Product design in AI-drug discovery: how to design products for and with scientific users | BenevolentAI

Do you know what “tool compound”, “target expression” or “IC50” means? If not, that’s ok, when I started Benevolent as a Product Designer, I didn’t know either. That’s why I work with expert users: our drug discoverers.

‍We meet this challenging design brief by calling in our closest allies: our internal users (BenevolentAI’s scientists)‍

At BenevolentAI, I both closely collaborate with Product Managers and Software Engineers and work with different scientific experts, such as Drug Discoverers, Medicinal Chemists, Bioinformaticians, Cheminformaticians. Luckily, these users are also our colleagues, and we closely collaborate with them on a daily basis. They help me understand their thought process, their motivations, and their field of expertise; they are not only users but collaborators. By building this close relationship, I can, without having any scientific background, deliver designs that help them discover new and better drugs for patients in need.

Designing for, and collaborating with, expert users can be challenging. Here’s an outline of my design process to help other designers who find themselves designing products in a complex, collaborative scientific environment‍

The first rule is not to panic, and to take design tasks step by step. That is ok to not know everything from the outset and it is ok, and even encouraged, to ask questions if you’re not sure. Scientists at Benevolent are passionate about their work and field of expertise and can communicate difficult topics in a way that others from outside their domain can understand. After all, no one at the company has all the answers, so we’re used to collaborating on solutions. With this in mind, we can embark on a 3 part process; research, design and consult, and check usability.‍

1. Research

Research in our design team is conducted by the user researcher, designers, and product managers; it depends on the scope and complexity of the project . On some projects, I perform the research, and on others, I use our researcher’s findings.One of my favorite research and empathy-building techniques is observation.

2. Design and consult

My design process consists of design and expert scientist evaluation. When needed, I get feedback from my domain expert colleagues to get a scientist’s perspective before I move on to evaluating my designs from a usability perspective.

‍Usability checks with prototypes‍

The form of the usability check differs depending on the project.

  • If I have time, I might also ask expert users for their feedback on the design. This is an opportunity to get new insights from domain experts who weren’t involved in the project earlier. The more people I ask, the more small details or requirements I can find that are missing. Because in the research phase I observed the users while they work, I can prioritize my findings better.
  • For bigger projects, I hold longer usability check meetings.
  • In case I need more input from a diverse group of users I set up a remote user testing session (using allowing users, in their own time and terms, to participate in usability checks.
  • When using Maze, I don’t have the opportunity to ask them about their expert input face to face, so I use it at the end of the process, or when the nature of the design is more into interaction design and doesn’t require much of the expert input.
  • In the case of complex interactions or hypotheses, I try to use interactive prototypes and check them in users’ normal workflow. For one of the first large projects I had at BenevolentAI, I had to redesign a key flow in Drug Discoverers’ workflow. I had an idea about improving how Drug Discoverers collect their assessment about a gene, and I was interested in finding out how users would respond to this idea. The design wasn’t as complex, but the new flow was different from what users are used to. The quickest and, frankly, the best solution was to use Google Sheets with pre-formatted functional cells which were arranged to simulate new product features. The prototype setup was BenevolentAI’s platform taking up 70% of user’s screens and on the right side, they had the Google Sheet prototype. In this way, users were able to enter real data into the proposed design and I was able to observe how they work with it, all of which cost very little from the prototyping perspective.
Róża Turowska, Product Designer



Uniting human and machine intelligence to discover new ways to treat disease | #Becauseitmatters #AI #DrugDiscovery #Innovation

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store

Uniting human and machine intelligence to discover new ways to treat disease | #Becauseitmatters #AI #DrugDiscovery #Innovation