The Next Challenge Isn’t Explaining AI Decisions. It’s Designing AI’s Role in the Experience.
The first of a 3-part article series on how AI is redefining user experience in life sciences from the UXLS Community of Experts at the Pistoia Alliance.
Stijn Janssen, Strategic Design Leader, Novo Nordisk
This article is written in a personal capacity. The thinking here grew out of conversations in the UXLS Community of Experts and reflects the author’s own view rather than a formal position of their employer or the Pistoia Alliance.
AI is increasingly invisible. It “hides in plain sight,” appearing as a feature on SharePoint sites, in workspaces, and in the corners of digital applications, framed by organizations as a coworker or assistant. It is there, waiting to collaborate and help us with our Jobs To Be Done (JTBD). But do we really understand why it is there, or whether it is actually helping us become more productive and efficient?
It’s one of the blind spots we have when it comes to Enterprise AI implementation and adoption. We embrace new technology by allowing everyone to build and use tools that make our lives easier. At the same time, we are adding more complexity to our digital ecosystem, infrastructure, and user journeys as they become increasingly tailored and personalized. A general way of explaining the purpose, behavior and intended use becomes a big challenge because there is no transparency or consistency. Two roles can use the same agent for different purposes and therefore need different explanations.
Contradictions create blind spots
As we democratize AI across the organization, everyone has the power to create their own assistant to support their way of working. At the same time we are mining processes to understand how they are followed as described in the standards and identify deviations and bottlenecks so we can decide what AI solutions we can build to help us work more efficiently.
We now have grassroots-built agents that are used by individuals (e.g., a CRA builds a solution to help generate monitoring reports, a medical writer uses a tool for Q&A documents) and centrally developed solutions we want (user) groups to work with without understanding if there is overlap and if the solution is understood the same way or explained in different ways. The result is a fragmented landscape of people using their own AI solutions, centrally offered solutions or none of the above. That does not eliminate friction; it merely relocates it to the interaction between human expectations and AI autonomy.
This illustration shows how key roles in life sciences are using multiple agents serving multiple roles/personas with different skills, workflows, and purposes.
Explaining is designing
There is a fundamental shift we need to make to create consistent experiences as we are moving to a world where agents are not only positioned as our coworkers but act as our colleagues because they participate autonomously in experiences, much like humans do. It means we need to focus not only on how we instruct AI, but also on how we define and design the behaviors, expectations, roles, and responsibilities it should have.
We need to answer the question: Which jobs belong to humans, and which belong to AI? As AI owns part of the journey, what are its goals and boundaries? Designers have always been good at uncovering these questions by mapping out journeys or service blueprints yet these maps rarely include AI as an actor, so designers should claim their position in designing for this explainability by defining the behavior, principles and guidelines that set the standards for working with AI.
The future of experience design is not just human-centered; it’s about orchestrating understanding. Aspen Snow takes up this theme later in this series in Finding Our Place in the Age of AI: The Future UX Researcher, calling for participation and collaboration between humans and AI in ways that are consistent, trustworthy, ethical, and intentional. Explaining UX and AX (Agent Experience: orchestrating the technical setup and infrastructure) separately is not going to cut it. We should instead think about a new discipline: Actor Experience, the practice of designing how human and AI actors collaborate, interact, and share responsibility to achieve consistent outcomes.
As Peter Hummel will argue in the article that closes this series, Start with the Workflow, Not the AI: “Scientists need to understand what the AI is doing, where its suggestions come from, when to rely on it, and when to challenge it.” It is no longer sufficient to explain how AI works. We need to explain what role AI plays within the experience.
The digital evolution
When I explain my role in digital transformation, I always start by saying it is not a transformation; it is an evolution and therefore, we need to look ahead to keep up with AI developments and prepare for the next shifts in focus:
- focus on capabilities
- then on workflows
- and the next focus will be on relationships.
Not simply the relationship between people and technology, but the relationship between people and AI actors within the same ecosystem.
Explainability will always matter (as described in this NN/g article) because every participant in the journey needs a shared understanding of what the AI is there to do, what decisions it owns, where its authority ends and how it contributes to the outcome.
But perhaps the future isn’t about better explanations; it’s about giving AI a clear role in the experience.
Continue the conversation in October
Stijn is speaking at the 2026 UX for Life Sciences Conference, 20–21 October, hosted by GSK in Stevenage, UK. His session, “Implementing a Digital Operating Model in R&D: Connecting Strategy, Portfolio, and Evidence Across the Lifecycle,” is where this article goes operational: designing how human and AI actors share responsibility in an experience, builds on the operating model that makes that coordination real across R&D.
Two days, four themed sessions and four hands-on workshops, with practitioners from Roche, Elsevier, GSK, AstraZeneca, EMBL-EBI, Harvard Medical School, Novo Nordisk, Novartis, MSD and others. This is a working conference, not a lecture series.
Want a first dive into the topic?
This article grew out of a UXLS panel discussion in June titled How AI Is Redefining User Experience in Life Sciences. Stijn was joined by UX leads from Novartis, Merck/MSD and AstraZeneca. The full recording is free to watch here.
And if you want something you can put to work this week, the UXLS Community published free, practical resources:
- Setting up a UXLS Community of Practice: What makes a CoP succeed or fail? How to start small and build momentum? A proposal template to help articulate value internally.
- Quick-start Learning Resource for AI: A curated selection of materials to help UX professionals explore the intersection of user experience and artificial intelligence.
- The UXLS Maturity Model: A tabular guide to assess the integration and impact of your UX strategy.
About the UXLS Community
User Experience for Life Sciences (UXLS) is a community of experts at the Pistoia Alliance. We bring together UX designers, researchers and leaders from pharma, technology vendors and academia to raise the standard of user experience across life-science R&D through best-practice guides, toolkits, and open discussion of what works.
We meet regularly, and because the Pistoia Alliance provides a pre-competitive framework for these conversations, members can compare notes across organizations without the usual hurdles.
If you work on the tools scientists use, you are welcome. Find out on the UXLS Community webpage.
Your turn
In your organization, when someone spins up their own AI agent, who decides what that agent is allowed to own, and does anyone else know it exists?
Be specific if you can. Do you have real examples of overlapping or invisible agents? This is the kind of input from which this community works and may shape the workshop discussions in October.
The rest of the series
This is Part 1 of three articles from the UXLS Community on how AI is redefining user experience in life sciences. Two more follow. In Part 2 next week, Aspen Snow (Merck Sharp & Dohme LLC), explores how AI has shifted the value of UX researchers from producing understanding to orchestrating it. In two weeks, Peter Hummel (Novartis Biomedical Research) shares in Part 3 how AI opportunities are found by mapping workflow friction, not by looking for places to apply AI.