• Identity: A portal that cannot distinguish which kind of partner it is talking to routes everyone incorrectly, no matter how well the downstream flows are built.
• Consequence: Wholesale vs. Revenue Share, presented as a radio button, was the highest-risk moment in the flow. The consequences had to be visible before the choice was made.

Scale: Partners had built personal relationships to compensate for the absence of a structured support system.

The IA was the first deliverable and the most consequential. Before wireframes, before screens, the portal needed a structural decision: what is the organizing principle, and does it hold as the product grows?

Partner type became the principle. Not product catalog. Not deal stage. Partner type. Every navigation section, every workflow branch, every default view flows from that single decision. An accountant and a VAR log into the same portal and see different things because the IA resolves their identities at the entry point, not mid-task.

The sitemap was built to be extended, not replaced. New partner tiers, new program types, and new billing models can be added without restructuring what already exists. The decisions made in July 2025 created the conditions for a product that can scale without a redesign.

That is the standard for foundational IA work: it should read like a document that anticipated the future, not one that described the present.

Information architecture delivered July 27.

Wireframes delivered August 15.

Presented to Aprio in Atlanta on August 28, a 2.5-hour working session with the most sophisticated partner firm in the study. Four go/no-go decisions were brought to leadership with design tied to each outcome.

AI products fail the same way when they do not resolve user identity at the right moment. A system that treats all users as equivalent routes the right output to the wrong person. The question is identical: does the system know who it is talking to before it routes them anywhere? In AI products, a wrong first response often means a closed window and a user who does not come back. That is not a design preference. It is the architecture.