Project Ideation

Project Ideation

Turn high-level goals into framed, evaluable project candidates. Use customer/data gaps, cross-functional creativity, and engineering feasibility to avoid surface-feature ideation.

Core Principles

Start from outcomes. Deconstruct goals into measurable drivers before proposing features.

Ground gaps in evidence. Customer feedback, data, support pain, sales objections, product usage, and incidents beat brainstorming alone.

Diverge then converge. Generate broad options, then pressure-test and narrow.

Frame projects as hypotheses. A project candidate states problem, hypothesis, target outcome, rough scope, and success metric.

Feasibility shapes scope. Infeasible ideas loop back to reframe or downscope before roadmapping.

Inputs

• OKR, strategic goal, customer problem, or opportunity area.
• user/customer data, feedback, support tickets, sales notes, product analytics.
• current roadmap, constraints, technical context, known platform limits.
• stakeholder personas and decision owner.
• run id for run-telemetry.

Tool Awareness

Use grounding-brief with long-term memory MCP for past ideas, product decisions, customer notes, and strategy docs. Use code RAG for technical feasibility and existing capabilities. Use agent-fleet /council as cross-function jam with product/eng/design/GTM/data/executive lenses when council risk triggers. Use success-criteria-metrics, hypothesis-testing, handoff-packaging, and run-telemetry.

Council Policy

Use agent-fleet council levels. Ordinary ideation can use no council or one product/engineering lens. Use minimal/full council for cross-functional jam, high-opportunity-cost bets, major strategic tradeoffs, or conflicting constraints across product, engineering, GTM, data, and strategy. Required minimal-council/full-council must block if agent-fleet preflight fails.

Workflow Modes

• lite: fuzzy idea or OKR needs initial framing. Budget: one brief, no council, no full gap analysis unless evidence exists.
• default: product opportunity needs evidence-backed candidates and feasibility. Budget: focused customer/data/context refs, one converge loop, council_level max minimal-council.
• deep: cross-functional strategy, high opportunity cost, major GTM/design/data constraints, or leadership-visible bet. Budget: full jam/council allowed with reason and roadmap-ready handoff.
• Escalate: conflicting constraints, high opportunity cost, major feasibility uncertainty, or leadership commitment.
• Load: start with goal and evidence; load agent-fleet council protocol, metrics, handoff, or code RAG only when that phase is active.

Execution

Deconstruct goal. Break objective into drivers, constraints, target users, measurable sub-outcomes, and non-goals.

Analyze gaps. Compare current vs desired state using customer feedback, analytics, support signals, competitive context, and internal constraints.

Run cross-function jam. Decide whether ideation needs no council, one lens, agent-fleet minimal council, or full cross-functional council across product, engineering, design, GTM, data, and strategy.

Converge candidates. Cluster ideas by user problem, value lever, feasibility, and strategic fit.

Frame strongest projects. Use success-criteria-metrics to define problem, hypothesis, target outcome, rough scope, primary metric, guardrails, and assumptions.

Check technical feasibility and eng lift. Use code RAG, architecture context, and engineering lenses to estimate effort, dependencies, risk, reversibility, and unknowns.

Package for roadmap. Use handoff-packaging to send candidates to roadmapping with evidence, metrics, cost/risk, dependencies, and open questions. Use human-approval-gate when candidate framing needs accountable stakeholder selection before roadmap scoring.

Loop Controls

Diverge/converge loop. Run ideation passes until option space is broad enough, time cap hit, or repeated ideas dominate.

Framing feasibility loop. If candidate is infeasible, reframe, downscope, or discard; do not pass fantasy scope to roadmap.

Delta-only re-jam. Re-open only rejected or changed assumptions, not the full idea space.

State carry-forward. Keep idea clusters, rejected ideas, assumptions, feasibility notes, and candidate scores.

Output Contract

# Project Ideation Output

## Goal Deconstruction
- goal:
- drivers:
- target users:
- measurable sub-outcomes:

## Gaps
- gap:
  evidence:
  user/business impact:

## Candidate Projects
- name:
  problem:
  hypothesis:
  target outcome:
  rough scope:
  success metric:
  feasibility:
  eng lift:
  risks:
  dependencies:

## Rejected Or Deferred Ideas
- idea:
  reason:

## Roadmap Handoff
- recommended candidates:
- evidence links:
- telemetry path:

Success Criteria

• goals are decomposed into drivers and measurable outcomes.
• gaps cite evidence.
• jam produces multiple distinct options and constraints.
• candidates are framed as testable project hypotheses.
• feasibility and lift are realistic enough for roadmap scoring.
• handoff package is ready for roadmapping.
• run-telemetry events emitted.

Common Failure Modes

Feature-first brainstorm. Fix by returning to goal drivers and gaps.

Evidence-free ideas. Fix by requiring source links for gaps.

Fantasy roadmap input. Fix by feasibility loop and downscope.

Council homogeny. Fix by diverse lenses and explicit dissent preservation.

Self-Improvement

Track ideas that later scored well, failed feasibility, got rejected by leadership, or produced real OKR impact. Feed outcomes into backprop to improve gap sourcing, persona rosters, and framing templates.