Exciting New Worlds in AI Technology Design

Famed designer, Charles Eames said "Design is the art of constraints." My teacher, Michael Large, at Sheridan College, phrased this somewhat differently as "design loves constraints." Both are true, and like all things in design, the details matter. When a designer is faced with a problem, the job is less difficult with just the right level of restrictions. Too many and your design can be handcuffed. The opposite of this is the 'blank canvas.' There's no existing design, or product. It's uncharted wilderness, and like any wilderness, that means it's easy to get lost.

This is what we faced with the Athos Omics AI Platform. There was no product. There was no design system. To make things even more challenging, the whole realm of Omics AI software is greenfield.

Usually, the solution to a blank canvas is to see what others have done before. Researching the competition is a standard first step in product design. Part of this is to understand user expectations, after all, some of your users may have used those products. Part of this is to not 're-invent the wheel.' As Steve Jobs said, "great artists steal." Of course, he 'stole' that line from Pablo Picasso, who stole it from T. S. Eliot. Are there better artists? Not many. The goal is always to design the best product, not to be original for originality's sake. That being said, if you plan to be original, you first need to know what others are doing, so you can do it differently, and hopefully better. Our research told us that there were no products quite like the one we wanted to build. There was, of course, some overlap, but often those products were inaccessible, even as screenshots. You can't steal an idea if you can't find one worth taking.

What we wanted to build was simply too new and too innovative to copy what others were doing.

In that case, you become utterly reliant on three things—user research, best practices, and your own experience.

For Athos Omics AI, this meant spending quality time with biologists, chemists, and bioinformaticians. We looked at how they worked together, and how they didn't. We discovered that often there were surprising knowledge gaps between them. This meant challenges when finding a design language and basic things like product labeling, but also opportunities to serve a bridge to collaboration features. It meant finding out what caused the greatest pain points and frustrations. Time delays and inflexible analysis reports were number one, followed by intimidating technicalities. This meant we needed to make processing fast and easy. We need to make analysis like having your very own bioinformatician at your beck and call.

Above all, it meant remembering what users really want. They don't want to upload and process files, they want analysis. Processing RNA Seq Data files is just the means, not the end. Uploading and processing is fraught with technical complexities, command-line tools, and endless opportunities for error: it is, for most users, a necessary evil. It is expensive in terms of time, effort, resources, and money. For us, this meant making it easy, fast, reliable, and secure was the first, best way we could serve their needs. It also meant making a complete end-to-end solution.

Building a product that does everything is hard, but it makes my work easier, which is to make the users' job easier. Once you eliminate moving big data from one system to another, you reduce the number of tasks and chances for things to go wrong. You also provide one consistent user-experience for the entire process. Learning curves drop, productivity curves rise. If this was easy to do, everyone would do it, but the pay-offs are enormous.

For analysis, we knew what we wanted. We wanted what scientists wanted, familiar and powerful reporting with the added dimension of interaction. Biologists, chemists and other scientists have developed a common visual language with bioinformaticians for clear, effective, and accurate analysis. In this respect, we didn't want to change anything. Users first get exactly what they want and need. Traditional R generated reports, however, are static. Athos T includes fully interactive reports, that can be easily configured and re-configured. They do this in expected ways, but also in exciting new ways I'm not yet able to divulge.

AI is where things get the most exciting. As always, Artificial Intelligence should present information in the most human way possible. In user experience, we have something called the 'user model' and the 'machine model'. The user model is what the user expects and understands of the task they wish to accomplish. The machine model is what the technology requires to function. It's UX's job to serve as the (literal) interface between both. We always push the machine to do more work, so the user can do less. AI makes this easier than ever, but still requires the designer to know both. For a general AI model, like ChatGPT, this means a minimal interface that assumes nothing. The user, after all, can be anyone. For us, it does not. We know our users and what interests them. We can therefore prompt their prompts, and target responses to help them ask the right questions and, more importantly, get the right answers. Sometimes, without even asking. We tackled the 'blank canvas' so our users don't have to.

Charles Eames also said, "Recognizing the need is the primary condition for design."

Which brings us back to where we started—complaining about a blank canvas. Except, of course, as we see, it's not really blank at all. It was always painted with the needs of our users to discover new drugs, develop new strains of blight-resistant crops, create non-irritating cosmetics, make more shelf-stable food, brew better beverages, and more. There's no truly blank canvases in product design, when you know what to look for. Once you do, you learn that this is an incredible opportunity to deliver the boldest design, and to reveal to your users the picture they need to see to do great things themselves.

This is what we have done here at Athos, with much more to come in future blog posts. I'll walk through in more detail what this means and some of the specific challenges of building a platform like Athos Omics AI in the future.