Conversational Design: An Alexa Skill for Coffee Lovers

All good ideas start with a good brew of coffee.

Personally, I’m a bit of a coffee fiend/nerd (the same can be said for a few members of our team), I love the variety of flavours in coffee and brewing it from the comfort of my own kitchen. The tactile process of grinding, stirring and pouring it in the right way, playing with all the variables to get a great flavour out of a bag of beans and experimenting with different processes to brew the perfect cup of coffee.

The challenge?

Those variables are a double-edged sword though, there is nothing worse than getting an expensive bag of fine Ethiopian Single Origin beans that promise dark chocolate, honey and cherry notes and turning it into burnt dishwater by doing the wrong thing in the brew. I wanted to build upon our previous work in conversational design and see whether we could create an improved experience, in brewing coffee.

The solution?

A guide to brewing, tailored to an individual bag of coffee and brew method, delivered through an Alexa Skill. Following up a brew with a rating flow could also help tweak subsequent brew recipes and even recommend other beans to reflect the user’s own taste. This would give it a lasting benefit to the user and help coffee roasters match their products to customers individual tastes.

To guide users through the process of brewing the perfect cup of coffee with an Aeropress, you have to first open the Skill with the invocation “Alexa, open Brewing Guide”.

1. Select which recipe you’re brewing from a number of options, in this case, it was simplified to options ‘one’ ‘two’ and ‘three’, the Skill will then confirm your selection and give you a description of the coffee you’re brewing.
2. The Skill will then read out everything you need to get started, from putting the kettle on to making sure your Aeropress and coffee mug are at hand, waiting for the next invocation before moving on to the next step.
3. Step-by-step instructions differ between brews, but include information on the optimal boiling-water temperature, how to avoid burning your coffee and the time it takes to brew in your Aeropress before it’s ready to be enjoyed.

Once all steps have been completed, the Skill will finish things off with a friendly goodbye and a reminder that users can customise their recipe’s with the invocation “Alexa, open Brew Guide and rate my brew”.

The complexity of designing for conversation comes down to handling every use case and every permutation of how a user can respond to questions and queries, what I really wanted to get out of my week was a proof of concept. I pared back the scope to a ‘Happy Path’ flow which includes a choice of three coffee blends from our awesome friends over at Naked Coffee, a local coffee shop we share a passion with for empowering independent businesses. Making the assumption that the user is brewing a black coffee with an Aeropress.

Making the Skill sound more human:

One of the main complaints about voice interactions is a cold, robotic feel. To make this something that users are comfortable and at ease interacting with (especially important when your users have just rolled out of bed) I modelled lines of dialogue on how our friendly local barista, Rob, chats to us, focusing on the terms used by Baristas, not customers ordering coffee.

This proved to be extremely valuable in finding alternative or variations of words, as well as the specific terminology used with Aeropress machines, such as:

• ‘Doppio’ is an alternative way of saying ‘double’ in context to espresso shots
• The ‘plunger’ (top cylinder) and ‘chamber’ (bottom cylinder) parts of an Aeropress
• Differences between ‘tones’, ‘flavours’, ‘roasts’ and ‘aromas’ in coffee brews
• A number of different brands and machines associated with brewing coffee

Arguably the biggest challenge that voice assistants and technology like Alexa Skills are facing, at least when it comes to being adopted on a wide scale, is trust. Nowadays people are far more wary of how devices are being used, and in the case of Alexa, being used to gain information and seen as ‘spying’ on people in their own homes.

I want to develop the project by implementing machine learning and artificial intelligence, more specifically, teaching the Alexa Skill to personalise conversations to individual users. This personalisation would make every experience unique, intuitive and personal, allowing me to use conversational design to create an experience that builds trust and meaningful relationships between Alexa and it’s users.

A few ways I could implement this is by cutting out unnecessary steps or questions on a user-by-user basis, use the same terms and phrases as different individuals, even make user-specific recommendations based on their history of interacting with different Alexa Skills.

Ratings, improvement and discovery:

This is a massive aspect of this concept; the loop that allows the Skill to be continuously improved from a user-experience perspective. Helping people achieve better and better brews with an interface that feels more natural and human to them.

I’d see this working as a few follow-up questions after your brew about the overall taste and (if the response was negative) going deeper to ask about the balance of different attributes (sweetness/bitterness/acidity) and modifying subsequent brews to draw out less or more of these depending on the feedback.

Barista training:

To ensure continuity in how coffee is brewed in a team of baristas, the Skill could provide step-by-step guides that help new team members replicate consistent quality across their team. A re-imagining of how the user-flow for his case would give the Skill another facet.

Progressive simplification:

As the Skill got used to brew habits and your knowledge, it could start to remove steps and dialogue to continuously be streamlining itself. This would remove the feel of robotic repetitiveness and allow the user to only hear the steps they need through conversational design.