Analogy as an AI Product Capability: Designing Systems That Support Creative Reasoning

Many AI products focus on retrieval, summarization, classification, and generation. These capabilities are useful, but they do not exhaust what knowledge workers need. Keith J. Holyoak’s The Human Edge, based on the excerpt and table of contents, highlights a deeper cognitive capability: analogy. If AI products are to support innovation, education, strategy, and expert reasoning, they should help users work with analogies more rigorously.

Holyoak describes analogy as the ability to create new ideas by seeing one thing as another. The book’s scope includes relations, causal reasoning, analogy in the brain, child development, language, teaching, computational models, semantic relations, visual analogy, large-scale AI systems, and creativity. The key product insight is that analogy is not just metaphor. It is structured relational mapping.

Current generative AI systems are already good at producing analogy-like language. Ask for a metaphor for cloud computing, and the system will offer libraries, power grids, highways, or nervous systems. But product teams should not confuse analogy generation with analogical reasoning. A useful analogy must identify which relationships carry over from the source domain to the target domain, which do not, and what inferences are justified.

This distinction creates an opportunity for AI product design. Instead of building tools that merely generate creative comparisons, we can build tools that scaffold analogical thinking. A strategy assistant could ask users to define source and target domains, list key entities, map relationships, mark causal links, identify mismatches, and generate testable implications. A learning platform could use analogy to explain abstract concepts, then check whether the learner understands the limits of the analogy. A consulting tool could compare a client’s business model to cases from other industries while making structural differences explicit.

A product architecture for analogy support might include four components. The first is analogy retrieval: finding potentially relevant source cases from a knowledge base. The second is structure mapping: identifying entities, relations, and causal patterns. The third is critique: highlighting where the analogy breaks or becomes misleading. The fourth is application: translating the analogy into hypotheses, experiments, or teaching explanations.

This is especially relevant for enterprise AI. Teams often use analogies informally: “our platform is like an app store,” “our data architecture should be like a mesh,” “our AI assistant is like a junior analyst.” These analogies shape budgets, roles, and technical architecture. If they are shallow, they create strategic debt. AI tools can help by making the analogy explicit and contestable.

Consulting teams can productize this as an “analogical reasoning workshop” supported by AI. The workflow would combine human domain expertise with AI-generated candidate analogies, structured mapping templates, risk analysis, and decision outputs. The goal is not to let AI decide which analogy is right. The goal is to improve the quality of human reasoning.

As AI systems become more fluent, the market will value tools that help people think better, not merely write faster. Analogy is one of the most important places to start.