Voice. Six Lessons Beyond the Hype – Part 1

In this two part series Erik Rave (Technology Director)and Jeroen Thissen (Creative Director) from Creative Digital Agency CODE D’AZUR look back on three years of developing voice applications for the likes of KLM and LeasePlan. In this first part, they will touch on their first three learnings: It’s not an ad; Stop talking. Start building; and Think about your architecture.


Image Source: Reviews.com

Voice is not advertising

While everybody is talking about ‘voice is coming!’ and ‘the rise of the smart speaker market in China’, throwing numbers around about Amazon’s investments in the channel and super-fast adoption rates, there’s far less talk about actual learnings from developing voice applications. That’s why, after three years of developing Google Actions for brands such as KLM Royal Dutch Airlines and LeasePlan, we wanted to share six learnings in six short bits of writing. The first learning may be obvious, but it’s a very important one if you’re considering starting to work with the medium: It’s not advertising.

It’s not advertising

Voice is not a medium you can use to push out messages people don’t want to hear. At least not yet (we’re curious to see the first voice activated commercial on Google Home). It’s about creating something people actually want to use, it’s about adding value. There’s no advertising structure within the platform itself. It’s similar to the beginning of the internet when we just had a couple of websites around, without Google. You can advertise your service in other media of course, but only if you have something useful to offer.

It’s a service

So, it’s about being useful. And that might just be the hardest thing to do – ever. What do your customers actually need? We always start with exactly that question.

At the same time, we’d be the first to admit we’ve made mistakes. When we developed Pack Assistant for KLM in early 2017 and that attracted a bit of attention. But we also found out that the service was too long and elaborate. It was overcomplete and people dropped off half way through. Since then we’ve come a long way. KLM is now running four voice services simultaneously in two different languages. The central A.I. backbone is also managing conversations on Messenger, WeChat and WhatsApp.

A voice service is not something you build once to get some attention and show you’re innovative (that ship has sailed). Instead, it is an iterative service design process that requires constant monitoring, maintenance, improvements, development, design and writing of a team of people to do it right. The business potential, however, is huge. So, one would say the investment is worth it.

In the next five parts we will explore more concrete learnings. Why is important to just get started? What are the right ways to set things up? And what can you do to make sure the service fits your brand? Maybe our learnings can help you get started. Because if you move now, you can be up and running as opportunities to serve and market to your audience open up.

Stop talking and start building


Image Source: Appinventiv

When dealing with innovation, there’s a universal truth about the future that is often perceived as a liability but is actually your greatest opportunity: no one really knows what’s coming.

Empirical research does agree on one thing: being the first to have a great new business model can bring huge profits, happier (returning) customers and positive brand perception. But at the same time, it can be difficult to move into the unknown. You pay the price of being first. Will voice technology be profitable any time soon? Perhaps. Will it reach a critical mass of people soon? More likely. Will it become your main channel of service within three years? Could well be. But we’re not sure.

So. How to develop for the unknown?

  1. Be flexible

One thing you do know is that things will change. It is important to have a clear picture of the service you eventually wish to provide. How you get there may change many times. Not just because of your own organisation. Google’s changing and constantly updated API might affect your scope of work. When we were building the ‘Take Me There’ service for KLM (a travel inspiration action) we built a piece of functionality to be able to ask the user for a departure location. Suddenly, a Googleupdate made it possible to read out the device location on approval. We could now determine a location automatically and did a lot of work for nothing.

So stay agile (to use another buzzword). Stay flexible in deliverables (backlog), stay flexible in budget and stay flexible in expectations (KPI’s).

  1. Training

The real optimisation starts when the Action is released. The key word, is training, and this basically means that you get lots of people to use your service in a staging environment and track where DialogFlow misinterprets. When you say you want to book a flight and fly to ‘the Big Apple’ DialogFlow might not recognise this as New York. But you can allocate this point of interest to these words in this context manually. That’s how the system learns. This continues after release of your service and new misinterpretations might arise. Keep training!

  1. Get started

When we know the only thing that really makes your service better is training, we understand the importance of getting started. Make sure you have a clear idea of the user’s needs you are trying to fulfill, define an MVP, create a backlog and get started. Do it.

Think about your architecture

Previously, we focused on the need to stay flexible. Our next point is very much connected to this: Focus on your architecture.

It’s not just your organisation has to stay flexible, but also your voice service itself. The system behind the voice application might be considered even more important than the application itself.

It’s important to get started with your first action (otherwise there will never be an architecture). But, realise at the same time that this action is very likely to be altered in the future or will even fail and get killed. So you’ll have to ask yourself what happens if it does. Do you have a framework set up where you can easily plug in different actions for different platforms? And what happens if your A.I. ecosystem broadens to, let’s say, also your customer service team? Much like your website there’s a front end and a back end (and an API). Don’t forget to think about the long-term ecosystem as well.

You want to ensure you offer a consistent user experience throughout all channels. If your customer tells you something via Google Home it would be great if this information will be available to the human agent in the service center that contacts the customer the next day.

An important strategy we used in our collaboration with KLM was to not build everything on Google’s DialogFlow. This would have been the easiest thing to do, but it would also mean that we would be dependent on its functionalities in controlling our actions. On top of that, all of our data would run through Google. And with the eye on data-ownership and privacy, this isn’t desirable. So instead, we built our own autonomous A.I. ecosystem that sends information to Google when requested. We use our NLP as a service to just us. Google isn’t able to use us as a service. This means that (in theory) we could swap out Google for any other system at any time if we deem it to serve our interests better. This also made it easier for us to ‘plug and play’ new actions and make our services available on Alexa.

We’ve seen that staying flexible is key, both in your organisation as well as your technical architecture. Next, we’ll look at another key learning that fits in the slipstream of this topic: why it is important to focus on the intent of your user rather than the conversation itself.


Erik Rave, Technology Director, CODE D’AZUR

Erik Rave is Technical Director at creative digital agency CODE D’AZUR. In his position, he oversees the development and technology department and ensures the technical realization of show-stopping creative concepts for the likes of KLM, Red Bull, ABN AMRO and LeasePlan.

With a passion for tinkering, Erik has spent close to two decades honing his skills and knowledge all things technical, from coding web platforms and building installations, to developing augmented reality games, artificial intelligence systems and voice technology solutions. When he’s not knee-deep in code or arms full of tech-parts, Erik enjoys hip-hop and techno tunes to fuel his inspiration.

Headshot-Jeroen Thissen-LR

Jeroen Thissen, Creative Director, CODE D’AZUR

Jeroen Thissen is Creative Director at CODE D’AZUR. He creates digital campaigns, services and products that people love for brands such as KLM Royal Dutch Airlines, LeasePlan, Red Bull and ABN AMRO Bank. Recently he’s worked on creating AI ecosystems, voice applications as well as AR games for various brands. His projects have been recognized by The Webby Awards, The Lovie Awards and Cannes Lions.

Why Brands Should Not Approach Voice Like Advertising

Introducing a Maturity Model for Voice & Conversational Marketing