3 tips for winning with Alexa

3 tips for winning with Alexa

For this blog in what will hopefully become a regular stream of consciousness, I thought I’d talk about the rapidly emerging world of voice assistants, the impact they are having, and how we’re trying to use this as a business. Most importantly, I’ll share some hard earned experience about developing and launching your own Amazon Alexa skill.

The future is voice

A recent survey of more than 5,000 consumers in the US, UK, France and Germany revealed that nearly a quarter of respondents would rather use a voice assistant than a website (a figure that is actually expected to rise to 40% in the next three years). With the Amazon Echo’s Alexa app coming out on top as the most downloaded app for iPhones and Androids on Christmas Day, voice interaction is rapidly becoming consumers’ preferred method of communication in our evermore digitised age.

The real appeal of voice interaction lies in its application as a ‘virtual concierge’ – easy to use, minimal effort, real time answers with the ability to initiate tasks. Although the technology will no doubt become more sophisticated, the numbers of people already using it for fairly simple transactional activity or for entertainment has grown massively in just a short space of time.

By allowing people to ask questions about the actual problems they need fixing (“What can I do if I’ve been made redundant unfairly?”), we can provide information in real time, and set out some next steps to take.

Launching our own ‘legal’ skill is about challenging ourselves to be at the forefront of new developments, helping us to communicate with our customers, and making our legal expenses insurance product more accessible and ever more relevant.

3 tips for building your own voice assistant

Having navigated the process of developing an Alexa skill with our ‘innovation training wheels’ firmly in place, we’ve learnt a few lessons on what works and what doesn’t. Here are my top three tips…

It's not about us, it's you

We’re realistic enough to realise that people are unlikely to ask Alexa questions about legal expenses insurance (as interesting as it is…). For voice interaction to work for us, we needed to think about how our customers are really likely to use it – the problems they are likely to face, and the questions they might ask. We looked at existing customer research we’d conducted, as well as data from our legal helplines and claims MI (Management Information) to help guide this.

One of the challenges of talking about technical areas like the law or insurance is that customers may not always know the right questions to ask to get the information they will need. We therefore spent some time working on a set of questions and answers that would guide you around a subject – including several ‘help’ prompts as well as suggestions built into our answers to “try asking about this”.

Whilst we think we know what will work with customers, the best way to actually know is to learn from how customers use the content you’ve created, and to listen to the feedback they give. If possible, talk to some users about their experience. Using the concept of rapid prototyping to test and learn with real customers seems a scary option as customers can be both complimentary and critical. However, this is really is the only way to get to ‘know’ what customers want and gives confidence that further iterations of the product will be going in a direction customers want – which will save a lot of time and effort later on!

Beware the large batch death spiral

Keeping our target audience in mind, we carefully mapped out the content we wanted to provide as well as several layers of navigation to help customers make the most of that information. Content written, we then programmed, tested, fixed, re-tested, added ‘utterances’ (different ways of asking the same question), re-tested and so on.

The whole process took far longer than we’d wanted, for the simple reason that our ‘minimum viable proposition’ had grown larger and more complex with our desire to deliver the best customer experience – all at a time when we were still learning the nuances of the programming. This complexity greatly increased the number of problems we had to resolve, and made solving them more challenging.

So, long story short, my next tip would be to launch an initial version with the smallest possible amount of content. By doing this, you can understand the development mechanics and launch process before you broaden the content. A small handful of questions would be fine to test out different question formats, session variables and utterances.

Co-creation is best

After you have nailed your initial content, and begin the process of dealing with errors, we found the most productive method was to get our programmer in the same room as our content owners, and to deal with each problem one-by-one; make the fix, upload it and then re-test. Repeat as needed.

Alexa sometimes pronounces words awkwardly – try altering line spacing, punctuation and also the preceding words (which can alter the pronunciation of the following word). We managed to resolve most issues this way, and the ability to quickly iterate through potential solutions for each problem allowed us get things resolved quickly.

Obviously if you’re both the programmer and content owner, you have this one sorted!

We’re just at the start of our journey using voice assistants, and it will be interesting to see how their use develops in the coming months and years. As a company we’ve begun a process of constant innovation, and are looking to accelerate this through the year, so I’ll share any further learnings as the journey continues…