# Roadmap Planning > Roadmap Planning - Skill: `uitbreidenos/roadmap-planning` (Agent Skill) - Install (CLI): `npx skillmds@latest add uitbreidenos/roadmap-planning` - Raw SKILL.md: https://api.skillmd.com/api/skills/uitbreidenos/roadmap-planning/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Product & Planning - Author: UitbreidenOS (https://skillmd.com/u/uitbreidenos) - Updated: 2026-09-22 - Page: https://skillmd.com/skills/uitbreidenos/roadmap-planning --- # Roadmap Planning ## When to activate You need to sequence features across 2+ quarters, align stakeholders on dependencies, or model team capacity constraints. Use this skill for quarterly planning cycles, major product strategy shifts, or when capacity allocation becomes contentious. --- ## When NOT to use - For single-sprint backlog prioritization (use Feature Prioritization instead) - When market direction is completely unclear (validate assumptions first with Market Research) - If you lack engineering velocity estimates or team size (gather baseline data first) - For tactical bug fixes or technical debt threads (roadmap is strategic, not tactical) --- ## Instructions ### Step 1: Gather Input Data **Required:** - Feature backlog (with estimated effort: S/M/L or story points) - Team capacity (points per sprint, number of sprints per quarter) - Strategic themes (OKRs, business priorities, competitive pressures) - Known dependencies (feature A must ship before feature B) **Optional but valuable:** - Risk flags from technical team (what could delay us?) - Customer commitments (contracts with timeline requirements) - Competitive threat timeline (when do we need to respond?) - Resource constraints (shared resources, hiring delays) **Template to gather data:** ```markdown ## Roadmap Input ### Team Capacity - Team size: [X engineers] - Velocity: [Y points/sprint] - Sprints per quarter: [Usually 3] - Quarterly capacity: Y × 3 = [Z points/quarter] - Buffer for bugs/debt: [20-30% typically] - Net capacity: [Z × 0.7 to 0.8] ### Prioritized Backlog (Top 20 features) | Feature | Theme | Points | RICE Score | Notes | |---------|-------|--------|------------|-------| | [Feature 1] | [Theme] | [Points] | [Score] | [Dependencies] | | [Feature 2] | [Theme] | [Points] | [Score] | [Dependencies] | ### Strategic Themes (Next 4 quarters) - Q2: [Theme] — rationale - Q3: [Theme] — rationale - Q4: [Theme] — rationale - Q1 2027: [Theme] — rationale ### Known Dependencies - Feature B ← Feature A (must complete first) - Feature D ← Feature C (must complete first) ### Risk Flags - [Risk] — Probability: [H/M/L], Mitigation: [how] ``` ### Step 2: Build the Roadmap Framework Create a **quarterly structure** using this model: ``` Q2 2026 (13 weeks × 3 = 39 work days available) ├─ Committed (75% capacity = 19 days) ├─ High Confidence (15% capacity = 3 days) └─ At-Risk (10% capacity = 2 days, watch closely) Q3 2026 ├─ Committed (75%) ├─ High Confidence (15%) └─ At-Risk (10%) ``` **Why this structure?** - **Committed:** Validated features, dependencies clear, team confident - **High Confidence:** Likely to ship, some unknowns, good to communicate - **At-Risk:** May slip to next quarter depending on Committed performance ### Step 3: Map Features to Quarters **Algorithm:** 1. Sort backlog by RICE score (highest first) 2. Check for dependencies — if Feature B depends on A, A must go in earlier quarter 3. Allocate to Committed bucket first (respect capacity line) 4. Allocate overflow to High Confidence 5. Flag risks and mark as At-Risk 6. Flag anything that blocks dependent features **Worked example:** ``` Backlog sorted by RICE: 1. Multi-turn conversations (9 pts) — no deps 2. RAG integration (11 pts) — no deps, but needed for #5 3. Model switching (5 pts) — no deps 4. Fine-tuning framework (6 pts) — no deps 5. Agentic workflows (8 pts) — depends on #2 (RAG) 6. Tool use framework (7 pts) — depends on #5 (agentic) Capacity: 26 pts/quarter (20 available after safety buffer) Q2 assignment: ✓ Multi-turn (9) — Committed (9/20) ✓ RAG (11) — Committed (20/20 FULL) At capacity, must defer rest. Q3 assignment: ✓ Model switching (5) — Committed (5/20) ✓ Fine-tuning (6) — Committed (11/20) ✓ Agentic workflows (8) — Committed (19/20, tight!) ✓ Tool use (7) — High Confidence (7 pts, but blocked by Q2 RAG) ``` ### Step 4: Surface Dependencies as a Dependency Graph ``` Q2 ──────────────────────────────────────────── ├─ Multi-turn conversations (9 pts) ✓ └─ RAG integration (11 pts) ✓ ↓ (must complete before Q3) Q3 ──────────────────────────────────────────── ├─ Model switching (5 pts) ✓ ├─ Fine-tuning (6 pts) ✓ ├─ Agentic workflows (8 pts) ✓ ← blocked on Q2 RAG └─ Tool use (7 pts) ⚠️ HIGH RISK ← blocked on Q3 Agentic Risk flags: 🚩 Agentic depends on Q2 RAG; if RAG slips, cascades to Q3 🚩 Tool use is month 2 of Q3; high risk of overflow to Q4 ``` ### Step 5: Create the Roadmap Artifact **Format: Quarterly Roadmap Markdown** ```markdown # Product Roadmap: Q2-Q4 2026 ## Q2 2026: Conversational AI Foundations ### Committed (19/20 points) - **Multi-turn conversations** (9 pts) - Context retention across exchanges - Session memory management - Ship: End of April - Success: 95%+ multi-turn retention rate - **RAG integration** (11 pts) - Document ingestion pipeline - Vector store integration (Pinecone) - Relevance ranking & reranking - Ship: Mid-May - Success: 85%+ NDCG@10 on benchmark ### High Confidence (0/20 points) - Capacity fully allocated this quarter ### At-Risk - None flagged --- ## Q3 2026: Reasoning & Agents ### Committed (19/20 points) - **Agentic workflows** (8 pts) - Tool invocation framework - Agent loop orchestration - Multi-step planning - Ship: End of August - Dependencies: RAG integration (Q2) - Success: Users create 5+ agent recipes - **Model switching** (5 pts) - Multi-model support (GPT-4, Claude, Llama) - Cost routing logic - Ship: Mid-August - Success: 30% of users switch models - **Fine-tuning framework** (6 pts) - User-provided training data support - Tuning parameter UI - Cost estimation - Ship: End-August - Success: 10+ customers run tuning jobs ### High Confidence (7/20 points) - **Tool use framework** (7 pts) - Custom tool integration - Tool parameter extraction - Error handling & retry logic - Ship: Early September (may overflow) - Dependencies: Agentic workflows (Q3) - Risk: Blocked by Agentic delays ### At-Risk - 🚩 Tool use framework may slip to Q4 due to Agentic dependencies --- ## Q4 2026: Enterprise Features ### Committed (18/20 points) - **Fine-grained access controls** (6 pts) - RBAC framework - Org/team/resource permissions - Audit logging - Success: SOC 2 Type II ready - **Custom model endpoint support** (8 pts) - Self-hosted model integration - vLLM, Ollama, SageMaker support - Success: 5+ customers deploy custom models - **Advanced monitoring** (4 pts) - Latency/token/cost dashboards - Alerting rules - Success: Ops teams adopt monitoring ### High Confidence (8/20 points) - **API rate limiting** (4 pts) - Tier-based limits - Burst allowance - Overage billing - **Batch processing API** (4 pts) - Async job submission - Cost optimization for large jobs - Success: Cost per token drops 40% ### At-Risk - None expected; this quarter has buffer --- ## Dependencies Summary ``` Q2 RAG integration ↓ Q3 Agentic workflows ↓ Q3 Tool use framework ↓ Q4 [unblocked] ``` --- ## Capacity Utilization | Quarter | Committed | High Conf | At-Risk | Total | % Capacity | |---------|-----------|-----------|---------|-------|------------| | Q2 2026 | 20/20 pts | 0/20 pts | 0 pts | 20 pts | 100% ✓ | | Q3 2026 | 19/20 pts | 7/20 pts | 2 pts | 28 pts | 140% 🚩 | | Q4 2026 | 18/20 pts | 8/20 pts | 0 pts | 26 pts | 130% 🚩 | **Recommendation:** Q3-Q4 overloaded by 30-40%. Either increase capacity or defer High Confidence items. --- ## Risk Register | Risk | Probability | Impact | Mitigation | Owner | |------|-------------|--------|------------|-------| | RAG integration delays | Medium | High | Spike engineering; reduce scope | [Eng Lead] | | Agentic complexity underestimated | Medium | Medium | Prototype first; validate architecture | [Tech Lead] | | Tool use blocked if Agentic slips | High | High | Parallelize where possible; reduce scope | [PM] | | Q3-Q4 overload causes burnout | Medium | High | Hire 1 engineer by August or defer features | [Eng Manager] | --- ## Success Metrics (Per Theme) **Q2: Foundations** - Multi-turn adoption: >50% of new users in first week - RAG NDCG: 85%+ on benchmark set - System reliability: <0.1% user-facing errors **Q3: Agents** - Agent creation rate: >10 recipes/day - Tool invocation success: >95% - Model switching adoption: 30%+ **Q4: Enterprise** - RBAC adoption: 40%+ of enterprise customers - Custom model deployments: 5+ - Monitoring dashboard usage: 2 logins/day per customer --- ## How Roadmap May Shift This roadmap is live and responsive to: - Competitive moves (flag in competitive-monitoring.sh) - Customer