Over the last six months, the Voice Lab has explored the challenges and opportunities the emergence of voice-enabled smart devices presents for news publishers. We approached this problem through experimentation and hands-on learning to inform the Guardian’s strategy team around voice and the future of audio.
The project was designed to be limited in scope and this week we are winding down active work on the Lab. Before we close this chapter, here are a few of our key findings.
A distinctive new medium
Smart speakers are a catalyst for a disruptive transformation of how we engage with machines. But the current generation of devices represents an entry point rather than a terminus.
The technology powering this revolution provides natural language access to a range of services enabled through artificial intelligence. From streaming on-demand media to bespoke new interactive models, designing for voice requires novel approaches to building meaningful content experiences.
Assembling teams from different disciplines, who are ready to rethink and relearn how to approach problem-solving, will be vital. It’s been an exciting and rare opportunity to be at the nexus of the content creation process and product development methodology.
Audio-centric, not audio-only
Access to digital assistants is proliferating to an ever-increasing variety of form factors with different surface capabilities. With device categories ranging from headphones to automobiles, building for the future of voice will require a flexible approach.
Undoubtedly, the most basic response to any voice input is to respond with audio. Crafting high-quality sound has the potential to be a differentiator in an area lacking many traditional visual elements. While audio might be the default, it’s not the only output possible.
In the same way responsive design influenced the process of building for the web, creating solutions for voice requires a similar set of principles. Investing in structured and flexible content will pay dividends when adapting to new surface capabilities and rapidly evolving user expectations.
The prospect of creating seamless multimodal interactions across devices used to feel like the realm of science fiction. Through smart-assistant platforms, using both visual and aural methods as inputs and outputs is becoming an exciting reality.
Emergent content patterns
Habits on smart speakers in particular are converging around two distinct types of experiences; short and transactional vs long and ambient.
On one end of the spectrum, demand for succinct solutions to everyday needs is clear. Users are keen to get news summaries, weather updates, or even order a taxi on demand. The end of the interaction is to complete a straightforward request by providing something concrete and finite. Our Guardian Briefing project was an attempt to fit this mould.
Conversely, long music sessions and radio streams fill the background with indefinite sound as a companion to typical household activities. Easy to request and simple to control, these linear experiences provide a different type of value. More forward-thinking formats will take advantage of the interoperability of audio on smart devices to add layers of control and personalisation.
Notably, podcasts haven’t yet found a significant audience on this platform. Perhaps they live in the middle ground between short and long form, or maybe they’re challenged by some other structural problem.
One theory is podcasting requires too much upfront specific knowledge to request compared to the relative ease of asking for a broadcast stream or music radio experience. The current model is based on aggregating feeds on a mobile device, choosing from a number of recent episodes a la carte, then listening during fixed increments in the day, such as commuting or during household chores. Many of us may have started to ask for a podcast from the Assistant before realising halfway through the command we lacked the exact vocabulary to get what we want.
Many people are working on new formats and discovery mechanisms for podcasts to help address these problems, but until then smart speakers are likely to account for only a modest amount of listening.
Horizons for search
Voice has the potential to become a dominant avenue for search. Predictions assert by 2020 half of all searches will be done through dictation, and 30% of all queries will be performed without a screen. Asking general knowledge questions is already consistently ranked as one of the core use-cases.
The nature of the medium poses challenges for primarily page-view-based business models. News media and platforms need to work together to better understand and quantify this behaviour. Explaining our fast-paced world through answers based on reporting is one of the core values news publishers provide. Especially for content creators whose main medium is text, exploring formats like speakables in combination with structured content systems could pay dividends as more of our audience relies on voice.
More than just hype
It’s likely we have yet to reach “peak hype” around voice. While smart devices continue to sell faster than mobile phones 10 years ago, it’s also likely this trend will plateau at some point, reaching a saturation point similar to what we’ve seen in other device categories. But equally it’s difficult for us to foresee a future where we speak to machines less.
Publishers should avoid myopic strategies on voice. Instead of taking a one-size-fits-all approach, we encourage media outlets to grapple with the existential questions around voice and how this new medium will affect them.
If you’re starting a new project in voice, feel free to dig into our work. We’ve open sourced the code from our two big projects, the Guardian Briefing and Year in Review. Hopefully these can be helpful to get you up and running.
Exploring this frontier has been challenging and exciting. We can’t wait to see what comes next.
