Identification starts with evidence
Generic lists begin with what AI can do. Evidence-backed identification begins with what is true about the company: where work slows down, which decisions depend on scarce judgment, what data is reachable, where customer pressure is building, what constraints apply, and which capabilities competitors cannot easily copy.
Opportunity patterns fire on facts
Foundry's 52 operators represent recurring AI value patterns across automation, augmentation, analytics, software, physical operations, R&D, compliance, growth, domain workflows, and new offerings. An operator fires only when its required graph facts and relationships are present. A support assistant, for example, should not outrank a painful core workflow when no proprietary support knowledge or linked pain supports it.
Prioritization is not a single ROI guess
Each candidate carries written reasons across three core axes:
- Impact — does the linked pain, frequency, cost, or strategic pressure matter?
- Feasibility — are the required data, skills, systems, ownership, and controls available?
- Moat — would the opportunity compound proprietary data, scarce expertise, reusable capability, distribution, or switching cost?
Quality, risk, and timing checks sharpen the decision without hiding the underlying evidence.
Sequence before funding
The roadmap distinguishes quick wins from strategic bets. It also promotes instrumentation, data access, eval coverage, risk guardrails, and other prerequisites into visible roadmap items. Missing evidence becomes a precise interview or validation question rather than a vague caveat.