churn signals (NPS drops >5 points) - Tech blockers (if architecture spike blocks 2+ features) - Market TAM changes (>20% shift) - Unplanned opportunities (partnership, acquisition) Quarterly review cadence: Lock roadmap on the 1st of each quarter; allow repriorization within locked capacity. ``` ### Step 6: Share & Align **Deliverable format:** 1. **For executives:** One-pager with theme, capacity, risks 2. **For engineering:** Detailed roadmap with dependencies, effort, technical notes 3. **For sales/CS:** Customer-friendly version (benefits, not technical details) 4. **For investors:** Narrative (why these themes, market signals, competitive response) **Alignment meeting agenda (1 hour):** - 10 min: Theme rationale (why these features, why now) - 15 min: Sequencing & dependencies (why this order) - 15 min: Capacity & risks (what could go wrong) - 15 min: Trade-offs (what we're deferring and why) - 5 min: Next steps (owners, review cadence) ### Step 7: Maintain Through Quarter **Weekly standup review:** - Burndown: Are we tracking to committed? - Blockers: Are dependencies on track? - Scope creep: Any unplanned work sneaking in? **Risk escalation triggers:** - Any Committed feature >20% behind schedule - Any dependency at risk of missing its window - Any At-Risk item trending toward Committed - New higher-priority item requiring reallocation **Mid-quarter repriorization (if needed):** - Did market conditions shift materially? - Did a competitor make us react? - Did customer churn signal a need to change? - If yes: convene 30-min alignment, adjust High Confidence, communicate change --- ## Example: AI Ops Platform (Real-world) ``` Team: 4 engineers, 1 TPM Velocity: 9 points/sprint, 3 sprints/quarter = 27 points Buffer (30%): 8 points Net capacity: 19 points/quarter Q2 2026: RAG + Evaluation - Backtested retrieval (8 pts) [Committed] - Cost optimization (11 pts) [Committed, tight!] - Total: 19 pts (100% of net capacity) Q3 2026: Agentic + Scaling - Tool integrations (7 pts) [Committed] - Async execution (6 pts) [Committed] - Model routing (5 pts) [High Confidence, at-risk] - Total: 18 pts [Committed], 5 pts [High Confidence] → Q3 looks healthy; Q3 model routing can slip to Q4 if needed Q4 2026: Enterprise + Reliability - Fine-grained RBAC (8 pts) [Committed] - Monitoring dashboards (6 pts) [Committed] - SLA management (4 pts) [High Confidence] - Total: 14 pts [Committed], 4 pts [High Confidence] → Light quarter; good for debt paydown Risks: - If RAG (Q2) slips by 2+ weeks, cascades to Tool integration (Q3) - Agentic complexity may exceed 12 points (current split is 7+6) Mitigation: Spike first sprint of Q3; architecture review with CTO ``` --- ## Common Pitfalls & Fixes | Pitfall | Why It Happens | Fix | |---------|----------------|-----| | Roadmap ignores dependencies | Dependencies felt invisible in planning | Build dependency graph before assigning quarters | | Over-committing (120%+ capacity) | Feature pressure, FOMO on competitive threats | Lock Committed at 75% capacity; use High Confidence as buffer | | Features slip without consequence | No mechanism to escalate or reprioritize | Weekly burndown + risk register; mid-quarter repriorization trigger | | Roadmap doesn't talk to technical feasibility | PM and Eng teams planning separately | Joint planning session; Eng estimates effort, not PM guessing | | Roadmap disconnected from user research | Features picked by intuition, not validation | Each Committed feature tied to RICE score from User Research Synthesis | | Roadmap never shipped | Scope creep + one-off requests | Lock Committed; redirect asks to High Confidence or next quarter | --- ## Tools & Templates **Gantt chart template (Markdown):** ```markdown ## Q2 2026 Timeline Week 1 2 3 4 5 6 7 8 9 10 11 12 13 Feat1: [=======] Feat2: [==============] Feat3: [=====] (blocked on Feat2) ``` **Capacity model (spreadsheet CSV):** ``` Feature,Points,Q2,Q3,Q4,Dependencies,RICE Multi-turn,9,✓,,,None,78 RAG,11,✓,,,None,85 Model switching,5,,✓,,None,52 Agentic,8,,✓,,RAG,68 Tool use,7,,✓,,Agentic,55 ``` **Risk register template:** ``` | Feature | Risk | Prob | Impact | Mitigation | Status | |---------|------|------|--------|------------|--------| | RAG | Inference latency | M | H | Cache warm, optimize reranker | Monitoring | ``` --- ## Example See `/ai_product_manager_stack/README.md` — `roadmap-generator.sh` section for a worked example of generating a 4-quarter roadmap from scratch.