Concept·Concept Design·Augmentation·Developing·CON-082
Information Architecture Development
Value hypothesis
Explore multiple organizational options for product information architecture deeply in the same time typical assisted methods spend to create one or two.
Velocity · Quality
The designer drafts the sitemap, navigation hierarchy, content groupings, category structure, and entity relationships within a digital product in iterative dialogue with an LLM. Product context is supplied at the outset, including target users, their goals and abilities, a content inventory, and any constraints that could impact structure (technical platform, existing URL format, stakeholder preferences). The model returns organisational candidates, which the designer revises into viable options.
Risks in application
Homogenization
LLMs converge on the canonical IA patterns their training corpus contains: e-commerce taxonomies, SaaS feature hierarchies, documentation trees. Drafts produced this way look competent and familiar, but risk inheriting maladapted conventions, so the architecture fails to address needs inherent to the specific content.
Empathy Gap
Generated proposals risk organizing according to precedent, rather than according to users mental models. The hierarchy that reads as logical to the LLM may not match how the audience navigates.
Expertise that differentiates
Information Architecture
Plausible-but-wrong IA fails at the first card sort, and the designer's job is to sense what will survive contact with users and what will collapse. This sensibility depends on experience, prior research sessions, support-ticket patterns, and internal debates about what a thing is actually called - none of which the LLM can reach.
AI Fluency that assures
Goal and Task Awareness
Provide context the LLM cannot typically access: research session observations, support-ticket naming patterns, internal debates about what a category is actually called, and technical constraints on URL structures.
Possible Indicators
Iteration speed
Number of distinct structural alternatives the team evaluates per concept cycle, versus baseline for manual IA drafting of equivalent product scope.
Expert assessment delta
Difference in IA quality between AI-assisted and manual drafting baselines, rated by experienced information architects against user-task coverage, consistency, and scalability criteria.