AI Brand Assistant


Designing a trustworthy AI agent for brand governance and decision support
 

I led the design of the AI Brand Assistant, an AI-powered conversational experience that helps users navigate brand guidelines, assets, and compliance rules. The goal was not just to answer questions, but to create an assistant that users could trust when making brand-critical decisions.

The core challenge was designing an AI system that feels helpful and intelligent without becoming opaque, overconfident, or authoritative in areas where human judgment still matters.



Background


Brand guidelines are complex, fragmented, and often underused. Users frequently struggle to:
  • find relevant rules quickly
  • understand how guidelines apply to their specific context
  • verify whether their content aligns with brand standards

This made the problem well-suited for conversational AI — but also introduced high risk: incorrect or overconfident answers could directly impact brand integrity.


The Agentic UX Challenge


From the beginning, the main challenge wasn’t technical feasibility — it was agency and trust.

Key questions we had to solve:
  • How do we prevent the assistant from sounding “certain” when the answer is ambiguous?
  • How do we keep users in control instead of outsourcing decisions to AI?
  • How do we make the system’s reasoning understandable and contestable?

Early pilot results confirmed this risk:
  • Only 10% of users were satisfied
  • 70% had no strong opinion, signaling uncertainty rather than delight

The assistant was functioning, but not yet credible.


My Role


I led the design effort end to end, from concept to release, working closely with Product, Engineering, and Research. My responsibility extended beyond UI design to:
  • defining agent boundaries
  • shaping conversational behavior
  • setting principles for transparency, confidence calibration, and user control.


Key Design Decisions


︎ Designing the assistant as a collaborator, not an authority

Problem
Early versions answered questions confidently, even when brand guidelines were open to interpretation.

Decision
We deliberately designed responses to:
  • reference source guidelines explicitly
  • highlight uncertainty when rules depended on context
  • encourage users to review or confirm decisions

Why
The goal was to support decision-making — not replace it. This preserved user agency and reduced blind trust.



︎ Making AI reasoning visible

Problem
Users struggled to understand why the assistant gave certain answers, which reduced confidence.

Decision
We designed responses to surface:
  • where information came from
  • which guideline sections were referenced
  • what assumptions were being made

Why
Transparency was essential for trust, especially in brand-critical workflows.



︎ Iterating through feedback to calibrate confidence

The assistant was first released as a pilot to a limited audience. Feedback revealed where the assistant felt:
  • too vague
  • too confident
  • or insufficiently actionable

Through iterative refinement of tone, structure, and response patterns, satisfaction and engagement improved significantly, leading to a broader release less than a year later.


Beyond Q&A: Conversational search and asset discovery


The assistant is being extended to support:
  • natural language search across assets
  • conversational discovery workflows
  • seamless transitions between answers and actions.

This positioned the assistant as an entry point into the system, not just a chat interface.



Customization and brand alignment


I explored how organizations could customize the assistant to reflect their visual identity and brand tone. This reinforced the idea that the assistant is an extension of the brand team, not a generic AI layer.


Impact


  • Simplified access to complex brand guidelines through conversational interaction
  • Improved brand compliance by supporting informed, contextual decisions
  • Increased engagement by reducing friction in asset discovery and guideline navigation
  • Positioned the platform as a credible, forward-looking AI-enabled product — without compromising trust


System-level outcomes


This work established:
  • clear principles for AI behavior and boundaries
  • reusable conversational patterns for future AI features
  • a foundation for safely extending AI capabilities across the platform via APIs

The Brand Assistant became a reference point for how AI should be designed across Frontify — transparent, assistive, and user-controlled.
© Anna Lukyanchenko 2025 — all rights reserved