Helping AstraZeneca design the next generation of enterprise AI.

When I joined AstraZeneca, generative AI was evolving almost weekly. New models were appearing, interaction patterns were changing, and there was no established playbook for designing enterprise AI products.

The programme set out to build an internal AI assistant that could help employees discover knowledge, reports and data spread across a complex global organisation. It wasn't intended to replace people or make decisions on their behalf. Instead, it was designed to become an everyday assistant that helped employees find information faster and work more efficiently.

The challenge wasn't simply building the product. It was introducing AI into a highly regulated environment where trust, privacy and responsible use were just as important as the technology itself.

My role was to help turn an emerging technology into something people could understand, trust and confidently use.

One of the earliest design decisions wasn't visual—it was language.

Internally, we deliberately referred to the product as an assistant rather than an AI tool. At the time, AI was often associated with replacing jobs, creating uncertainty across organisations. Positioning it as an assistant helped reinforce that it existed to support people, not replace them.

Designing trust became just as important as designing the interactions themselves.

Users needed clear guidance about what the assistant could do, where its limitations were and when human judgement was still essential. We introduced warnings around sensitive information, encouraged users to validate responses and designed interactions that reinforced the assistant as a starting point rather than a source of absolute truth.

Helping people feel comfortable using AI became one of the biggest design challenges of the programme.

I joined several months into the programme, taking over responsibility for design at a point where delivery was already underway. The engineering team was moving quickly and stakeholder expectations continued to grow as interest in AI accelerated.

As the sole designer across a large multidisciplinary team, my role extended far beyond interface design. I led discovery, service design, interaction design, conversation design, workshops, stakeholder engagement and product strategy, while ensuring design work remained aligned with engineering delivery.

A significant part of the role involved creating clarity. AI capabilities were evolving rapidly, stakeholder priorities regularly shifted and delivery plans needed constant adjustment. Prototypes became an important communication tool, helping teams understand complex ideas, align on direction and make informed decisions before development began.

Rather than trying to keep pace with changing technology, the goal was to create a product flexible enough to evolve alongside it.

The assistant became a new way for employees to access knowledge across AstraZeneca.

Rather than simply returning search results, it helped employees discover reports, documents and information through natural conversations that felt familiar while remaining appropriate for an enterprise environment.

The work included:

• Conversation and interaction design.
• Prompt and response patterns.
• Information architecture.
• Service design across multiple user journeys.
• Prototypes for testing and stakeholder engagement.
• Design systems and reusable interaction patterns.
• Roadmaps aligning design with engineering delivery.

Throughout the programme, the experience continued to evolve as AI itself matured. New interaction patterns emerged, user expectations changed and the product adapted alongside both.