Designing for Google's AI products
AI assistants, voice agents and research tools: interfaces that show how the AI reached its answer, taken from concept to executive pitch to build.
Making an AI's answers something people can check.
The challenge
Generative AI can sound sure when it is wrong. In travel, security, healthcare and customer service, that is not good enough. People need to see where an answer came from, how sure the system is, and when to take over.
My role
Since 2023 I have been a Product Designer in Generative AI at Cognizant, designing enterprise AI products for Google's AI products. I take ideas from first concept through executive pitches to the screens teams ship.
Five areas, one way of thinking.
Different industries, the same questions: what does the AI know, how does it show it, and who is in control?
AI travel assistants
- Designed and shipped AI travel assistants for leading travel platforms, covering trip planning, booking and navigation.
- Presented AI product and technical strategy to senior stakeholders and executive technology teams, and won buy-in for the initiatives.
AI-driven security workflows
- Designed AI security workflows for financial services and healthcare organisations: threat detection, trend analysis and faster response.
- Made AI-assisted decisions, automation and workflow orchestration easier to act on, which cut operational effort and security risk.
Research assistant for scientists
- Designed an advanced research assistant and a literature search and retrieval system for healthcare and life sciences.
- Helped scientists and clinicians run complex queries, synthesise findings and form new interpretations from large bodies of literature.
Voice agents and support tools
- Designed AI voice agents and support systems across automotive, travel and healthcare.
- Automated issue resolution, service scheduling and information retrieval, and supported live customer-service agents with intelligent assistance.
Photo evaluation and enhancement
- Designed an AI photo evaluation and enhancement system for a photography business, to lift image quality and support photographer sales.
- Applied image analysis to connect content quality to commercial results.
Good AI products make the model's limits visible, not hidden.
Show the source
Every answer should point back to where it came from, so experts can verify it in seconds instead of taking it on faith.
Design for being wrong
Uncertainty, errors and edge cases are part of the main flow, not an afterthought on the error screen.
Keep people in control
Automation handles the routine work. Clear handoffs bring in a person when the stakes or the ambiguity are high.
From concept to executive pitch to build.
Frame the problem with the client, sketch how the AI should behave, and decide what people need to see to trust it.
Take the idea to senior stakeholders and executive technology teams with clear product and technical strategy.
Work with engineers until the screens that ship match the intent, including the edge cases.