Design a privacy-safe personalization system for Firefly

Problem Statement Description

Product context: Adobe is a creative, document, and marketing software company; its products include Creative Cloud, Photoshop, Illustrator, Acrobat, Adobe Express, Firefly, and Experience Cloud.

Adobe Firefly serves creators who use generative AI to produce images, graphics, marketing assets, and creative variations across professional and collaborative workflows. The question asks you to design a personalization system that can make Firefly feel more relevant to each creator’s style, intent, workflow, and history while maintaining strong privacy expectations.

Your task is to define what “privacy-safe personalization” should mean in this context, what user experiences it should support, and how the underlying technical system might responsibly use signals without exposing sensitive creative work, client assets, enterprise data, or personal identity. The focus is not only on personalization quality, but also on trust, control, transparency, and safe operation at Adobe scale.

As a Technical PM, you should frame the product requirements, data boundaries, system interfaces, reliability needs, consent model, security considerations, rollout plan, and trade-offs between utility and privacy. Consider both individual creators and teams or enterprise customers who may have stricter governance requirements.

The experience should consider:

- Which Firefly users and workflows need personalization, such as question assistance, style suggestions, asset generation, brand-aware outputs, or creative recommendations.

- What data signals may be used, how they are collected, stored, minimized, retained, or excluded.

- User controls for consent, opt-out, data deletion, visibility, and personalization reset.

- API and system requirements for profile generation, recommendation, model inference, feedback capture, and policy enforcement.

- Privacy, security, and compliance expectations for personal projects, client work, team libraries, and enterprise-managed accounts.

- Reliability, latency, fallback behavior, abuse prevention, and observability for a production personalization layer.

- Rollout strategy, experimentation boundaries, guardrail metrics, and how to detect privacy or quality regressions.

The goal is to describe a technically sound and user-trustworthy personalization system for Firefly that improves creative relevance without compromising privacy, professional confidence, or Adobe’s responsibility to creators and enterprise customers.