Help
How to use AI Product Factory — submit ideas, read scores, and act on results.
Submitting an idea
Title
A clear, specific name for the product or service. Avoid vague names like "AI app" — be concrete.
Description
What does it do, how does it work, and what problem does it solve? 2–5 sentences. The more specific, the better the score.
Target audience
Who pays for this? Name the job title, company type, or demographic. Avoid "everyone" — it signals unclear thinking.
Business goal
What does success look like? Examples: "€1M ARR in 18 months", "10 000 paying users", "replace our manual process".
Mode
Core = scored against your company strategy. Explore = any industry, wild ideas welcome. Choose before submitting — it affects how the Council evaluates the idea.
Attachments
Optional. Attach images, PDFs, or text files for additional context. Council agents read them as part of their evaluation.
Core vs. Explore mode
Core
- → Scored against your company strategy
- → "Strategic fit" is a rated dimension
- → Company context injected into all agents
- → Best for evaluating ideas in your domain
Explore ✦
- → Any industry, no domain restriction
- → Council evaluates purely on merit
- → Generator picks a random industry
- → Best for discovering unexpected opportunities
How the Council scores ideas
Eight AI agents run in two phases. In Phase 1, seven specialists evaluate the idea in parallel. In Phase 2, a Judge synthesises their assessments into a final verdict.
Market size & growth potential
How large and fast-growing is the addressable market? A score of 8–10 indicates a massive, expanding global opportunity. 1–3 means a tiny or shrinking niche.
Strategic fit (Core mode only)
How closely does the idea align with the configured company strategy (e.g. circular economy, ESG, refurbishment)? Not scored in Explore mode.
Competitive moat
How defensible is the business against competition? High scores mean strong network effects, proprietary data, or switching costs. Low scores mean anyone can copy it.
Technical feasibility
How realistic is it to build with today's technology and a small team? High = proven tech, low complexity. Low = unsolved research problems or heavy infrastructure.
Revenue potential
How much revenue could this realistically generate? High scores indicate a clear path to €10M+ ARR. Low scores mean unclear or structurally limited monetisation.
Speed to market
How quickly could this reach first paying customers? High = weeks with an MVP. Low = years of development before any revenue.
Customer demand evidence
How strong is existing proof that people want and will pay for this? High = documented complaints, waiting lists, paid alternatives. Low = speculation only.
Risk level
⚠ Inverse scale. HIGH score = LOW risk. 8–10 means a safe bet with proven model and clear demand. 1–3 means high regulatory, technical, or market adoption risk.
Interpreting Go / No-Go
✓ Go
The Council judged this idea strong enough to pursue. It does not mean the idea is guaranteed to succeed — it means the fundamentals are promising enough to warrant further investigation, a hypothesis test, or a PRD.
✗ No-Go
Insufficient market potential, too risky, or not differentiated enough in the Council's assessment. The analysis_points in the verdict banner explain the key reasons. Consider refining the idea and re-scoring.
The overall score (1–10) reflects the average of all dimension scores. A No-Go idea can still score 5–6 if some dimensions are strong — read the full rationale before discarding.
PRD flow (Go ideas only)
- 1
Go verdict
The Council returns a Go verdict. A "Generate PRD" button appears on the idea detail page.
- 2
Generate PRD
Haiku generates a structured Product Requirements Document: problem statement, goals, user stories, P0/P1/P2 priorities, and success metrics.
- 3
Edit & approve
Edit the PRD inline. When ready, change status to Approved → In dev → Shipped.
- 4
Push to GitHub
Optionally push the PRD to GitHub as an epic issue with child issues for each user story (requires a GitHub PAT in Settings).
Glossary
Council
The 8-agent AI system that scores every idea. Seven specialist agents evaluate different dimensions in parallel; a Judge synthesises them into a Go/No-Go verdict.
Go / No-Go
The Council's final verdict. Go = strong enough to pursue further. No-Go = insufficient potential, too risky, or not differentiated enough.
Core mode
Scoring aligned with a specific company strategy. The Council weighs strategic fit and uses company context when evaluating each dimension.
Explore mode
Open-ended scoring across any industry — no domain restriction. The Council evaluates the idea purely on its own merits.
Overall score
Average of all dimension scores (1–10). Displayed on the idea card and in the radar chart.
key_points
Factual bullets at the top of the idea describing what the product IS — no analysis, no opinions. Written by the Judge.
analysis_points
Sharp analytical bullets in the verdict banner — the decisive reasons the Council went Go or No-Go.
PRD
Product Requirements Document. Generated from any Go idea: problem statement, goals, user stories, feature priorities (P0/P1/P2), and success metrics.
Hypothesis
A testable assumption derived from a Go idea — used to validate the riskiest assumptions before building.
Prompt version
The version of the Council scoring prompt used. If an idea was scored with an older version, a warning badge appears and you can re-score.
Re-score
Re-runs all 8 Council agents on an idea using the current prompt version. Adds a new score row; the highest score is displayed.
Questions? Contact jyri.roselius@eco360.ai