Executive Advisor Board — Build Plan
Executive Advisor Board — Build Plan
Date: April 14, 2026 Source roadmap: executive-advisor-roadmap-2026-04.md Total features: 21 across 4 tiers Estimated duration: ~18 days (Phases 0–2)
Phase 0: Foundation
Duration estimate: 1.5 days Goal: Close three known v1.0.0 gaps that require minimal effort but affect production quality and completeness.
Tasks
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 0.1 | Wire CRO, CPO, CCO agents into runtime | S | — | Three agents registered in get_active_agents(); .txt prompt files exist in agents/; agents appear in evaluation pipeline; all 11 agents run successfully in a test evaluation |
| 0.2 | Enforce budget cap in API call paths | S | — | ea_budget_cap setting checked before each Claude API call; user-friendly error displayed when budget exceeded; admin can configure cap in settings; cap resets on configurable schedule |
| 0.3 | Fix model reference mismatch | S | — | Plugin header description matches runtime default model; admin setting allows model selection from dropdown; selected model persists across requests |
Phase 1: Critical Path (Tier 1)
Duration estimate: 5.5 days Goal: Ship proposal version comparison and pre-evaluation scanning — the features that enable iterative refinement and cost savings.
Epic 1.1: Proposal Version Comparison
Roadmap rationale: No competitor offers before/after delta analysis showing how revisions improve scores; users need iterative feedback loops, not one-shot evaluation. Gap type: White-space
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 1.1.1 | Add version tracking to evaluation storage | S | 0.1 | Evaluation results stored with version number and parent version reference; DB schema supports version chain queries |
| 1.1.2 | Build “Revise & Resubmit” flow | M | 1.1.1 | User can select a past evaluation and click “Revise”; form pre-fills with previous proposal data; user edits and resubmits; new evaluation linked to previous version |
| 1.1.3 | Generate delta report with per-agent score changes | M | 1.1.2 | After re-evaluation, a comparison view shows side-by-side scores per agent; highlights improved, unchanged, and declined areas; consolidation pass notes version-over-version progress |
Epic 1.2: Pre-Evaluation Proposal Scan
Roadmap rationale: Saves users money by catching incomplete or weak proposals before running the full 11-agent evaluation. Gap type: AI-native
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 1.2.1 | Implement lightweight pre-scan Claude call | M | 0.2 | ~500 token Claude call scans submitted proposal for missing sections, weak areas, and obvious gaps; runs before full evaluation pipeline; respects budget cap |
| 1.2.2 | Build pre-scan results UI with proceed/revise options | S | 1.2.1 | User sees scan results with flagged issues; can choose “Proceed to full evaluation” or “Revise proposal first”; pre-scan cost displayed separately |
Phase 2: High Value (Tier 2)
Duration estimate: 10.5 days Goal: Ship persistent organizational context, boardroom questions, agent deliberation, multi-model support, and industry templates — the features that maintain competitive positioning against Consensus AI.
Epic 2.1: Persistent Organizational Context
Roadmap rationale: #1 Reddit complaint about AI tools is memory/context loss (91+ hrs/year wasted); self-hosted WordPress storage provides data sovereignty no SaaS competitor can match. Gap type: White-space
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 2.1.1 | Design and create org profile DB table | S | — | Custom DB table stores industry, size, strategy summary, and key context fields; migration runs on plugin activation; CRUD operations via ITI Database Base class |
| 2.1.2 | Build org profile admin editor | M | 2.1.1 | Admin can enter and update organizational profile; fields include industry, company size, strategic priorities, key challenges, past evaluation summaries |
| 2.1.3 | Inject org context into agent prompts | S | 2.1.2 | Org profile data automatically injected into each agent’s prompt context; agents reference org-specific information in evaluations; injection respects token budget |
| 2.1.4 | Auto-accumulate context from evaluations | S | 2.1.3, 1.1.1 | After each evaluation, key findings are summarized and appended to org profile; user can review and edit auto-accumulated context |
Epic 2.2: Boardroom Questions Drill Mode
Roadmap rationale: Users want to “anticipate questions from different perspectives” — no competitor offers preparation coaching. Gap type: White-space
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 2.2.1 | Implement boardroom question generation | M | 0.1 | Post-evaluation feature generates 5-10 tough questions per agent; questions reference specific weaknesses identified in evaluation; coaching hints included per question |
| 2.2.2 | Build exportable preparation worksheet | S | 2.2.1 | Questions and coaching displayed in a drill format; exportable as PDF worksheet; grouped by agent/perspective |
Epic 2.3: Agent-to-Agent Deliberation Display
Roadmap rationale: Consensus AI has debate mechanics; observable deliberation differentiates by showing the reasoning chain. Gap type: Eroded differentiator recovery
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 2.3.1 | Spike: deliberation prompt engineering and token cost modeling | S | — | Document prompt structure for agent rebuttals; model token cost for deliberation round (11 agents × 1 rebuttal each); define conflict detection threshold |
