On 15 November 2017, FARM Digital hosted an evening for charity professionals to learn from Microsoft, Age UK and Cancer Research UK about platforms and prototyping of artificial intelligence to better support service users.
My background in leveraging personal data to deliver service improvements through personalisation drew me to this fascinating event.
Microsoftâ€™s platforms for building artificial intelligence solutions
Faris Haddad, Cloud Solution Architect and â€œData liberatorâ€ at Microsoft gave an overview of the platform that Microsoft has developed (including Cortana Intelligence Suite and Azure Machine Learning) to enable organisations to test out and implement artificial intelligence for better access to relevant information, for fundraising and for better support through enhanced Interactive Voice Response (IVR). Of particular interest was the speech API called Language Understanding Intelligence Service (LUIS) and Microsoft Payment API available within the bots framework, taking away some of the challenges of PCI compliance.
Faris shared an interesting nugget from Gartner: “By 2020, the customer will manage 85% of the relationship with an enterprise without interacting with a human.” It’s interesting how customers will not only be used to interacting with a non-human by 2020 but will likely come to expect a high level of speed and accuracy for most operational tasks.
Age UK’s approach to quicker access to information via chatbots
The digital team at Age UK have been working hard to deliver important information and useful resources to older people and their carers, as demonstrated through their recently relaunched super accessible website ageuk.org.uk
We heard from Rob Mansfield (@robram), Head of Digital Content at Age UK, about how they set up a virtual assistant on their website for their service users to easily access answers to more general questions. This freed up their call centre staff to address the more complex queries, bringing about cost savings for the organisation and user journey improvements for older people.
Key learnings about chatbots shared by Rob from Age UK:
- While most people donâ€™t understand what a chatbot is, itâ€™s OK – internally help them see the time and cost saving benefits, and for service users, let them experience a better service through quicker responses and improved signposting.
- Bots give the impression of simplicity, but it will take effort and ongoing enhancements to deliver a truly great user experience.
- You can never answer everything – some queries will just be too complex, or unrelated – in those situationsâ€¦
- Build a safety net to allow your chatbot user to get through to a real person – whether itâ€™s by phone, filling in a contact form, or patching them through to a live online assistant.
- Itâ€™s important to dedicate time and resource to the planning, implementation and ongoing maintenance (content and technical) of the chatbot.
- The easiest thing you can do is think linear and utilise a decision tree to lead them down one of a pre-defined set of paths to help you help them.
- When interacting with a non-human, it neednâ€™t be a sterile environment. Add some delight where appropriate, using emojis, GIFs, even a video.
- Itâ€™s good to make your chatbot sound human, but not too human! The aim is not to fool your users, but simply to deliver a better service to them.
- Real chatbots are still quite some way off. It can be really complicated to work out how to handle misspelling and mistyping and to really understand the context of what someone has typed. As technology and AI APIs develop, weâ€™re getting closer.
- Know what success looks like by gauging happiness of the end user through some kind of feedback mechanism, such as a poll at the end of the chatbot interaction.
Cancer Research UK’s use of a voice-activated virtual assistant
The Innovation team at Cancer Research UK (CRUK) identify and test out ways to bring about efficiencies and a richer user experience through the use of artificial intelligence, such as for external customer service, internal customer service (IT helpdesk) and fundraising. They use product-focused start-up approaches to build organisational readiness.
Rob Leyland (@robliteration), Innovation Manager at Cancer Research UK, discussed some of the approaches their charity has taken to implement artificial intelligence, using it to provide health information to the public and to influence behaviour change.
It is scientifically proven that 40% of cancers are preventable. Knowing that overconsumption of alcohol can lead to many types of cancer, CRUK prototyped an Alexa â€˜skillâ€™ that allowed people to interact with their Amazon Echo to set goals for the maximum units of alcohol they were going to drink in a given period and to record their actual drinks consumed over that time, just by speaking into the device, with Alexa providing ongoing positive encouragement.
Robâ€™s team at CRUK set up a five day design sprint, using the Google Ventures Sprint approach, to build a prototype. This involved:
- Day 1: Map – mapping out the problem and picking an important place to focus
- Day 2: Sketch – working in isolation and sketching competing solutions on paper
- Day 3: Decide – making difficult decisions and turning ideas into a testable hypothesis
- Day 4: Prototype – hammering out a high-fidelity prototype
- Day 5: Test – testing out the prototype with real live humans!
Jake Knapp, formerly of Google Ventures, explains the process well in this video:
Key learnings about prototyping an Alexa skill, shared by Rob at Cancer Research UK:
- Keep the scope down and start simple. Make it quick to build by minimising how much content youâ€™re initially offering.
- When using voice, be concise – so that the person using it doesnâ€™t switch off 🙂
- Challenge to overcome: there might be confusion about invocation (how to initiate Alexa) and intents (e.g. add a drink to my log, not my shopping basket!)
- Soft launch it to the skills store and then take it on the road to test with real users (e.g. within charity shops)
- Acknowledge (make peace with) the limitations – test it, soft launch it, trial it, enhance it.
Current thoughts about using artificial intelligence in charities
I sense that while the use of artificial intelligence is starting to emerge in charities and the applications appear to be useful for service users, AI will really shine when the data that the charities and the platforms have about the supporter or service user is clean and is used mindfully. Watch this space!
- Microsoftâ€™s resources for building artificial intelligent services
- Rob Mansfield’s full set of slides are available at bit.ly/OverherdAgeUK
- Cancer Research UK’s science blog about their ‘My Alcohol Tracker’ Amazon Alexa skill
- Google Ventures’ Design Sprint
- Sprint (the book), by Jake Knapp
- Future events from Overherd, specialising in digital technology related events for charities.