Consumerizing AI to Drive Stickiness and Usability

Share this:

Just a few years ago, employees accustomed to the intuitive interfaces of Gmail and Facebook wondered why enterprise software couldn’t be as simple and accessible. Gartner dubbed this “the experience gap.” It’s the difference between the consumer-friendly software used at home and the interfaces business users navigate at work.

These weren’t just stylistic preferences, either. Soon, organizations investing billions in enterprise software realized the obvious: that easier-to-use technology was not only more scalable internally, but that it delivered better ROI. Accessible platforms could be optimized faster and were “stickier” across teams. This gave way to the consumerization movement in IT and enterprise.

As we head into 2019, the enterprise’s consumerization is well established. Yet when it comes to AI, which will see over $235 billion in investment by 2025, this idea of consumer-like UI has largely fallen by the wayside.

That has to change.

Keep in mind, the value of AI across an organization has exploded. AI tools are now more widespread, with everyone from marketing to HR benefiting from access. Yet, too often, only data scientists can use and make sense of AI tools. This has created an enormous cost burden on the organizations that invest, requiring high-priced data science hires to manage AI stacks and unpack findings. At the same time, there is a data scientist shortage, with global demand for expertise now exceeding supply by more than 50%. This model is unsustainable.

To manage cost and navigate the talent gap while taking advantage of data, AI solutions must be retooled. They need to be user-friendly, without deep knowledge of coding or engineering as prerequisites for use.

But how can we consumerize interfaces for AI and democratize the technology?

What makes an interface truly intuitive is hard to define. Like the cliché about obscenity, we know it when we see it. For instance, if you unbox a piece of hardware and don’t have to consult the manual, that’s intuitive. The hallmarks of intuitive design are that everything is arranged in a simple, logical fashion and it all works as we’d expect. When a design aligns with our mental model, we don’t need to expend energy filling in the gaps. A classic example: in 2012, the nonprofit group One Laptop Per Child dropped off tablet computers in Ethiopian villages. Within four minutes, one kid had unboxed it and powered it up. Five days later, they were using 47 apps per child, per day. 

AI technology needs to be built similarly. Here are four ways we can get there:

  1. Accessibility. Consumers can access Spotify, the music streaming app, from their desktop, tablet, or phone. For users, the experience is seamless. If you’re listening to a song on your desktop and open Spotify on your mobile, the app asks if you want to listen on the phone instead, and it shuts off desktop access. AI can and should be similarly accessible. Apps should be mobile-enabled with full fidelity. They should be available through a simple link or application button that is usable anywhere, including office, home or in transit, as well as accessible across diverse services and networks. This increases and simplifies collaboration opportunities as AI becomes more widespread among teams.
  2. Personalization. Imagine how frustrating it would be if every time you turned on Netflix you had to go back and find your current show and episode. Instead of forcing consumers to waste their time like this, Netflix has a “recently watched” prompt where viewers can drop in where they were before to continue binge-watching The Office. Netflix uses AI to suggest other programs and movies as well. In a similar way, an AI solution should be able to subtly assist the user by employing machine learning to learn users’ preferences. Then, assistance can manifest through recommendations or automation — just like Netflix. These “smart” capabilities can streamline onerous workflows and free up time for more important tasks.
  3. VisualizationLego sets can be complex, but the means to construct them are broken down into simple steps that are easily visualized. This means that Lego doesn’t have to offer multiple translations in every manual and that kids building such systems can backtrack if they realize they made a mistake. But it isn’t just kids who appreciate visualization; some 65% of the population are visual learners. Designers of AI systems should aim to make them as easy as possible to use by making them as visual as possible. This can mean more image-driven instructions over text, drag-and-drop options, and more. But visual UIs also need to be built intuitively — without clutter and endless metadata fields.
  4. Core features onlyFeature creep has always been a problem in IT, with too many software makers trying to address every nit through ancillary features and add-ons. Over time, non-critical features can jeopardize the user experience, making it slower, more complicated and even unstable. While well intentioned, feature creep can kill platform retention — and we see it happen far too often in AI solutions. The most intuitive and user-friendly tools stick to their core features and keep their scope focused.

As the benefits of AI become more obvious, organizations need to ensure that everyone can use such systems. That means closing the experience gap and making AI as easy to use as any well-designed consumer product.

On a practical level, that means giving design a seat at the table. That’s not necessarily the intuitive approach to designing AI systems, but it is the intelligent approach.

Natwar Mall is CEO of