F.2 Product Discovery

Engineering question. What should we build, and why?

Product discovery operates where uncertainty is often legitimate. Customer needs clash, features begin as hypotheses, requirements are negotiated, and experiments exist because important questions remain unsettled.

Product discovery therefore needs to preserve what the team knows and why it made particular choices. Useful models can include stakeholder needs, product hypotheses, experiment definitions and results, prototypes, decision records, provisional requirements, and relationships among them. These representations preserve what the organization knows about the product problem rather than primarily describing its implementation.

Product discovery therefore usually needs light Alignment. Hypotheses, evidence, and provisional decisions can be worth preserving without becoming obligations that realization must satisfy. Provenance matters: where did a claim come from? Relationships matter: which evidence bears on which hypothesis? Promotion matters: which provisional decision has become accepted product intent? Much of the judgment remains human because discovery is partly the work of deciding what the obligations should be. Rationales should be recorded, but extensive formalization is often unnecessary.

Agents can still provide substantial leverage. A reasoner can retrieve experiments relevant to a proposed capability, identify earlier product decisions it appears to contradict, trace stakeholder needs, compare competing evidence, or reconstruct why a previous direction was abandoned. Externalizing that knowledge reduces reconstruction without pretending the resulting judgment is mechanical.

Discovery also exposes the distinction among unknown, tacit, and free. An apparently unconstrained choice may be genuinely open, may conceal an obligation not yet discovered, or may depend on one stakeholders have not articulated. Discovery helps determine which.

MAGE profile.

• Characteristic models. Needs, hypotheses, experiment results, product decisions, provisional and accepted requirements.
• Lifetime. Mostly episode- and product-lived; decisions and requirements may outlive the investigations that produced them.
• Alignment posture. Light. Preserve provenance and decision status; avoid prematurely giving hypotheses mechanical authority.
• Role of autonomous reasoning. Synthesis, retrieval, comparison, ambiguity, and reasoning across competing evidence.
• Determinization opportunity. Limited while the underlying product question remains genuinely unsettled. Stable process obligations may still be mechanized.
• Degrees of freedom. Product choices that stakeholders have genuinely left open.
• Smallest useful adoption. Persist consequential product decisions, their rationale, and their relationships to needs and evidence; make that structure available to later reasoning.
• Lifecycle connections. Requirements entering Engineering and Assurance; change intent entering Product Management; operational evidence that tests product assumptions.