Reactive copilot** (user-triggered, M1) - **Autonomous experience center** (proactive, M3) M2 bridges the gap by demonstrating AI-driven UI layout changes and enabling data source agents to train themselves without developer intervention. M3 completes the vision: fully integrated experience center blending reactive and proactive modes across multiple data sources. --- ## Milestone Summaries ### M1 — Reactive Copilot _(P0, ~April 1)_ **What ships:** - Reactive copilot with base patterns - Redesigned UI connected to Invictus/Tata Steel data source - Multi-agent orchestration (not single-source queries) - Visual outputs: interactive charts, drill-downs, not just text **Two parallel workstreams:** 1. Integrate redesigned UI code from design team (code, not Figma) 2. Prepare Invictus data and train the data agent **Known gaps:** - "Orange box" (PostgreSQL for system prompts/instructions) missing; prompts hardcoded for now - Architectural databases not fully set up; delivering plumbing **Success criteria:** - User asks question → system orchestrates agents → returns visual + textual answer - UI reflects new design patterns - Invictus data queryable **Detail:** [M1-reactive-copilot.md](M1-reactive-copilot.md) --- ### M2 — Dynamic Workspace + Agent Training _(P1, ~April 10)_ **What ships:** - Rudimentary experience center with AI-driven UI dynamics - Rich patterns (Three.js, D3.js) selected dynamically by UI agent - Data source agents train themselves on the fly - Simulation mode: static snapshots demonstrate UI layout changes **Key innovation:** - Real-time data refresh NOT required for M2 - Pre-prepared snapshots can be linked/unlinked to show AI driving layout changes - Focus: demonstrate the AI layer's control over the UI, not data freshness **Deferred to M3:** - Complex trainer-role features (arbback) **Pre-M2 work:** - Analyze Nihar's design patterns - Establish self-describing UI widget framework (metadata-driven) - Each pattern describes itself; UI agent selects based on context **Success criteria:** - AI changes UI layout dynamically in response to context - Data source agents can be trained without developer code changes - Visual pattern selection works across multiple widget types **Detail:** [M2-dynamic-workspace.md](M2-dynamic-workspace.md) --- ### M3 — Full Experience Center _(P2 → P3 → P4, end of April)_ **What ships:** - Fully integrated experience center blending reactive + proactive modes - Multi-data source orchestration - Data refresh pipeline - Core system working end-to-end as originally envisioned **Priority order (explicitly decided):** 1. **P2:** Full experience center (reactive + proactive integration) 2. **P3:** Multi-data source orchestration 3. **P4:** Data refresh capability **Architecture scope:** - RT core agent manages cross-source snapshots - RBAC for daily org data storage - Integration with two red boxes + one pink box **Success criteria:** - System autonomously surfaces insights without user query - Multiple data sources queried and synthesized in single response - Data refresh pipeline operational - User receives both reactive answers and proactive alerts **Detail:** [M3-full-experience-center.md](M3-full-experience-center.md) --- ## Cross-Cutting Requirements ### Terminology Standards - **Reactive copilot:** User-triggered Q&A mode (replaces "chat") - **Dynamic workspace:** Output delivery canvas (driven by reactive or proactive modes) - **Experience center:** Autonomous proactive mode - Avoid "chatbot" in product-facing communication ### Platform Independence - DUX must work with Claude, Codex, Gemini, and Copilot - Project files as markdowns for LLM-agnostic access - No formal solution yet; work ongoing with Watana ### Visualization Framework - Patterns/widgets carry metadata describing capabilities - UI agent selects pattern based on context - Widgets don't need domain knowledge; they describe themselves - Three.js and D3.js patterns managed by Nihar's team ### Quality & Context - Dakshi quality issues traced to insufficient project context - Fix: comprehensive context capture (full meeting transcripts, salient daily points) - Skills already supporting this: Storytelling mode, Vasa skill, Retrospective skill --- ## Risks & Dependencies | Risk | Impact | Mitigation | | ------------------------------------------- | ------------------------------- | ------------------------------------------------------ | | Orange box (PostgreSQL for prompts) missing | M1 ships with hardcoded prompts | Accept technical debt; plan refactor post-M1 | | UI pattern code access not granted | M2 blocked on widget framework | Urgent: grant Yeshwanth access this week | | LLM compatibility unresolved | DUX adoption limited | Watana collaboration; temporary workarounds acceptable | | Designer workflow friction | UI generation inefficiency | Process refinement; doc-first or UI-first flow TBD | | Operational apathy (VS Code integration) | Reduced team adoption | Architecture council + daily standup enforcement | --- ## Related Documents - [MOM 20260326](../phase-2/conversations/20260326-mom.md) — Source discussion - [User Interaction Modes](user-interaction-modes.md) — Reactive vs proactive framing - [Current Sprint](../phase-2/current-sprint.md) — Active sprint tracking --- **Last updated:** March 27, 2026