Roadmapping

Roadmapping

Sequence project candidates into an executable and defensible roadmap. Make tradeoffs explicit across ROI, dependencies, themes, capacity, risk, and strategic narrative.

Core Principles

Roadmaps are tradeoff artifacts. Show why work is sequenced, not just what is listed.

Use consistent scoring. Compare candidates with shared criteria for value, cost, risk, confidence, and strategic fit.

Dependencies shape time. Do not schedule work before technical or organizational blockers are ready.

Capacity is real. Match work to actual people, skills, focus, and time.

Approval creates commitment. Leadership owns final plan and tradeoffs.

Inputs

• project candidates from project-ideation or existing backlog.
• OKRs, strategy, constraints, deadlines, dependencies.
• team capacity, skill availability, ownership, budget.
• historical delivery data from memory, git, project tools, or prior roadmaps.
• run id for run-telemetry.

Tool Awareness

Use grounding-brief with long-term memory MCP for prior roadmaps, strategy, team constraints, and decisions. Use code RAG for technical dependencies and platform constraints. Use success-criteria-metrics for ROI scoring, task-decomposition-planning for dependency-aware sequencing, agent-fleet /council for feasibility/alignment stress-test when risk triggers, handoff-packaging, human-approval-gate, and run-telemetry.

Council Policy

Use agent-fleet council levels. Simple sequencing can use no council or one execution/product lens. Use minimal/full council for leadership commitments, multi-team capacity conflicts, platform bets, high opportunity cost, or unresolved feasibility/risk disputes. Required minimal-council/full-council must block if agent-fleet preflight fails.

Workflow Modes

• lite: sequence a small set of candidates or prepare a concise decision brief. Budget: one scoring pass, no council, no full capacity model.
• default: roadmap planning with ROI, dependencies, horizons, and capacity. Budget: focused refs, one dependency/capacity loop, council_level max minimal-council.
• deep: leadership commitment, multi-team capacity conflict, platform bet, high opportunity cost, or disputed feasibility. Budget: full council allowed with reason and approval package.
• Escalate: resource commitment, conflicting priorities, irreversible sequencing, missed dependency risk, or executive decision.
• Load: start with candidates and strategy; load agent-fleet council protocol, capacity details, or approval templates only when needed.

Execution

Evaluate ROI. Score candidates by user/business value, strategic alignment, confidence, effort, risk, reversibility, time sensitivity, and learning value.

Map dependencies. Identify technical prerequisites, org dependencies, external commitments, staffing gaps, data/infrastructure blockers, and sequencing constraints.

Group themes and horizons. Organize into now, next, later with coherent themes that explain strategy and balance quick wins with long bets.

Allocate capacity. Match projects to teams/owners/skills and realistic time windows. Include buffers for support, interrupts, and unknowns.

Iterate allocation and dependencies. Adjust sequence when allocation reveals conflict or dependency mapping reveals hidden blocker.

Council review. Select council level from risk; use executive, product, engineering, architecture, cost, and risk lenses only when roadmap stakes justify agent-fleet minimal/full council.

Revise and approval. Apply council deltas, package final roadmap, and route to human-approval-gate for leadership commitment.

Loop Controls

Dependency capacity loop. Iterate until timeline is executable, cap hit, or blockers require leadership decision.

Council revise loop. When council level is minimal/full, council, revise, and re-review only raised issues until pass/pass-with-nits or escalation. For none/single-lens, record reason and only re-review if the lens finds a blocker.

Theme stability check. Re-group only when scoring or dependencies materially change the story.

State carry-forward. Keep scoring rationale, rejected sequences, dependency graph, capacity assumptions, and council issues.

Output Contract

# Roadmap

## Scoring Model
- criteria:
- weights:
- confidence notes:

## Candidate Scores
- project:
  value:
  cost:
  risk:
  confidence:
  strategic fit:
  score:
  rationale:

## Dependencies And Capacity
- dependency:
  blocks:
  owner:
  ready by:
- team/owner:
  capacity:
  assigned work:
  risk:

## Horizons
- now:
- next:
- later:

## Council And Approval
- council level/verdict:
- required changes:
- leadership ask:
- telemetry path:

Success Criteria

• candidates scored with consistent criteria.
• dependencies and prerequisites are explicit.
• capacity allocation is realistic.
• roadmap has now/next/later narrative and themes.
• council review either records skipped/one-lens reason or passes when minimal/full council is triggered.
• human approval package exists.
• run-telemetry events emitted.

Common Failure Modes

Loudest voice wins. Fix by shared scoring rubric and evidence.

Calendar fantasy. Fix by capacity allocation and dependency loop.

Theme salad. Fix by grouping around strategic outcomes, not org charts.

Approval ambiguity. Fix by exact human-approval-gate ask and tradeoffs.

Self-Improvement

Track roadmap commitments that slipped, delivered ROI, got descoped, or were blocked by missed dependencies. Feed outcomes into backprop to improve scoring, capacity assumptions, and council rubrics.