anthony galvin

Recently I’ve been doing lots of work with voice interfaces. Globally AKQA has been working with clients on VI since it started to go mainstream, and in the UK we’ve helped Jamie Oliver and Arsenal get onto the Amazon Echo platform.

VI is also a lens to use when looking at system or service and figuring out what an MVP really is. By it’s nature VI is a paired back user experience - so if you have an experience that works well for voice users then you probably have a good idea of your core use cases. 

From a technology perspective it also starts to place some new demands on engineers. It forces technical teams to think differently about infrastructure, code and server less at scale. The separation between code and infrastructure can get messy when working with AWS Lambda (and similar services), so it’s extra important to focus on code and build pipelines for anything but the most trivial use cases. 

Developing for voice also highlights something I think is increasingly important for engineers - the ability to bring service design thinking into software / sprint planning. There’s an increasing overlap between writing software that doesn’t have dead ends and handles exceptions well, and managing the user through options and occasions frustrations of voice interfaces.

Test automation, unit and integration tests are even more important than usual (though I’m not sure how much more important you can make these kind of tests!). Manual testing of voice services is slow (you can’t really speed it up), so you need to have good test coverage to have high confidence in your software and for the team to be able to work at a decent velocity.

It’s also important for everyone on the team to spend time using voice services and to understand the emerging design patterns and expectations of regular users. In the early days of television it was apparently common to meet people who appeared on TV shows but refused to own a television set. The same approach won’t work if you’re developing VI. Using an Echo or Google home on a regular basis is the only real way to understand the possibilities and frustrations of voice interfaces.