Mark Hurrell.

Voice interfaces are an interesting design problem graphic designers interested in structuring information. They entirely succeed or fail due to the quality of the information architecture underpinning them
A running joke in the team was how mansplaining voice interfaces tend to be. This is due to how differently structured verbal communication is from written words
Because information written for speech tends to be more concise and easier to read, so is also suitable for reusing in things like search engine snippets and pullouts

GOV.UK voice

Product design direction for Government Digital Service

2018  2018

More and more people use voice assistants. That’s especially true for people with access needs, where voice assistants are starting to replace more traditional accessibility tools like screen readers. At the same time, GOV.UK needed an interim launch to start demonstrating the value of the replatforming work, and there was a wider lull in GDS’s product announcements.

Voice struck me as a useful area to experiment with. “Innovation” is a sensitive topic in government — usually because it’s associated with snake oil sales pitches. For GOV.UK, though, making information available for users with access needs or limited digital skills is a core part of the proposition. And sometimes, big organisations need a good, fast-moving story. Traditionally, one of the problems with tools like voice assistants (and machine-learning, and SEO before that) is that they only really have access to commercial data and information — but barely any accurate public service information. Finding a pizza restaurant is easy. An abortion clinic, less so.

I did a bunch of desk research, market analysis and digging into GOV.UK’s new capabilities. I figured out that we could distribute public service information directly to personal assistants using APIs and the semantic web. I realised we could make the content and metadata APIs from GOV.UK step by step and replatforming work for voice. Pairing that with schema.org, we could distribute existing public information and data in a way that voice assistant software could easily consume.

We set up a small agile team to work on GOV.UK voice as a 20% project. To establish the user need, we began talking to providers of voice assistants like Amazon and Apple. (We also needed to talk to them about using our information.) Then we did some prototyping and experimenting. And it worked really well.

Public information from GOV.UK is now available as default on most voice platforms in the UK. This was done without duplicated publishing, and without users needing to install any new software. Other public sector organisations have also started using this approach to making their information available for voice. There’s a blog post about it here, and you can read more about the GOV.UK replatforming work.


Small agile team led by Sam Dub, Tijmen Brommet and myself.

Selected projects