| 2.3.2 | Implement agent rebuttal pass | M | 2.3.1, 0.1 | After individual evaluations, agents with conflicting scores exchange one rebuttal each; rebuttals reference specific claims from opposing agent; displayed as threaded debate |
| 2.3.3 | Build deliberation UI thread | M | 2.3.2 | Deliberation displayed as an observable debate thread below individual evaluations; user can expand/collapse per-agent rebuttals; deliberation is optional (user can skip) |
Epic 2.4: Multi-Model Provider Support
Roadmap rationale: Consensus AI and OpenClaw support multiple models; users want flexibility and cost control. Gap type: Parity
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 2.4.1 | Abstract model selection in shared Claude API client | M | — | Model provider interface supports Claude, GPT-4o, and Gemini; per-agent model assignment configurable in admin; API key management for each provider |
Epic 2.5: Industry-Specific Evaluation Templates
Roadmap rationale: Verve Intelligence does vertical validation; industry-tuned evaluation increases relevance for each vertical. Gap type: White-space
| ID | Task | Effort | Dependencies | Acceptance Criteria |
|---|---|---|---|---|
| 2.5.1 | Create 4 industry templates with form defaults and rubric adjustments | M | 0.1, 2.1.1 | SaaS/Tech, Healthcare, Manufacturing, Nonprofit templates each include pre-configured form defaults and agent rubric adjustments; selectable at proposal creation; template selection auto-populates relevant fields |
Phase 3: Strategic (Tier 3) — Epic-Level Only
| Epic | Features | Estimated Effort | Key Dependencies |
|---|---|---|---|
| 3.1 | Tavily-powered market context injection — real-time industry/competitor data injected into agent evaluations | L | Phase 2 complete; Tavily API integration pattern |
| 3.2 | RAG knowledge base per organization — Pinecone + user-uploaded company docs for context-aware evaluation | L | Epic 2.1 complete; Pinecone shared library client |
| 3.3 | Agent personality tuning — conservative ↔ aggressive risk tolerance slider per agent | M | Phase 2 complete |
| 3.4 | Executive summary → action plan pipeline — structured implementation plan from board recommendations | M | Phase 2 complete |
| 3.5 | Evaluation history dashboard — past evaluations, score trends, improvement tracking | M | Epic 1.1 (versioning) complete |
Phase 4: Exploratory (Tier 4) — List Only
- [ ] Parallel agent processing — concurrent agent execution (3-4 at a time) for faster evaluation; requires rate limit management and UI redesign
- [ ] Mobile-optimized evaluation view — dedicated mobile layout for reviewing results and chat (form submission remains desktop-recommended)
- [ ] WordPress Multisite support — shared agent library with per-site customization and centralized billing
- [ ] SaaS distribution option — hosted version requiring auth/storage layer rewrite
- [ ] Board composition presets — named board configurations (Growth Board, Risk Board) with one-click selection
- [ ] Webhook/Zapier integration — post-evaluation webhook for Slack, CRM, and project management integration
Risk Register
| Risk | Phase | Likelihood | Impact | Mitigation |
|---|---|---|---|---|
| Agent deliberation doubles token costs per evaluation | 2 | High | Medium | Make deliberation optional; model token budget in spike (2.3.1); consider using cheaper model for rebuttals; set deliberation budget cap |
| Multi-model provider abstraction breaks existing Claude-specific features | 2 | Medium | High | Maintain Claude as default; add providers incrementally; comprehensive regression testing per provider |
| Persistent org context accumulates stale or incorrect data over time | 2 | Medium | Medium | User can review and edit accumulated context; add “last updated” timestamps; flag stale entries after 90 days |
| Industry templates require deep domain expertise per vertical | 2 | Low | Medium | Start with 4 well-researched verticals; use proposal-evaluation Skill for rubric design; iterate based on user feedback |
| Consensus AI ships features faster as a funded SaaS | All | High | High | Lean into WordPress self-hosted differentiator; prioritize features Consensus can’t match (data sovereignty, extensibility, WordPress CPT customization) |
Dependency Graph
0.1 (Wire agents) ─┬─→ 1.1.1 (Version tracking) → 1.1.2 (Revise flow) → 1.1.3 (Delta report)
├─→ 2.2.1 (Boardroom questions) → 2.2.2 (Worksheet export)
└─→ 2.3.2 (Agent rebuttal pass) → 2.3.3 (Deliberation UI)
0.2 (Budget cap) ──→ 1.2.1 (Pre-scan call) → 1.2.2 (Pre-scan UI)
0.3 (Model fix) ──→ (no downstream blockers)
2.1.1 (Org DB) → 2.1.2 (Org editor) → 2.1.3 (Prompt injection) → 2.1.4 (Auto-accumulate)
2.3.1 (Deliberation spike) → 2.3.2
2.1.1 → 2.5.1 (Industry templates)
Critical path: 0.1 → 1.1.1 → 1.1.2 → 1.1.3 (version comparison)
Parallel track: 0.2 → 1.2.1 → 1.2.2 (pre-scan)
Parallel track: 2.1.1 → 2.1.2 → 2.1.3 → 2.1.4 (org context)
New Skills / Shared Library Needed
| Skill or Component | Products Affected | When Needed |
|---|---|---|
multi-agent-deliberation-design (new Skill) |
Executive Advisor, future multi-agent products | Phase 2 — Epic 2.3 |
business-proposal-evaluation (new Skill) |
Executive Advisor, Career Coach | Phase 2 — Epic 2.5 industry templates |
Multi-model provider abstraction in class-iti-claude-api.php |
All AI products | Phase 2 — Epic 2.4 |
| Persistent user/org context DB pattern | Executive Advisor, Career Coach, Estate Manager | Phase 2 — Epic 2.1 |
| Pre-evaluation input validation pattern | Executive Advisor, Career Coach | Phase 1 — Epic 1.2 |
| PDF export improvements (shared library) | Executive Advisor, Career Coach, Journey Mapper | Phase 2 — worksheet export |
