Your AI agent invents product rules. Make it establish context first.

When you request an account settings screen from an AI agent, it generates a layout with toggles and a delete account option, even though those features weren't specified. The agent fills in missing details, treating assumptions as established product rules. To prevent this, the Design Decision Method suggests establishing context first.

This involves reading the brief and existing product to identify who the screen is for, how often it's used, the cost of mistakes, the primary device, known features, and any assumptions. These facts and assumptions are clearly labeled as known or assumed to differentiate them from the agent's guesses. For example, an account settings screen might be for account holders who manage their own profile, with email or password changes occurring occasionally and account deletion being rare.

The cost of errors is medium for email/password changes and high for delete or billing changes. The primary device is phone, with desktop as the secondary. The product currently has email, password, and three notification toggles, but no social login or team accounts. By establishing this context before generating the layout, the agent avoids inventing features the product doesn't have, such as a dark mode toggle or delete control.

This step ensures the AI agent produces a realistic and accurate representation of the product, preventing the introduction of unvalidated features that could be costly or confusing to users.