Brands, do your best not to fall into the hype.
You should certainly consider a voice assistant extension to a service, but we’re not yet in voice’s heyday. Creating a robust, personified version of your brand (or specific products) that can answer every user question, convert and increase sales, and bring the online/in-store experience into the home effectively is a fool’s errand.
Instead, focus on ways to augment existing services by identifying critical “intents” (the mechanism that picks up what your user is requesting, e.g. “I want pizza”) and on-brand responses for select platforms. Talk to stakeholders, talk to users, identify which platforms your customers are likely to use, and create a low-level voice experience to gain feedback.
First, Identify If There’s a Need
Let’s begin with an example. Let’s say you’re considering building a voice experience for a professional sports team in a major metro area. The first thing to identify is, do the fans of this team need a voice assistant? Considering the rise in home sports viewing and the demographic overlap between smart speaker owners and sports fans, it’s a good fit.
Second, choose your platform.
For most marketers, this advice may seem counterintuitive: be on one platform, not all of them. Just as developing for web platforms and producing for social platforms are converging, so too will voice. In fact, it’s happening. With SiriKit you can be on hundreds of millions of iPhones and HomePod. With Google Assistant – Android, Google Home, and Lenovo smart speakers. And with Alexa multi-modal, Echo Dots, Shows, Spots, and FireHD TVs. Creating voice experiences that cross platforms is becoming easier, but you’re still restricted to one company’s ecosystem (e.g. Apple world, the Google-verse, etc).
But for now, the development ecosystems are fragmented. Services like Snips, Watson Assistant, PullString, and DialogFlow are changing that, but it’s all still emerging. Investing in cross-platform development isn’t worth the investment if you’re building your first bot/assistant.
Choose a platform based on two considerations: where is your audience and what’s your business objective?
- If you’re looking to innovate and create a splash for a campaign, go with something open-source or more flexible like Snips.
“Keep control of your data, users and branding! The more you rely on a third party like Alexa or Google Assistant, the more likely you are to be intermediated and eventually lose your customer touchpoint. Opt for white-labelled solutions that you can control.” -Rand Hindi, CEO, Snips.ai
- If you’re building something that’s a home utility, informational, or a delivery application, go with Alexa or Google.
- If you’re looking to augment your mobile application, develop for Google Assistant, Cortana, Bixby (Samsung’s voice assistant), or Siri using iOS12’s SiriKit (which, sidebar, is a huge upgrade from iOS11 where Siri Development was only available for messaging, payment, ride-sharing, and internet calling apps).
For our professional sports team example, let’s run with Amazon Alexa multi-modal. We know the team’s market overlaps with higher rates of Echo Show and Dot ownership, and we can use their visual devices like the Echo Show and Echo Spot to show game highlights, previews/recaps, stats, and boxscores on larger screens.
Third, map out possible intents that service both business and fan needs.
What’s the score of the current game? How many goals does [player.name] have? Where can I buy tickets/merchandise? The goal here is to categorize larger themes in types of intents (e.g. team information, FAQs, scheduling, etc), identify intents for each category, and then list potential responses in your desired brand voice and tone.
The fourth consideration is slot values.
This is what voice platforms commonly categorize as interchangeable variables intended to train the bot to understand and identify variations of similar requests. For example, you can make an intent “How are the [team.name] doing this season?” Instead of listing every variation of the team’s name – including city, not including city, nicknames, etc – the assistant is able to understand every variation after being trained on a set of [team.name] values. Identifying slot values early enables you to get an idea for the taxonomy breakdown for your voice experience, which can influence development resource allocation, training sets, intent categorization and more.
Admittedly, this is an obvious example to choose. For the Washington Capitals, we built the first Alexa Skill for a professional sports team. The above steps are major pieces of our voice process; building out for one platform and tracking popular intents will influence future iterations of the product.
Voice is ever-changing (check out our predictions for 2018). We can make inferences about the future of this technology, which is sure to continue becoming integrated into our daily routines but only for the immediate future. In the interim, follow the steps above to assess if you need a voice assistant. And if you ultimately build one, check any and all available analytics to see what your customers are and are not using (e.g. Amazon’s got a solid dashboard, see below). For 2019 and beyond, anything could happen.
- Voice Assistants in 2018: 5 Emerging Trends
- Four Pro Tips for First-Time Alexa Skill Developers
- We Published Three Alexa Skills. Only 2 Poke Fun at Trump.
- Team-Based Alexa Development: Heroku + Flask-Ask
As always, if you have questions or comments drop me a line: email@example.com.