We’ve been exploring people, culture and tech here in Australia on our daily blog. If you’re looking for your next dream role with 100+ of Australia’s top companies doing amazing things in tech, check out The Martec.
For many, chatbots mark the beginning of the intelligent machine interaction predicted by futurists and philosophers for decades. Popularity and functionality of chatbots have skyrocketed in recent years with a number of brands fighting to pioneer the technology.
But why are chatbots so enticing for brands? The technology is far from error free, as a customer facing service they open up a big risk of alienating users and they still don’t solve user needs as effectively as humans can.
In 2015 messaging apps overtook social networks; for the first time ever people were using messaging apps more than they were using social networks.
This opportunity sparked Boris Alzamora Sánchez’s interest in the technology when creating his first chatbot “Evento” last year; which went on to win the Wise Tech Global Prize for his team last week. Boris says “we wanted to explore this new technology which was actually showing some interesting results compared to social media” by offering an UNSW events aggregator in the form of a chatbot.
After completing Evento, Boris was approached by S1T2, in collaboration with Bastion, to create the Blackmores Well Bot, which launched at the Australian Open in January. Using the latest AI technology, Well Bot, a personal wellbeing coach, was designed to help you stay on track with your new year’s resolutions; engaging enthusiastically with you about your individual wellbeing plans and providing a tailored program in either nutrition, beauty, mindfulness or fitness to get you on track with your 2017 goals.
Blackmores Well Bot is a hyper-energetic, bundle of bot-energy with a great sense of humour and a disarming personality which lead me to ever so complicity hand over my email address and personal wellbeing preferences to appease the bots excitement in providing me with a wondrous plan for mindfulness. Exactly as it was intended to do.
“People try to design software in different ways” Boris explains “rather than thinking about how the interface will look, you have to think how this persona will be designed. It’s not like trying to design a user interface, it is trying to design a personality… without the persona of the bot it’s just a questionnaire” and without this persona you are not connecting to your users and you’re certainly not capturing their data.
Unlike designing a user interface, the team had to put pen to paper and ask themselves:
What kind of personality would the bot have?
How would the bot interact with people?
What questions would the bot ask?
What was its script?
How would people respond to the bot?
Script in hand, rather than going straight to code, Boris and the team spoke to colleagues around them to see what their reaction was; learning how they would interact with the platform, gathering insights and taking it back to code. They then worked with Blackmores marketing team to connect Well Bot to their CRM, ensuring all preferences given by users were captured and could be used to tailor future marketing activities.
Where to start when building a bot
Blackmores Well Bot was created using Python programming language. Python is dynamic, allowing you to add on features indefinitely. “Python has an API library of almost anything so you can connect to AWS Technologies, IBM Watson or try to apply machine learning on the go” says Boris.
Start with the Facebook Messenger documentation if you already know how to code using Flask or Django (if not learn the basics) then, read the forums about Python programming for facebook messenger. In addition, python tutorials related to machine learning would be useful, but not strictly required. Finally, and most important, try to identify what is beneath and work on creating a persona for your bot.
Research your engines
For Blackmores Well Bot, Boris and the team used API.AI, with great success. Certainly not error free but the team learn as the users interacted and updates on the cognitive model were made accordingly. The service was available on SMS using Twilio, Messenger and Web for the backend; which was introduced as a solution to direct any sensitive topics to human representatives. Yes, most chatbots still have humans as their back-up.
Recently, however, Boris has been testing IBM’s Watson, which has presented some surprising results compared to API.AI and WIT.AI, the reason being its simplicity of use. “The failure rate is really low and it is easy to train”.
The future of bots
Boris believes “eventually, whether it is IBM Watson or someone else, those technologies will leverage a really good level of natural language processing that may have almost negligible error rate.”
Right now many options are being identified for natural language processing using technologies like Alexa or even Siri. We can use this bot and present it in a chat window or we can use the same engine in natural communication channels using speech, we have many ways to connect the same model. This is something that IBM Watson presents now.
For the future Boris predicts “maybe this model will evolve to a point that the failure rate will be close to zero and rather than using interfaces we’ll be using, if not voice, than commands like ‘open window’, ‘close door’, ‘lights’, etc, that will have no error rate and will create conversations with an AI.” Think Mark Zuckerberg’s Jarvis.
This presents immense opportunity for humanity. Imagine not only helping traditional users with luxuries like closing doors or making their toast but assisting users that have problems such as low vision or hearing problems. This technology could provide to them a more connected world; connected solutions that are easily accessible.