When we were testing a client’s voice app, we noticed an interesting phenomenon.

User spoke: “May I have an orange juice?”
App understood it as: “May I have orange shoes?”

Sergio and I were working out of our separate offices (in Barcelona and Chicago, respectively) and upon seeing the testing output above, we immediately understood what was happening.

Sergio is Italian and I’m Korean American. In our lives, we have experienced people (and recently, digital services) misunderstanding us. We can completely empathize. So, we feel really passionate about making this misunderstanding less painful and awkward for all parties involved.

With the advent of human-to-machine communication, the importance of comprehension is at an all time high. AI-based experiences are becoming more commonplace around the world. It’s not uncommon to see restaurants take advantage of AI-powered menus and drive-throughs. Companies have also been utilizing machines in customer service for decades with interactive voice response (IVR) technology, yet only recently are they applying AI aggressively to make sense of the stream of incoming data.

Now more than ever, it is vitally important to design and deliver an elegant experience for people who interact with machines. It’s important because this interaction is another intimate touchpoint the customer experiences with a given company/brand. Just as someone at a restaurant taking a food order over the phone would never offer “shoes” even if they misheard “juice,” we should expect our machines to act in the same manner.

We at PacteraEdge help design and deliver an elegant experience by taking the following ingredients into consideration when we craft a lovable human-machine interaction:


The role that accents play in a lovable voice experience is often underestimated, even within the same country. One of the unexpected insights in our experience is the realization that everyone has an accent. This revelation should seem obvious to anyone who has ever been to Boston, but even in a location, multiple varieties of accents can be heard.


Besides what people sound like, it’s also important to understand the nuance of how people say the things they say. People from different cultures may not ask for a drink the same way, for example. One person might be more courteous and ask in the following manner: “I’d like a drink, please.” Someone else might be more direct and declare, “One drink, large.”

A machine should also be able to nimbly navigate the conversation and intuitively understand the intent of an order whether the person prefers to say “pop” or “soda.” In both of these examples, the experience with the machine should feel natural and familiar given the human participant’s background.


Referring back to the awkward testing output we mentioned in the beginning of the article, we know that humans will acknowledge that orange shoes do not make sense when ordering food (unless of course, the menu item was called Orange Shoes). Therefore, even if the person heard “shoes,” they would contextually understand the order was for juice — or at least attempt to clarify the misunderstanding.

Just like our everyday conversations with other humans, the context plays a critical role in our understanding of the communication and intent of the message.

As such, using natural language processing plays an indispensable role in an AI-driven experience as it ensures the machine can interpret our intentions and respond accordingly. And as the use of such technology increases, especially in regions of the globe where this experience is brand new, it is of utmost importance to continuously train the core intelligence to be aware and mindful of such considerations.

The training of the intelligence core is hard and laborious work. Just like training in real life, it is often kept hidden from public view, managed by trainers who are expertly skilled at exacting a high level of performance. This continuous education is what we must do so the machine can perform dutifully with elegance and power when it is time to perform.

How are you training your machines?

Mike Kim, Digitalization & Modernization Practice Lead (Chicago)
Sergio Bruccoleri, Solutions Development Lead (Spain)

Michael Kim

Michael Kim

Practice Lead, Digitalization