We sat down with Emma Pegg, head of content at Spixii: the insurtech startup based at Plexal that recently designed, built and launched an award-winning chatbot.
Zara, who’s getting more efficient by the day, is helping Zurich customers file non-emergency home or car insurance claims a whole lot faster.
What does the Zara chatbot do?
Customers can talk to the Zara chatbot to notify Zurich UK Insurance (Zurich) of a non-emergency home or motor claim.
Zara was released into the wild on the same day as the Beast from the East hit the UK. With large parts of the country covered in snow, many customers were facing fallen-in ceilings, burst water pipes, broken boilers and flooding damage. It was up to Zara to help fight the Beast and relieve Zurich’s dedicated customer service teams.
Without her, customers would have had to wait hours for the call centres to open and then join long queues to speak to someone.
With Zara, customers can now log their claim 24/7 on the Zurich website.
What was Spixii’s role in creating Zara?
Zara was designed in one week, then built and implemented within two weeks using Spixii’s chatbot technology, which includes the CXD (Conversation eXperience Designer): a user-friendly tool that helps companies visualise digital customer journeys. The tool also includes an analytics dashboard that provides unprecedented behavioural insights.
Are chatbots like Zara intended to replace humans, or support them?
Call centre staff are often spread thinly and have to answer many of the same questions over and over, which gives them less time and resources to handle other queries.
Zara collects the information online and passes this onto the human handler, so they have everything they need to start the call and make the process much faster. In fact, Zurich can now process claims within three working hours compared to the previous 24 hours.
How did you go about programing Zara’s responses to make this process smooth?
We held a series of vigorous workshops, aimed at uncovering Zurich’s tone of voice and testing how it resonated with people. Zara was actually iterated five times in just one day of user testing in Zurich’s office in Swindon!
After this, Zurich employees put her through her paces to ensure the language was on-brand, clear and engaging.
Now, Spixii supports Zara’s ongoing maintenance and suggests iterations to make sure Zurich’s digital customer experience is the best it can be.
She’s part of Spixii’s mission to make insurance more human and personal.
So by adding more technology, you’re making insurance more human – isn’t there something contradictory about that?
At first glance, it is quite the contradiction!
We want to add the human touch to insurance, but we’re using an automated bot to do that. It looks like you’re taking the human out of the equation completely, right?
The thing is, any technology starts with the people behind it. The chatbot conversation starts with an interaction between at least three people on paper.
While we’re building the “script” and doing user testing, we aim to make sure the conversation reflects what feels natural and intuitive. After all, everyone loves chatting. A chatbot is a scalable solution to helping insurance companies personalise their digital customer experience.
And is machine learning and natural language processing (NLP) advanced enough to be able to support chatbots right now?
Both machine learning and NLP are absolutely able to support chatbots.
That being said, insurance chatbot technology is a different ballgame because you’re working within a highly regulated industry. And until NLP can operate at the same level of accuracy as a human being, it’s difficult to incorporate it within certain use cases.
For instance, if you’re buying insurance, can you imagine how frustrating it would be to talk to a quote-and-buy chatbot using NLP and be sold the wrong product? Humans and machines communicate differently. Any ambiguity in language could cause huge problems further along the way.
However, machine learning is an incredible asset. We’re still a young company, having been incorporated in early 2016, and will apply machine learning with our partners once we have enough data to draw statistically significant conclusions.
Final question: what’s next for Zara the chatbot and Spixii?
Zara will be continually updated and iterated, getting more and more intelligent as she grows.
Her extraordinary net promoter score (NPS) – the percentage of customers wanting to recommend her to others – is currently at 78% and continues to rise.
As she quickly becomes much-loved by the Zurich team, Zurich customers and the industry as a whole (she even picked up an industry award recently), she may even start having new types of conversations with customers.
As for Spixii, we’re growing rapidly and will be opening offices across Europe later this year. If you’d like to collaborate on a chatbot project for your customers, please drop us an email at firstname.lastname@example.org. And who knows, the next time you need to buy insurance or update your policy, you could find yourself talking to Zara.