Product Talk: 3 Principles to Consider When Building AI for Behaviour Change


Product Talk: 3 Principles to Consider When Building AI for Behaviour Change


The world of AI is full of unknowns. That’s why we’re committed to bringing our #PeopleofAI to the forefront. We’ll illuminate what they do, how they stay on top of ever-evolving trends and research, and demonstrate why it matters for smart, human business. This week, Product Analyst, Ben Tang, will talk about various topics in the realm of AI product creation. Here, Ben offers insights for AI enthusiasts, product managers, analysts and data scientists focused on championing innovation within their organizations via artificial intelligence. 

1. Make your user’s goals easy to achieve — but don’t make every decision for them

Too often, product creators get caught up in the idea of making things “seamless” and “simple.” But that’s not always the best approach – especially if you’re creating something for highly qualified, cerebral people like data scientists and analysts, who are, across the board, interested in working smarter, not harder. That means removing tedious work when necessary, and providing a framework  wherein choices need to be made in the right place.

It’s our job to make goals easy to achieve, but the path to there sometimes means a user should be forced to make crucial decisions. By creating opportunities for users to make conscientious choices, we’re allowing them to be in the driver’s seat: you can have machine learning pipeline with default values that exist for less experienced users, but more sophisticated ML experts can change them to meet their unique needs.

In short, we can use technology to encourage users to be highly conscious of the choices they are making. By taking this approach, we can ensure the tools we build enhance and empower the user to efficiently design high performing machine learning pipelines.

2. Know that people get attached to doing things a certain way, and that’s OK! Evolution doesn’t happen overnight, nor should it

Anyone in AI knows about the fear that machines could “take over” jobs. Unsurprisingly, I’m not in that camp, and our organization strongly believes that AI, and AI-powered tools have the potential to enhance your workforce in ways that we’re only beginning to understand.

But that doesn’t mean adoption will be — or should be — simple and seamless. In fact, the sooner we realize that people really love the tech they already use, and the sooner we understand that fusing those functionalities with ease-of-use, the better off we’ll be when building out our AI roadmaps. Data Scientists are not traditionally programmers, and generally coding is only a means to an end. They are generally interested in data problems, not software problems, and have gotten used to using certain tools to achieve their goals. This means that to fit into their existing workflow, a productivity enhancing product should build on top of these current tools. They use various tools to manipulate and model data,  and communicate their findings.

That means the adoption of an AI-powered tool has two hurdles to overcome: first, assurance that it isn’t built to replace a human’s work, and second, that it is better than what they’ve come to know and love, for a multitude of reasons. So how can we ensure data scientists feel empowered to do their craft, but find innovative ways to free their time from mundane tasks, and pave the way for them to experiment more, test more, and evolve more?

This is part of building for behaviour change that is one of the most challenging, but one of the most exciting. Finding ways to ease fears and drive excitement with a single function of a product is what fuels my fire, day after day.

3. Remember that building product is about building a culture of who (and what) drives success in an organization

For some, success may mean communicating results to key stakeholders in an effective manner. For others, it might be all about driving efficiency for the business. No matter how you define success, it’s always important to remember that models themselves have no inherent value: how AI-driven organizations leverage models to generate business decisions is the key success factor.

Building AI that is directly tied to key performance indicators throughout an organization, is building AI that is directly tied to the success of individual team members. Knowing the goals of the user, whether it’s to spend less time writing boilerplate code so they have more time to experiment with new machine learning techniques, or implementing deep learning pipelines to get unparalleled insights on images, product creators should always keep a lens on how success is defined, taking into consideration the unique nuances for individual users and organizations as a whole.

Product creators should always keep a lens on how success is defined, taking into consideration the unique nuances for individual users and the organization as a whole.

This site is registered on as a development site.