Sounds like the future: voice tech
Part 2: how to get started
This follows on from part one of our Sounds like the future mini-series. To read about the basics of voice tech, start there. Here, in part two, Louise Corden focuses on how Alexa can be used for charities and how to get started.
At NSPCC, we’ve been looking at voice technology for a while as part of our remit to make our services available to our supporters via the channels they actually use, and broaden accessibility of our products and services. We launched our NSPCC Alexa skill in December. It helps deliver our core purpose: keeping children safe, through teaching them that their bodies belong to them, via a PANTS song and interactive voice game. And because we couldn’t deliver services without money, we’re also accepting donations through Alexa — becoming one of the first charities in the UK to do so.
Check it out on Alexa or watch our video of it in action:
We’ve also worked with our corporate partner O2 to create two skills that help parents understand the online world and help keep children safe. These are: NetAware — adding content from our NetAware website — and Parents vs Kids — an interactive game voiced by Richard Osman and Lauren Layfield to see whether parents or kids know more about the online world. We presented these skills at the CharityComms Digital Conference last November and had some great feedback — so we know there’s interest across the sector.
Voice technology offers huge potential for charities to deliver your services or collect donations, but how do you get started?
How to create your first Alexa skill:
1. Work out what problem you’re trying to solve. This will work best if it’s something that naturally lends itself to voice — e.g. do you need your hands free or is it something you can do together as a family. If you can co-create this with your intended audience: even better!
2. The Blue Peter bit: get a piece of paper and sketch it out. Start with the ‘happy path’ i.e. what will happen in an ideal world. Then consider error paths and how to respond if Alexa doesn’t understand the question or the user asks something you didn’t include.
Keep your responses for Alexa really short and never give people more than a few options to choose from (three max).
3. The Wizard of Oz bit (aka prototype). Write the questions a person might ask and Alexa’s responses. Get two people to say each part out loud — you can make this more sophisticated by recording Alexa’s responses or programming them in a Blueprint (Amazon’s free tool) and playing them on demand. You’ll soon realise the bits that sound waaaaay too long, too formal or just aren’t how people actually talk.
4. Build it. As technology projects go Alexa skills are relatively straightforward and there’s lots of documentation online, but if you don’t have anyone in house who knows a coding language, like Node JS, you’ll need an expert. We worked with the clever folks at Addition who are now offering a template skill for charities.
For simple skills without the donation element I would have recommended Storyline — a free service letting you build Alexa skills without coding — but they’ve recently changed their product. Voice tech is a rapidly evolving space so there will likely be other companies launching a similar service soon.
5. TEST TEST TEST and refine. Ideally test with your main audience, but at a push use anyone — the more varied the better! The way I as a 30-something Londoner would ask a question is likely to be different from a busy Cornish grandmother or teenage Liverpudlian. If you’re using Alexa to take advantage of its accessibility for blind or partially sighted people be sure to include them in your testing.
Things to watch out for:
- All utterances said to Alexa get logged in whichever Alexa App was used to set it up in the first place, including a transcription plus the voice recording. So a child asking Alexa a question on a shared home device will be logged on the adult’s app.
- Charities working with vulnerable groups need to be aware of utterances being logged, since it might mean Alexa skills are not appropriate for their service delivery. We decided not to offer anything that verged on advice or support because we didn’t want people to disclose sensitive information or lead them to believe (falsely) that it’s a two-way support channel.
- It takes time! If you’re looking to receive donations the setup process can take a while. Assume three months. Amazon are having a phased release of donation functionality for charities so contact them to check if you’re eligible.
- Once created your skill will need to be submitted to Amazon for certification before it can be made live. This review process can take a while — assume at least 2 weeks, with one or more rounds of changes within that.
- It’s all new! Amazon only enabled donations for charities in the last few months so it will be interesting to see how people adopt it
- It’s in flux! New = exciting, but also that things change quickly. For example, at the time we first started creating our NSPCC skill, there were rules about how audio files could be treated within Alexa. We chose the option that offered max. 90 seconds per clip, needing to do some hasty editing to cut the PANTS song down from 2mins. Thought part way through the development process Amazon changed the rules so you can now have audio clips up to 4mins!
Have a go creating your own via free tools including Amazon Blueprints — this free service lets you create your own Alexa skill with some easy to use templates. You can’t publish these for the public to access, but you’ll be able to access them via devices connected to your account. So as a way of prototyping your new skill to test out with some users, or to create an example FAQ skill to engage sceptical stakeholders in the office they’re great.
Voice design is different from designing for other channels. On a website you have layout, images, colours, CTAs and buttons all working together. If one element isn’t quite right it might look a bit odd and possibly affect your conversion rates, but overall impact is unlikely to be huge. But on Alexa it’s different. Voice is a very pure medium; there are only words, so every word matters.
We learnt this the hard way, in the development of our NetAware skill where our first draft took the content of our online safety site NetAware and put it verbatim in the Alexa skill. We quickly discovered that all the context we carefully supplied on the NetAware site was lost when our first draft of Alexa read “on Facebook there is a high risk of suicide, violence, bullying, suicide and drugs content” — and the end result was misleading (as well as terrifying!).
So we had to rewrite all content for voice delivery.
If you’re accepting donations you will have access to income reports and basic donor data via your Amazon Seller Account.
You can access Alexa skill summary stats from within the Amazon Developer Portal. This shows the total number of people who’ve enabled the skill, number of users, sessions & utterances (i.e. the statements made by users to Alexa) and which intents were activated (intents are the categories of info within the skill, and include: donate; play the Pants song; what is Childline).
For privacy reasons, standard reporting functionality means you can’t access the exact words an individual used, e.g. if our only user on a given day was someone who said, ‘I want to donate because I’ve been abused’ we would only be able to see this info in the reporting:
— Sessions: 1
— Unique customers: 1
— Utterance: 1
— Intent: Donation
Not the exact words they used — in this example, that they’d been abused.
Discoverability is difficult. And today there are nearly 30,000 skills available in the UK Skill Store. So think carefully about how people will find your shiny new voice app; on Alexa the main way is via searching the skill store. If your skill fits with a general campaign you’re running on other channels consider adding a CTA of ‘ask Alexa to open [your skill name]’.
So to wrap up with a single tip:
One of the best ways to get started with voice tech in your organisation is simply to use voice services and judge what works well and what doesn’t.
I know this is an emerging field so if you’ve got any questions about voice tech I’d love to help. Let me know in the comments below.