Chatbots? Seriously?

Some of us remember Clippy, Microsoft’s assistant in Office, now much derided as an interfering pain that never seemed to help.

clippy in action

And for years we’ve been seeing chat boxes on home pages of sites we visit, but how often have we ever used them? And how many horror stories have we heard about them going wrong?

So why do bots seem to be all the rage? For quite some time, I’ve asked this question incredulously, skeptical that conversational approaches to user interfaces will ever be of much value.

But in the last few months, I’ve started to come around to the idea that a conversational approach to user interfaces will increasingly have a place in a future of computing that will not always be screen-centric, whether text or voice based (such as Alexa, Amazon’s in-home voice based assistant) or as a way of connecting with a human or driven by machine intelligence.

chatbots going wrong, an infinite loop of question and answer

These are very early days, despite earlier false starts like Clippy, and even Siri. Advances in speech-to-text algorithms (already very accurate for many use cases), in Natural Language Processing, and other machine learning techniques for making sense of free form text, available through easy and inexpensive to use APIs from the likes of IBM’s Watson, AWS, Google Cloud Platform and Azure, alongside many startups, mean that exploring and experimenting are low cost and low risk, but with a potentially very high upside.

One area that is looking with considerable interest at AI based chat technologies is the insurance industry.

When a large storm is bearing down on an area, insurance companies know they are about to be hit with a huge number of enquiries. This is the time they must deliver on promises to their customers who’ve paid premiums for years, perhaps decades: the promise that in their hour of need they’ll be there for them.

Call centre based enquiry lines scale linearly with the number of people affected. And yet, increasingly, it seems people don’t actually want to call, they want to visit a website, or use an app.

‘Chat’ (the term seems too trivial, and non-serious for applications like reporting a claim to an insurance company) based systems will also only scale with the number of humans able to engage with enquirers. But AI based chatbots (seriously, we need a far better name for these types of experiences – sadly, Conversational UXs will get abbreviated to CUXs and the less said about that, the better), even if they might help triage a relatively small percentage of claims – the simplest, the ones that are least emotionally charged – and escalate the most significant claims – like a family home destroyed – to people as quickly as possible, might save money and make for a better experience both for the person whose claim or enquiry was dealt with promptly by a chatbot, and the person whose claim needed the human touch, who waited less long, and who dealt with someone less under the pressure of time to address a huge and growing backlog of claims.

I feel  the intelligent, sensible, sensitive use of technologies that are available today might already improve the customer experience in so many areas of commerce, in government services and elsewhere.

In fact, I think it’s so interesting and potentially fruitful an area of exploration that we have a whole session at Respond focusing on it. This features Elizabeth Allen from Shopify, which has been exploring and investing extensively in ‘conversational commerce’- a term coined by Chris Messina (who also happens to have invented the hashtag, among many other contributions to web technology and practice), and locals Laura Summers and Warwick Cox, who are using conversational approaches to UI in interesting ways.

I really do feel we’re at the tip of an iceberg here (where’s my intelligent agent to point out a terrible cliche when I need it?).

Let’s see how my prediction plays out a year or two from now. But my advice is at least start exploring.

We are, and we’ll have a lot more to say about this area in the coming weeks and months.

2 responses to “Chatbots? Seriously?”:

  1. Chatbots are great for triaging.

    Considering most first-level conversations with call-centers are pretty much entirely scripted, you might as well automate those and get away from the frustration of knowing you’re essentially talking to a chatbot through a human being that’s reading the chatbot’s script.

    If the chatbot can resolve your issue, so much the better, if it can’t you’re off to talk to a real human with (hopefully) more latitude to fix it.

    As tech improves, the level at which you’re escalated to a human being will rise. Cost-conscious companies will raise it earlier than customer-focused ones. You’ll mostly still be left with the same frustration that you have now when contacting the former. One day the technology will be so good we’ll barely ever talk to a human for such interactions. Much like we don’t remember talking to operators to dial a number, or a bank clerk before ATMs were a thing.

    • By: John
    • April 28th, 2017

    Yes to all, and it’s happening now.