How has the platform faired? Should medical communications agencies be considering adding chatbots to their marketing mix?
In April 2016, Facebook opened up their Messenger platform to allow developers to build custom ‘bots’ that can (in theory) hold natural language conversations directly with users. The rest of 2016 saw an almost overnight explosion of chatbot startups. As a medical communications agency we felt this needing exploring further.
For a while, the hype was real. The model offered a low-barrier entry point for companies itching to engage customers on the platforms they already use; enticing marketers and developers alike with the chance to have a commercially relevant use case for a foray into ‘artificial intelligence’. The real win for marketers came from the wealth of user information available via the platform, offering unprecedented insight into their user base.
However, the hype died almost as fast as it started; chatbots promised much but poor execution often meant experiences ranged from hilarious to downright frustrating. For your enjoyment, check out https://chatbot.fail/
Despite this, chat-led user interfaces have been confidently making meaningful inroads across a broad range of sectors, and bot-based startups continue to receive plenty of VC funding (check out Botfunded.com)
We’ve been keeping an eye on the best bots in health and social well-being. Here’s a few that have caught our eye:
https://www.your.md/ Making better health decisions
https://www.swell.wtf/ Social decision making
https://chat.itskoko.com/ Emotional “safety net” support for social networks
Swell is undoubtedly the poster child for rapid growth hacking in a multi-platform chatbot world – rather than release a standalone app, the company took their offering to existing platforms that audiences were already engaging with.
However, Koko is my personal favourite here – the platform has composed a wonderfully engaging social & emotional support network with the potential to make real, positive change. The story behind Koko’s development is equally engaging and a great case study in how a (still) small company can make great waves through a well-designed and architected platform.
The big takeaway from the advent of chatbots is that we have a new axis of engagement to contend with. When developing chatbots, UX & UI are no longer just about pixel-perfect layouts, but are first and foremost about curation of conversational user interfaces as an extension of your product, service or campaign. With a conversational UI, we have an opportunity, for brands to – dare we say it – demonstrate some personality.
Adrian Zumbrunnen, Conversational Designer at Google, offers an insight into this:
My website’s design is what some would call very Swiss. It has a lot of whitespace, it’s pretty minimalistic and there are almost no colors in it. In short, it’s not exactly emotional. The chat bot on the other hand sends emojis, laughs and makes jokes. It has emotions. In other words: it has personality.
For those looking to start out on the road to developing their own chatbot, the tooling and set up can seem fragmented and confusing. As a specialist medical communications agency, we’d recommend that anyone starting out take a good look at Wit.ai, the Facebook-owned NLP & intent parsing platform. It has a reasonably intuitive interface for developers and non-techies alike to start “training” a chatbot to being parsing intent user-submitted sentences.
The screenshot above shows an example of a YourMD-style conversational we’ve trained to recognise the intent behind each part of a sentence. You can see how being able to match free-form user inputs to defined intents allows us to programmatically evaluate the response required to the combination of these three intents.
Using tools like Wit.ai has drastically lowered the barrier to entry into this exciting sector. For marketers and dev teams in agencies, there’s nothing to stop you taking that first step towards developing a chatbot for your next project, product or service. Of course, it’s up to you to make sure that you chatbot actually has something to say…