Three Keys To Building Effective AI-Driven Products
Artificial intelligence, while already at work in many applications, remains a mystery and even a buzz term for many working in technology. AI will only continue to evolve, and its use will increase. For companies developing products that will utilize AI, it is critical to have clear objectives.
In clinical research at my company, we use AI in our mission of making trials more efficient while increasing access to study participants globally. Through our work, we have gained practical experience when it comes to developing and augmenting a software platform to utilize AI, including allowing our systems to self-educate based on the enormous amounts of study data that passes through them. The following are some thoughts for companies seeking to develop new AI-driven products.
Use AI to improve the user experience.
Obviously, with any application with a user interface, the user experience is extremely important. If a consumer app is challenging to use, it will likely be ignored or quickly abandoned. In the clinical research space, however, it isn’t as easy. Our software applications are designed solely to gather valuable clinical data. We help customers run decentralized clinical trials (DCTs) at my company — studies that allow remote patient data collection.
This means that the users of the applications we deploy to collect data are the study participants themselves or their caregivers. If our users find our applications too challenging to use, their engagement typically drops, and they often leave the study altogether.
AI can help reduce the software-related friction experienced by patients. For example, having the patient constantly log in to enter information can be a burden for users who are tired and unwell. It can make the study seem like too much work, discouraging patients from continuing. AI-driven automation can help fill out many kinds of assessments based on previous data and behaviors. The AI can anticipate and predict patient entries, allowing the software to suggest to the user that they can confirm with one easy button press versus taking time to type out lengthy assessments.
The second user-specific benefit is that AI can help keep patients and caregivers in the loop about the study and how they are doing. For the vast majority of clinical trial participants, their involvement helps to benefit future patients more than it benefits them, sort of like planting a seed that will provide shade for future generations.
Many patients may grow weary, feeling like they give much and receive little when participating in a clinical trial. AI can help provide patients with visibility into their participation, giving them information that they can share with their other healthcare providers to improve their quality of life.
Don’t limit how much the AI can learn.
Many currently utilizing AI algorithms apply them to static data sets to increase efficiencies. This is great, and it makes sense. However, if you are doing this, you are only scratching the surface of what these algorithms are capable of. To allow your AI algorithms to keep evolving, consider introducing "black swans" — potential but unpredictable events — into the algorithms. It isn’t dissimilar to vaccination; by introducing black swans, you can strengthen algorithms and provide them with the data they need to solve future problems.
Many companies are not yet taking a long view regarding the potential of AI to impact their businesses and don't have a foundational architecture to begin optimizing their data's value. To move forward, companies can look at building centralized data lakes capable of holding massive amounts of data and then building an architectural layer on top of that. Then you can configure a way for your AI algorithms to interact with and learn from your data.
Seek tools and processes that increase the value of your experts.
You cannot overstate the value of automation in clinical research. In a typical DCT, there can be thousands of data points gathered through electronic clinical outcome assessments (eCOA), sensors, medical devices, video visits and more. In an industry where many trials are still literally being run on paper, the ability of an AI application to auto-populate forms and assessments and instantly share information between multiple, disparate software platforms can save hundreds of hours of labor while eliminating opportunities for human error.
In clinical research, and specifically in DCTs, there are enormous amounts of data that need to be aggregated. Now that you have participants not going to a site where, for example, the clinical team would capture measurements like blood pressure, glucose levels, weight, etc., at that moment in time, DCTs allow you to collect data multiple times per week or even per day.
Unlike other industries that are much further along in terms of dealing with such vast banks of data, the clinical research space is tackling this now. Luckily, we have access to AI and machine learning tools to help manage and mine value from all of our data, and this allows us to drastically reduce the amount of time and money needed to run a clinical trial.
Implementing these technologies helps free up hours usually taken up by sifting through and attempting to analyze data. This allows study team members, particularly clinicians, to focus on other vital aspects of clinical research. They can spend more time with patients, offering them support and answering their questions and concerns. This one-on-one time is essential and can be the difference between study success and failure. Further, the insights gleaned from the AI tools can provide new and unexpected insights for clinicians that they can use to help make patients’ experiences more positive.
While our experiences with AI are taken from applying them to help our customers optimize their clinical trials, the takeaways apply to those in any technology industry. Using AI to improve user experiences, effectively use data to evolve and glean new beneficial insights and automate all applicable processes are essential considerations for any organization seeking to create new solutions capable of meeting present and future challenges.
This article was originally published by Forbes
Scott Pearson
Chief Product Officer