Cris Cristina // February 26 2018

Customer Experience in the Age of AI

Businesses today are shifting from “tech-led” to “experience-led” product and service development. In 2017 alone, Facebook, Google, and Amazon have increased the size of their design teams by an average of 65%*. It’s easy to see why the Customer Experience (CX) is becoming more and more of a business priority. The latest edition of the Design Value Index, which tracks and compares stock performance, shows that companies which integrated design as part of their business strategy collectively outperformed the S&P 500 baseline by 211%**. So if good design is good business, how does that fit in with the latest trend in development — machine learning and artificial intelligence as a whole?

Customer experience in the age of ai

Apply AI-focused design techniques.

Machine learning is of rapidly growing interest, and the applications are exciting and seemingly infinite. The opportunity is vast, and the challenge is not in what we can do, but how we do it. The smart application of machine learning is critical in translating to good CX. Applying a design lens to AI projects highlights a few focus areas to build great products and services.

  • Design for humans — Machine learning algorithms introduce more uncertainty in the human interaction, making it much harder to script the outcomes. When integrating AI components, it is more important to proactively represent user concerns so users don’t lose their voice in the system.
  • Design for communication and collaboration — Addressing how the human agents and autonomous agents communicate is a design challenge in and of itself. When there are conflicts of interest, what determines priorities? How are decisions and intent communicated? It is critical to map handoff, authority, and clarity of actions.
  • Design for transparency — The implementation of any machine learning project can be inherently opaque. Showing what data the algorithm uses and demonstrating how the algorithm uses those data can go a long way to enabling trust between the users and the AI implementation.
  • Design for inclusivity — We unintentionally build our own implicit biases right in to every product, and that can have an enormous impact on the CX. In AI applications, we need to proactively plan for these biases by both 1)including a broader range of voices contributing to the algorithm creation and application and 2)monitoring outputs to make sure the CX is within the range of planned outcomes.

The one big takeaway here is to make sure we craft the application of algorithms with the user in mind. In short: use design tools to enable relevant, useful experiences.


Double down on the design process.

Developers talk about “human-in-the-loop” machine learning algorithms, whereby humans help train the model (also referred to as semi-supervised learning). While it’s debatable whether it will be necessary to include human feedback over the long term, humans will remain critical in the creative application of these algorithms. Think of a product or service that does not have a human in the loop. It is impossible. How about a building’s fire sprinkler system? On its face, it is totally automatic… except for the variety of people who design, install, monitor, respond, maintain it, and whose lives are saved when the alarm triggers. Maybe most of those jobs can be eventually replaced by robots, but the point is that there will always be a human element. To create a new thing, there will always be someone who asks, “what if we built a water distribution system in the ceiling that automatically douses a fire if the temperature of the room gets too hot?”

How do we get really good at coming up with those questions? Double down on design. Research the needs, goals, and desires of users. Try to understand their behavior, culture, and the context of the environment. Look to see if there are any established design patterns. Sketch. Draw storyboards. Service design techniques are particularly useful as a precursor in mapping how and where algorithms can help. Design can help make machine learning applications more meaningful, engaging, and useful.

Good design drives better customer experiences. As design-driven companies’ performances show, it can make them more profitable as well.

Cris is a product strategist, designer, and researcher at Digitalist. If you liked this article, check out a more in depth exploration of design and machine learning in The Designer’s Guide to Machine Learning.

*Maeda, Design in Tech Report 2017

**DMI, Design Value Index 2015

More from Cris Cristina