Jade Software Thu, Dec 5, '19 14 min read

We built the world’s first digital barista. But why?

Bella is a conversational AI-enabled digital assistant, who can talk to you about most things coffee, and obviously make you a coffee of your choice. To pay homage to Simon Sinek, we’re going to start with why and then move onto how and what.

So why did we build a AI, IoT, and chatbot-enabled coffee machine that interacts with you like a real human?

Speed. Innovating at speed is an essential ingredient when competing in today’s fast-paced business environment. Responding to opportunities or market forces within days, weeks, or months could be the difference between leading and lagging the pack. So we set ourselves two weeks to pull this project off, showing what’s possible when you focus your energy.

Chatbots. Has there ever been a word that evokes such polarising views? There are several reasons why chatbots failed to live up to the lofty expectations that burdened their robotic shoulders. With Bella, we’re demonstrating what’s now possible through conversation and how different intelligent chatbots (what we refer to as digital employees) are to their simpleton ancestors: powered by AI, the ability to learn, and carry out certain tasks rather than just responding to basic questions.

Customer centricity. Bella, with all her coffee knowledge and technical skills, bridges the gap between technology and experience – you get what you want through conversation. In your business, you could easily replace the coffee machine with a number of digital or physical technologies and apply the same customer-centric, conversational approach to the problems your business is facing.


Here’s how our story of speed, digital employees, and customer-centricity unfolded:

The timeline: How exactly did we bring Bella to life?

  • Day one: we assembled our A-Team because we knew that innovation couldn’t happen in silos. Plus, it was clear that our deadline was going to stretch our capacity, so we brought in expertise – Ara Institute (Chris Wordsworth).

  • Day two: we bought the coffee machine through one of our contacts at Noel Leeming, then pulled it apart, then let Chris get all technical – starting to reverse-engineer the coffee machine.

  • Day three: let the UX research begin… We immersed ourselves in the world of coffee (not that we weren’t already familiar with it nor needed an excuse to drink more). We talked coffee. We tasted coffee. We observed coffee. Using our re-useable cups, we drank plenty of takeaways, and we discovered some interesting takeaways (insights) too.

  • Day four-six: Normally we all love public holidays and long weekends. But when they put your project under even more pressure, perhaps not. It was down tools for three days.

  • Day seven: in breaking down the typical conversation flow of ordering a coffee, we found there are five bits of critical information required to make the right coffee (assuming there’s only one option for the type of bean). And this is what we call intent matching, which we facilitate on Ambit’s conversational AI platform. From these five intents, our UX design team got to work in designing the conversation flows. Here’s an outline of the intent matching the uncovered:
    • You want to coffee (duh)
    • Size = Regular
    • Type of coffee = Flat White
    • Type of milk = Trim (or Skim)
    • You also want one serving of sugar

  • Day eight: we wanted to bring Bella to life (as humanly possible), so we reached out to our partner UneeQ, and they came to the party with one of their digital humans! This meant we could put a face to the name, letting people visually interact with Bella, rather than having a conversation with a coffee machine (because that would be weird, right?).

Learnings from (just over) the halfway point:

It’s all about experience – people love catching up with their barista, background music and all, as much as they do the coffee. So while people really crave coffee, there’s a craving of relationship they’re seeking to fulfil at the same time.

There are no "smart" coffee machines out there. None. "Connected" does not equal "smart" (and few are even connected).

Things we thought should be easy are really not; things that we thought would be hard turned out to be a breeze.

The need for speed (innovating under deadlines) requires the use of platforms and partners – you can’t innovate in silos.

It’s surprising how many people are willing to jump on board and help when there’s an exciting project at stake - Ambit and UneeQ, Ara Institute of Canterbury, Jade – a great team.

Even simple orders like a large flat white still require a bit of back and forth. Only 15% of the conversation was around coffee.

  • Day nine: we finished reverse-engineering the coffee machine, which put up a worthy resistance, but Chris and some of our developers were up to the challenge. There were a lot of ones and zeros, zeros and ones, plus something that resembled a flux capacitor too. The physical tech was in place, but now we needed to make the software to talk to it.

  • Day twelve: the weekend happened again, but no matter. Bringing together software (the conversational AI and other digital smarts) and the hardware (the DeLonghi coffee machine and other tech Chris built to make our connected tech smart) was proving to be harder than first thought. More coding, more development, more sweat and tears, and more tweaking to the WiFi tech we had to install. There was progress, but it was (dial-up) slow.

  • Day thirteen: because around 85% of the conversation is small talk, we consulted other well-known baristas around Christchurch and others on social media channels to pick up on the intricacies of all things caffeine. We then added these into Bella’s arsenal so she could drop them in conversation while making coffees. But would it all work?

  • Day fourteen: one of the biggest sighs of relief was witnessed in a workplace near the fringe of New Zealand’s largest cities. Bella was alive. Alive! The project team featuring human baristas, technologists, multiple businesses, developers, UX designers, coffee runners, and more pulled off creating what’s possibly the world’s first digital barista. And she was on display for all the world to see at the 2019 Canterbury Tech Summit.

Learnings from the end of the project:

Great conversational AI doesn’t just happen; it’s not just designed either – it’s trained. So make sure you give your digital employee the training it requires to be able to offer the ideal experience. If you’re not sure what that ideal experience should be, partner with someone who can help you discover what that should be.

A connected coffee machine allows you to trigger it via Bluetooth. A smart coffee machine can talk to you about your coffee (and of course the weather).

Get the right people on board early on in the project, to ensure you maintain momentum. Don’t be afraid to look outside your organisation too, as partnering can be a great way to increase your productivity and level of output.

Once you launch the product, it won’t be perfect to begin with, but take regular steps to attempt to achieve it. Expect that you might initially have ‘the wrong type of beans’ or don’t ‘offer organic lite almond milk’. But don’t expect your customers to accept that will never change. Keep improving your offering, so your customers keep getting delighted.

Amazing things can happen when you take bets (however large or small) on innovative projects. You meet and often work with amazing people too, as projects like this seem to have magnetic properties.

Technical matters: What’s under the hood?

If you’re wondering what exactly is under the hood of Bella, here’s a brief outline. Bella interacts with people through a contact centre (through Amazon Connect) or with a digital human (from UneeQ). She has been taught to speak using a conversational AI platform (from Ambit). Bella’s choice of coffee machine is an automatic, connected Delonghi. The Delonghi was turned into a smart coffee machine with the help of Chris (from Ara Institute) and APIs created by developers (from Jade). The overall experience was concepted and crafted by a UX design and technology team (also from Jade).

Coffee up: now what?

We do bring Bella up in our conversations with people, and we even sometimes get her out of the box and put her to work – including a return trip to Australia for a conference where she entered the country in someone’s baggage – no passport required.

But Bella was never really about the coffee; she was about the people, the conversation, the experience. Through this, you’ll hopefully have seen it’s entirely possible to chat with a ‘machine’ and get something done while having a good conversation and a good experience. It could be something simple like adding a meeting to your calendar. Or it could be something with greater complexity and regulation such as financial or insurance advice. Whatever it may be, taking a partner-centric approach can be the difference between just landing the project and landing the project that gets meaningful results.

Talk to us about making data meaningful