AI vs UI

Over the past 6 months, I have been involved in numerous AI/ ML projects. Each project had a different use case, some were consumer facing, others were internal facing; despite this, they all involved some form of UI. It is interesting because these projects are mostly a proof of technology where we would show that something is doable. However, most of our efforts are actually spent on building an interface for whatever AI solution we developed.

I am not sure why it is surprising for me, maybe because I am a developer by nature. It is well-known that one of the most important skills in data science is story telling and communication. The same thing applies to AI; without being able to convey what value your solution brings to the customer or stakeholder efficiently, what you’ve done goes to waste. This is a very important concept in AI, delivering something usable is as important as the model you build.

That is why, it is important when you are building an AI related solution, you spend some effort on the UI and give the stakeholders a taste of how it will look eventually. It can be a simple bootstrap template, but as long as the UI is attractive and conveys the message clearly, it adds another good point to whatever you are delivering.

I Like to think of AI as end-to-end applications that helps the end user perform a certain task more efficiently and in an easier way.

Now UI is not my forte, but I have learned to develop full-stack applications to compliment the machine learning and data science skills I have. This has actually proven useful as there are instances where I needed to quickly spin up an application to demonstrate a model I developed to non-technical stakeholder. If I did not have that skill (albeit an amateur skill), we would have lost a few opportunities. Even if you are a professional UI designer, it is import to set the expectations of the stakeholders and make sure they understand that this doesn’t fully reflect the end results. Most of the time, the final AI application will look significantly different that preliminary demonstrations you have done.

This also brings about another important point: the user experience. Since AI is really an application a Machine Learning model in a real world scenario, it is important to focus on the value and experience of the user. That is why you see many companies like IBM, Microsoft, and Google who have come up with various AI building components to facilitate the process of developing AI applications. For example, IBM has Watson which enables subject matter experts with little to no development and data science experience to come up with their own models. It then becomes a developer’s role to stitch everything together: a bit of speech to text here, a bit of visual recognition there, with a chat-bot on top and you will end up with a superb AI application that brings value to your customers.

As you can see, AI is very application oriented, where you have to make sure the business case brings value and is very user oriented. Without having the ability to convey the value your AI model brings to the table, you will not be able to bring about this value to your consumers and stakeholders.