Smart Voice Assistants in Retail Banking. Lessons from deploying a Retail Banking Voice App
In our recent thought leadership paper, Journey 2022, we discuss the questions being raised by consumers and society regarding certain technologies and how these may lead to, what we describe as, Digital Dilemmas. Voice assistants are an example of a key opportunity banks have today to reinvent the customer experience in order to stay relevant. When turning to Voice Apps, what are the key success factors banks should look to in order to be successful in the face of digital dilemmas?
The smart assistants we have in our homes are now entering the business world. We’ve been working on introducing one of the first retail banking voice applications to the UK market. In this blog I will share our lessons from this achievement.
Firstly, why build a Voice App?
LMGTFY “ownership of voice activated speakers”. Ownership of voice activated speakers has seen a 100%+ growth in less than a year and now more than 10% of people in the UK own one. At £30 a pop, expect growth this Christmas. That’s the numbers.
But let’s think about this. Our most natural form of communication is through speech, fact? Our larynx evolved for voice. Our thumbs evolved for, well, certainly not smartphone usage.
Voice interaction is simple. Consider. Whilst I’m serving up the Sunday roast dinner, my wife, with her impeccable timing, says “have you transferred that money to your brother?”. I’m now faced with a dilemma. Do I wash my goose fatted fingers to apply my thumb print to my mobile phone and suffer the starved family wrath. Or do I keep serving dinner and forget, again. I know what I’d like to do. I’d like to say ‘Alexa (or Hey Google), transfer £20 to Richard for the birthday I forgot’.
What we learnt
1. Designing User Experience for voice is significantly different to designing for web or mobile devices
Speaking with a voice activated speaker is communication in its purest form. When you interact with a mobile or web device, you’re forced to interact in the language the device understands. Variants in language and the interpretation of intent is a key element of user design that has limited, if any, consideration in mobile app builds, but is of critical consideration in voice app design.. Holding context of what has happened and what a customer might be referring to is an additional feature of the user design for voice. Some of our key points of design learning are as follows:
- Be clear on the use case. Why is this voice app being built?
- Create conversation flows but then use process maps to understand the details and flush out complications
- Keep it simple. For every element that is added you need to consider error and silence handling as well as process and systems checks
- Understand the device landscape and if you are going to need a visual UI too (e.g. for multi-factor authentication)
- Use a simple prototyping tool when designing. It is quick to add text into an online tool and then hear it.
- When developing, load utterances in sooner rather than later. When you start testing beyond dev teams, testers get frustrated because they don’t know the “happy path” words that should be used.
2. Work out early your test strategy
How do you intend to test? What about the data you will use and the test environments? How will you represent the use cases and customer journeys?
3. Hardware and intent needs consideration
As users perform differently, so does hardware. There are a number of parameters within the hardware that define how a device hears. This is to distinguish between “two”, “too” and “to” or “four” and “for”, as an example.
4. The value of delivering fast and joint teams: Use Agile!
Release early and learn, or spend years hypothesising perfection. I would argue that agile is a mandatory delivery method as the underpinning technology enables it.
1. “Familiarity breeds customer satisfaction”
I’ve googled it, can’t find it, therefore I can lay claim to coining the phrase. Let me explain. You go to your website bank login page. Button expected top right. You make an application. Progressing through the form you expect button on the right to progress, button on the left to go back. If these “familiarity standards” are not met, it impacts CSAT.
In voice applications, those standards have not yet been established. Early movers will find themselves either setting the standards or quickly knowing how to pivot to meet the new standards. There’s a first mover advantage.
2. Experimentation, insight and analytics
A/B or Multi-variant testing is a common principle employed by many with an online presence. One that we don’t see much of across voice channels. Experimentation will certainly be a route to learning what works and what doesn’t.
3. Cracking the case for compliance
Registration, (end point and multi-factor) authentication and logging and redaction of data are challenges I can see in moving to fully serve customers through voice assisted channels.
4. Device agnosticism
I would expect to see the customer and user experience and, de facto the underpinning technology, to harmonise on anything that uses voice as the means of customer service.
5. Treating Customers Fairly
For many retail banking customers, an app is their preferred channel of choice, available 24-7. However, more than two million people in the UK live with sight loss and that is predicted to rise to 2.7 million by 2030. Treating customers fairly and in particular, a digitally underserved segment of the population, is a clear use case for voice apps on voice activated speakers.
6. Open banking and aggregation of life services
I’ll save this one for another day 😊
 LMGTFY, a term that requires googling
 Atos Analysis; Smart speaker ownership doubled in six months, YouGov; Tech Tracker Q1 2018, Ipsos Connect
 How many people in the UK have sight loss, RNIB