Return to site

The Chatbot opportunity – especially for healthcare

The chatbot will see you now

· Chatbot,AI,Healthcare

You will have heard about chatbots. There has been a lot of chat about them chatting a lot. The tech titans have been investing heavily in them from Facebook to Google to Microsoft. Brands from Tommy Hilfiger to Pizza Hut have published chatbots. Even Joey from Friends now exists in chatbot format. We are hearing that they are revolutionising the way we interact with the web and that branded apps are finally dead. So what are chatbots? Why would people engage with them? And what could they do for healthcare?

1. So what is a chatbot?

A chatbot is a computer program designed to simulate conversation with human users, especially over the Internet.

OK, sounds a bit techy. What does that look like?

Typically they look like a conversation one might have on a messenger app:

broken image

They could also be hooked up to an avatar of some form to give them a ‘human face’:

broken image

It’s important to realise there are broadly two types of chatbots: one that functions based on a set of rules, and those that use machine learning. In many ways the chatbots that function on a set of rules can be a great place to start, particularly in a heavily regulated sector like healthcare. They are regulator’s dream as they cannot answer any question other than with the statements programmed into them. So they are not capable of breaking regulations. You can’t say that about human representatives. The workaround to the fact they may not be flexible enough for all questions at the outset, is a rigorous process of analysis and iteration.

We are at the start of this world of virtual agents to speak with. With devices like Amazon Echo and Google’s Home, increasingly it is likely virtual personal assistants will become part of our everyday lives. So it’s a good time to be getting involved.

2. So why would people want to engage with a computer programme?

Won’t people want to talk with someone rather than use text to carry on conversation? Well, not really. Look at the rise and rise of messenger apps. They are now more popular than social platforms. From Facebook’s Messenger to WhatsApp, message apps are the preferred way for many to communicate. In any minute, 20.8 million messages are sent on WhatsApp.

broken image

Yes, there will be cases where people want to talk rather than use text, but chatbots can also use voice functions to hear and communicate using language. Interestingly, they can be programmed to use any language at all, including the use of regional dialects and slang. Of course there are limitations – currently – but a secret of successful chatbots is that they know when to hand over to a human, such as a healthcare professional.

Another change in people’s behaviour is their expectation of receiving quick personalized information. People don’t want to wade through unhelpfully worded FAQs. Good chatbots use the language people use. In fact, one of their strengths is that they can constantly learn the language people use to interact with them. This is also gold for organisations to learn more about their audiences. People also don’t want to wait in a phone queue while being told ‘your call is important to us’ and, worse, be told someone will email them back in 24 hours. People expect information any time, any place, anywhere. If they don’t get it instantly, they move on.

And here’s another strength of chatbots. They don’t require an organisation to wait for people to visit their ‘owned channels’ of call numbers, websites, bricks and motor buildings or branded app. Nor do they require organisations to pay for advertising to get them there. Chatbots can live on the channels that people are already using: WhatsApp, Facebook Messenger, Slack, Skype etc. They can seamlessly fit within the flow of people’s lives.

Chatbots are discreet. It has been shown that people are more likely to open up to virtual humans, such as chatbots, on sensitive topics that are still rather taboo in society like mental health, sexually transmitted diseases or even zits. People have a strong tendency to want to look good in front of others, including doctors.

But with chatbots, there is none of that human instinct to hide away situations that may be seen as socially unacceptable.

And note that chatbots don’t need to have the feel of a computer programme. As you will see from the above examples, they can take on the look and feel of a human. Having said that, it’s important to note that the magic of an effective chatbot is not in the technology alone. It is also in the language it is programmed to use. The tone of voice is crucial. But that is the case in any communication. Organisations that forget this in their standard communications do so at the peril of being called ‘faceless’.

It is true though that there is a particular challenge for technology in that it must avoid the ‘uncanny valley’ effect of a computer simulation being similar to a human but not quite the same. Don’t leave chatbots purely to techies, make sure there’s a diversity of input that represents your brand and your audience.

3. What could chatbots do in healthcare?

If we look at what is already going on, we’ll see that chatbot and chatbot style services are already being adopted:

  • In China, the search engine giant, Baidu, has launched a medical chatbot, called Melody, to help with the diagnoses of illnesses. This is being built into the Baidu Doctor app on iOS and Android. Which allows people to contact local doctors, book appointments and ask questions.
  • Babylon Health is a UK company that offers mobile health services such as consultations with GPs, but now also offers AI chat functions.
  • has launched Molly the virtual nurse that helps ensure discharged patients follow treatment plans. mimics the bedside manner patients need. It’s not quite a chatbot but shows a lot of the characteristics.
  • Your.MD’s chatbot offers a symptom checking function and is available on Android, iOS, Slack and other channels.
  • MedWhat is a medical assistant powered by artificial intelligence that can give proactive information as well as answer questions.
  • HealthTap’s chatbot allows people to ask health related questions on Facebopok and the chatbot will either get an answer from its database instantly, or seek an answer directly from its network of 100,000 doctors in less than 24 hours. 

4. So what should I do?

  • Don’t be scared that chatbots are too techy: the magic of effective chatbots is only partly down to the technology. The real magic is in the user experience. The best chatbots will have a diversity of input. Do not leave your chatbots to the IT Department alone.
  • Form a chatbot strategy: Think of it as part of a patient centric strategy. People are driving the adoption of technology such as chatbots. Increasingly they expect the world to shape around their needs. We see this with the adoption of smartphones so people can get information wherever they are and in the effectiveness of personalization and curated services. Effectively created chatbots will meet people’s needs. In the healthcare sector, patients and carers will be able to access relevant information faster. Health Care Professionals will be able to quiz chatbots on data releases to get their immediate questions answered. Part of the strategy should be to understand the potential of chatbots as well as any current limitations. But those limitations will largely be dictated by the level of commitment and investment an organisation is prepared to put in.
  • Build a business case: chatbots can do many things including:
    • increase your organisation’s reach, relevance and awareness by being on popular channels like WhatsApp and Facebook Messenger;
    • drive revenue by linking to ecommerce platforms or other sales channels;
    • save money by reducing the need for paying for people to be online to answer basic questions;
    • Increase the dissemination of accurate regulatory approved information by powering the chatbot with pre-approved statements;
    • increase adherence and product use by providing timely follow-up; and
    • increase audience understanding by providing a goldmine of understanding on the way people really talk about a subject.
  • Prototype: We live at a time where it is possible to move fast and learn quickly. Using agile development processes, test chatbots can be developed and tested early to learn quickly what the intended audience wants. Participatory development will enable the chatbot to meet your audience’s needs.
  • Don’t over-promise: Yes chatbots can be an effective member of your team, but start modest and build from there. For the moment, they are about quick dissemination of information and it will be absolutely necessary to hand over to humans for medical advice.
  • Be human: It’s important to build in plenty of character to your chatbot. Tone of voice, quirks of language and humility are all important to effective communication.
  • Chat to me or my chatbot (but go easy on her as she is a first stage prototype and I’d love to hear your feedback).