Personalization

What it means

The system adjusts rankings, tone, defaults, or suggestions based on who you are, what you did before, or what similar users did.

Why designers should care

Personalized AI needs controls (“Not relevant”), cold-start defaults, and parity for new users so the product does not feel broken on day one.

Example

A prompt library surfaces “Recently used” and “For your role: PM” sections, each with a “Why am I seeing this?” link to preference settings.

Common mistakes

• • Personalization that locks users into wrong personas with no escape hatch.
• • Dark patterns that use AI to maximize engagement over task completion.
• • Assuming personalization fixes bad base content.

Related terms

• Memory: Persistent stores (profile fields, summaries, vector memories) re-injected into future prompts so the assistant “remembers” you.
• Embeddings: Text is converted to vectors in high-dimensional space; closer vectors mean more similar meaning, enabling “find like this” search beyond keywords.
• Fine-Tuning: Additional training (full or lightweight) on your labeled data so the model’s default behavior skews toward your product’s patterns.
• Machine Learning: The product learns behavior from examples (clicks, labels, documents, feedback) and updates its predictions over time or across user segments.
• Few-Shot Prompting: You show the model one to several demonstrations of the task done correctly, then ask it to handle a new input the same way.