— PROJECT NAME
Ai conversational search experience
— ROLE
Lead UI designer
— CLIENT
Squiz
Santander
University of Greenwich
— SCOPE
Design an intuitive conversational ai experience connected to generic global search
— SKILLS
UX strategy
Prototyping
Governance workflows
Ai integration
The Challenge
Users across higher education, government, and enterprise websites expect quick, natural answers to complex queries but traditional keyword search is not always sufficient. The goal for Squiz Conversational Search was to introduce AI-powered Q&A functionality to surface trustworthy, conversational answers, while maintaining content governance and institutional voice.
Each sector posed unique challenges, for example:
Universities - Vast academic content, multiple departments, decentralised ownership.
Public sector - Accessibility, compliance, and clear audit trails.
Private sector - Brand voice, conversion goals, rapid content change.
— PROJECT OBJECTIVES
Objectives
- 🔹 Replace traditional keyword search with a natural language, conversational interface.
- 🔹 Ensure responses are factual, source-linked, and contextually relevant.
- 🔹 Provide content governance tools to control scope, tone, and quality.
🔹 Offer an embeddable UI with sector-specific configurability (accessibility, branding, tone).
Mockup created for University of Greenwich which was eventually sold as a phased implementation package
— DESIGN STRATEGY
We defined four core design principles:
🔹 Clarity – Answers must be easily scannable and source-linked.
🔹 Trust – Every response should reference real, approved content.
🔹 Control – Clients needed full transparency and control over AI behaviour.
🔹 Continuity – Multi-turn conversation history must enhance, not confuse, the user journey.
— DESIGN EXECUTION
Designed a flexible conversational UI component that could be embedded across client websites.
Introduced:
- 🔹 Real-time multi-turn chat interface
- 🔹 Inline source citations for traceability
- 🔹 Fallback flows for unsupported or sensitive queries
- 🔹 Sector-specific tone and language control (e.g. academic vs corporate vs civic)
Built prototypes in Figma and tested with real content across university, council, and commercial sites.
Created a scalable governance UI for content scoping, tone configuration, and usage analytics.
— REFLECTIONS
🔹 One-size-fits-all doesn’t work: tone, content scope, and UX needed tailoring by sector.
🔹Clients who invested in content curation early saw better results—good AI still needs good content.