Bloomberg
2025 - 2026
ASKB and Emerging AI patterns
Evolving shared components and AI patterns through close collaboration with Bloomberg’s leading AI product.

Overview
An ASKB-First Design System
As Bloomberg’s future product direction centered on AskB, our design system adopted an ASKB-first approach. We prioritized the needs of Bloomberg’s leading AI product while considering how those designs could extend across the product ecosystem.
I worked from the Design System team, collaborating closely with the ASKB product team on component requirements and design decisions. I also organized the Feedback and Loading working groups within our initial AI patterns effort. My contribution connected immediate product needs with shared system decisions: what to standardize, where to allow variation, and what needed further validation.
Challenge
Supporting New Experiences Within an Established Visual Language
ASKB brought together conversational answers, citations, financial data, charts, documents, and analytical code. These features introduced new requirements for shared components and patterns. We worked with the ASKB team to discuss what each feature needed and how the design system should support it. Alongside industry AI conventions, we considered Bloomberg-specific requirements and existing visual meanings.
The central challenge was determining which solutions should become shared standards and which should remain specific to ASKB.

Inline citations connect AI-generated summaries to source documents, with relevant passages highlighted for review.
Practice
Defining the Button Hierarchy
While defining primary, secondary, and tertiary buttons, we encountered an additional ASKB requirement: an amber prompting button with greater visual emphasis. Amber already had established meanings within Bloomberg. It was the brand color, and legacy input fields used amber backgrounds. Making amber the standard primary button color could cause users to confuse actions with input fields.
We established white as the standard primary button treatment.
For ASKB’s amber button, we explored two approaches:
Ultra primary: an additional emphasis level above standard primary.
Branded button: a Bloomberg-specific treatment for prompting Bloomberg AI.
These interpretations have different implications. One extends the action hierarchy; the other expresses brand identity within an AI interaction. The amber treatment remains under validation and is currently used only in ASKB. Keeping that question open allows us to support the product while evaluating its broader system meaning.
Patterns
Developing AI Patterns Through Working Groups
We initially prioritized six AI patterns, including Feedback, Loading, and Citation. Each had a dedicated working group. I organized the Feedback and Loading groups. The Feedback group arrived at three approaches:
Answer-level feedback on an entire summary or response.
Inline feedback on an individual paragraph.
Product-level feedback through an overall rating scale.
These options make the scope of feedback pattern explicit. Evaluating a complete answer, flagging a particular passage, and rating the overall product capture different aspects of the experience. Across the AI patterns effort, we considered industry conventions, Bloomberg-specific needs, and potential applications beyond ASKB.

Image is showing an example of Feedback pattern approach #1 - Answer Level Feedback. Feedback controls sit beneath the complete response, allowing users to evaluate the answer as a whole.
Application
Using AskB to Ground System Discussions
ASKB provided concrete contexts for component and pattern decisions. Its screens combined several content types: compact artifact references, detailed charts, tables, news lists, code, and citations. These examples grounded discussions about shared structure, content-specific requirements, and visual hierarchy.
Working closely with the product team kept the design system connected to real feature needs. Our broader perspective brought another question into those discussions: how could a component or pattern remain useful when adopted by a different Bloomberg product?
Outcomes
Building a Shared Foundation for Bloomberg AI
ASKB became the starting point for Bloomberg’s design system component work and a leading context for developing shared AI patterns. This connected system development to real product requirements while keeping broader Bloomberg applications in view. The AI patterns effort initially covered six priority areas, including Feedback, Loading, and Citation, each supported by a dedicated working group. Together, this work created a foundation for considering industry AI conventions alongside Bloomberg-specific needs and identifying how patterns could extend to other products. I organized the Feedback and Loading working groups and collaborated closely with the ASKB team on component needs. My contribution connected detailed product discussions with broader questions of consistency, applicability, and system evolution.
The broader outcome was a product-grounded approach to developing Bloomberg’s AI design language: using ASKB’s requirements to shape shared patterns, distinguish common standards from product-specific treatments, and guide future cross-product expansion.



