Conversational AI: what's the point of chatbots?

What role should conversational AIs play in the future, and how will we interact with them?

Chatbots have become a hot topic recently, with Apple, Facebook, Kik, and Google opening their platforms to developers. These conversational AI programmes are being explored by organisations across a wide range of industries, but what role should they play? What will they look like in the future, and how will we interact with them?

My first experience with a modern chatbot was pretty abysmal. Flush with hype from spring’s keynote speeches, I decided to check out a news chatbot on Facebook Messenger. I spent most of the next half hour trying to get it to stop shouting news at me and unsubscribe. Far from an insightful, machine-learning, digital sidekick; it revolved around recognising keywords and feeding news stories back to me. A week or so down the line it started feeding me news again without prompt and I ended up blocking it.

It would be ridiculous to write chatbots off just yet, of course. It’s early days and naturally they will improve, but my experience with the news chatbot raised a question — what was it doing that couldn’t be done with an app or webpage? Companies have promised that chatbots will make engaging with services more natural and eliminate the inconvenience of switching between apps. This made the bot seem even more odd; although it could offer me a summary of news stories, I still had to leave Messenger to read the full article. What benefit is it offering me over a news agency’s own app or webpage?

Will chatbots replace apps and webpages?

If we ignore the hype and try to understand the mechanics of chatbots, it’s not difficult to imagine useful and practical applications for the technology. For example, using chatbots in a messenger capacity works well when it is asked to return individual pieces of information — asking what the weather will be tomorrow, what my bank balance is, or how much that pair of shoes are. In this context, chatbots could cut the length of time required to complete the task. But using chatbots for tasks that return more than one result can impact negatively on the user experience. If I don’t know which pair of shoes I want, it’s more useful to see multiple options at once on the same screen, as opposed to one at a time in a carousel.

So how can chatbots be designed to accommodate large amounts of results? Do they even need to? Would a more personalised engine that knows exactly which handful of results to show be the way forward?

The monetisation of personal relationships

One thing that stood out to me during this spring’s tech keynotes was the prevalence of chatbots being used in an eCommerce capacity; lots of great demos of how to call an Uber or order food as a group, but few demos demonstrated other possible applications of the technology.

Messenger and chat apps like Whatsapp, Facebook Messenger, Google Hangouts, and iMessage are one of the last few places on the Internet which are largely free from adverts and commercialisation. Although the logic of ‘going where the users are’ is solid, some of the keynotes felt almost like personal relationships between people were the next target for commerce and monetisation. In our opinion, chatbots can be so much more than this.

The opportunity for brand building and customer support with chatbots is huge: they could facilitate the kind of personalised two-way communication that marketers have been chasing for years. The chance for brands to tell their stories and build relationships while delivering services and support through the same outlet could be a game changer. Research has shown that people prefer support options which don’t involve having a discussion with staff, either by phone or email. Being able to distil the qualities of a good customer support agent into a bot could be one of the defining challenges of the chatbot era, and the first companies to get it right will profit from their endeavours.

Public service bots

Another field that could benefit from chatbots is public services. Think of your local council or municipality website. Is it easy to navigate? I would guess that the likely answer is no. A chatbot could be a natural, friendly interface for finding and interacting with complex services, helping to relieve pressure on human staff and allowing them to focus on the more exceptional cases escalated by the bot.

These benefits could also extend into healthcare. To meet the demands of an ageing population health systems are going to need to rely much more on technology and patient self-management, but health is a field in which people truly value a human. A well-designed chatbot could offer the necessary information while remaining reassuring and comforting throughout, with the ability to call for help where necessary.

What are the implications for UX designers?

Chatbots will no doubt begin to play an increasing role in designing the customer experience, but as they are in their infancy the field is wide open for experimentation. This is what makes them so exciting — being able to try out ideas before standards are set and customer expectation begins to form.

Providing the transition from app to chatbot is smooth, it could contribute to an improved user experience. But who should initiate the conversation? While having a bot pop up with a welcome could help people find their way faster, malicious or poorly-designed ones may lend themselves to spamming. The number of users installing adblockers is rising; it’s not unreasonable to assume that we may see the emergence of ‘botblockers’ if chatbots are not designed and implemented responsibly.

Another facet of the UX of chatbots to consider is the design of the bots themselves. Should they have persistent personalities? A single brand voice? Or should they tailor themselves to the user? Personalisation lends itself to better engagement with customers, but it could come at the cost of consistency and control in branding.

Ultimately, the overarching question designers must ask themselves when designing chatbots is whether the bots should be identified as personified entities or simply software. Facebook’s Messenger app is focused on getting people to speak directly with bots, and connecting the bot to the associated personality of its parent brand. Meanwhile, Google sees chatbots more as omnipresent assistants — ready to jump into a conversation and help out, but dismissed to the background again when they have served their purpose.

How to plan for chatbot failure

As more experiments occur around this technology, failure mapping should also be taken into consideration. Websites and apps offer clear paths for the user to follow. With chatbots, users build their own paths through interaction. Many of these interactions will inevitably lead to dead ends, and measures must be put in place so the bot can respond intelligently to these failures. In my experience, the Messenger news bot’s ‘failure to understand’ message was cute at first, but after the sixth time in a row it became frustrating and caused the mask of intelligence to crumble. Bots can also be asked things completely unrelated to their primary function — think of the abuse hurled at Tay, Microsoft’s ill-fated Twitter bot, or the million off-topic questions posed to Siri each day about what it’s wearing and what it thinks about Terminator. Users will expect answers to these questions, and designers have to cater for them.

What’s next?

Chatbots are in many ways a natural commercialisation route for the growing investment in AI tech over the past few years. Similarly, sales-focused bots were always likely to be one of the earliest out the gates, but we believe that chatbots represent a huge opportunity for enhancing the user experience. Cutting edge UX is something that we have been incorporating into our services for years, and chatbots are no less hot a topic around our studios than they are in Silicon Valley. What excites us most are the opportunities for guiding users through a smoother and more natural experience in using digital products, but huge challenges lie ahead and it may be years before we see standards emerge and the first universally appreciated chatbots elevate the field. And who knows? We could be welcoming our first bot to Team ‘Meeja very soon.

Screenmedia is a design and innovation practice and consultancy. Get in touch to find out how we can support you in becoming a more innovative company.


Innovation Lead

Let’s build something amazing together

Let's talk