Generate concepts from verified opportunity signals
Use AI to recombine a curated signal pack into traceable concepts, then have humans assess the concepts against real-world criteria.
Why
The workflow problem
Volume prompting recycles generic concepts disconnected from the research, market, operational, or technology signals that should constrain the opportunity.
A verified signal pack anchors generation in material the team can inspect and reduces generic ideation. The trade-off is that traceability can create false confidence: a cited signal does not validate the proposed mechanism.
Concrete output
What you will produce
A concept portfolio in which every concept names its intended user, source signals, mechanism, value proposition, assumptions, and evidence IDs.
How
Run the workflow
Curate and verify the signal pack
Humans select concise, decision-relevant signals from research and approved business, operational, or technology sources. Give each signal an ID, source, date, scope, and confidence note; remove unsupported trend claims.
Define the opportunity boundary
The team writes the intended user, target outcome, opportunity statement, excluded solutions, and evaluation criteria. An accountable owner confirms that the boundary does not overstate what the signals establish.
Generate traceable concept candidates
Give AI only the curated signal pack and require it to cite the IDs it combines. Treat concepts as hypotheses, not research-backed recommendations.
Recombine verified signals into concepts
Generate concepts only from the verified signal pack below. Each concept must combine at least two signal IDs and use a distinct mechanism. Intended user and opportunity: [insert] Evaluation criteria and exclusions: [insert] Verified signal pack: [ID | source | signal | confidence] Return: concept name | intended user | signal IDs | mechanism | value proposition | key assumption | criterion likely to fail first. Do not invent signals, quote sources not supplied, or rank concepts.Inspect evidence and operational reality
Humans open the original sources for every retained concept, correct misreadings, add operational knowledge, and reject concepts whose difference is only branding or language.
Rank and commission next tests
A multidisciplinary human group scores retained concepts against the stated criteria and selects tests. Preserve dissent and uncertainties in the portfolio rather than averaging them away.
Human–AI partnership
Who contributes what
AI contribution
Recombine supplied signals, structure concept descriptions, and surface assumptions. It cannot verify a signal, infer market demand, or select a concept.
Human responsibility
Verify the signal pack and originals, define the opportunity and criteria, add domain reality, evaluate trade-offs, and make all keep, kill, and test decisions.
Stop rule
Stop AI use when a concept lacks evidence IDs, misquotes a signal, adds unsupported claims, or produces diversity only through surface branding.
Check before use
Watch out
Signals may be outdated, non-representative, or commercially motivated. Check source authority and scope, protect confidential evidence, and avoid treating weak signals as a substitute for direct user research.
Evidence base
What supports this workflow
A Board of Innovation practice account documents signal-grounded AI concept generation, but it provides no independent benchmark. The traceability requirement is a risk control and an AIdea synthesis, not proof that grounded concepts will succeed.
- Signe Damgaard · How AI is rewiring the innovation function
Direct practice source for tying AI-generated concepts to consumer signals before generation.
- Stanford d.school · Idea Expedition
Provides canonical method context for moving from opportunity framing to concept alternatives.
- IDEO.org · Methods
Method library supporting concept development as a human-centred design activity.