This is the write-up from the fifth in our series of virtual coffee breaks – designed to provide a caffeine-enriched forum for discussion amongst senior executives who might otherwise be going stir-crazy during this period of lockdown.
The topic for this session was: ‘With human operators sick or self-isolating and vital call-centres struggling to cope, can intelligent chatbots deliver the answers?’ and our guest speaker was Andy Feltham, Innovation Lead in Conversational AI at Filament AI.
Andy started the call by thanking everyone for taking the time to join this conversation, focused on the pressing need to innovate. As a former IBM Master Innovator and manager of their Emerging Technologies team, Andy has spent several years applying Machine Learning to a range of innovation challenges.
In the past there have been times when people looked to innovation for novelty, but right now the need to innovate has taken on a new level of urgency, as we look for ways to respond and adapt to the challenges of the Covid-19 pandemic.
We all know the score: Stay at Home; Protect the NHS; and Save Lives. The significance of this situation, and how we have to deal with it, has a huge effect on all of us. The impact on businesses, large and small, is immense.
Ocado, for example, has 1% of the UK market share for supermarkets. Shortly after the lockdown was announced, their website was inundated. They had more than 12,000 users per minute and hundreds of thousands of customers waiting in a virtual queue.
If just 1% of these 12,000 users per minute had questions for the call centre, that would be 120 questions per minute. And how many of these questions would be unique?
Businesses are struggling to keep up and often falling back on the only things they know they can do – reduce their levels of service or even close their doors. For example, Nationwide has had to close its branches at weekends.
Call centres are particularly badly affected, as they are very much in the front-line and are having to cope with increased volumes, reduced staffing (due to a combination of isolation and sickness) and logistics (especially related to social distancing).
One call centre has segregated the staff on each of its three floors – allowing staff on the upper floor to use the lift, staff on the middle floor to use the right hand staircase and staff on the lower floor to use the left hand staircase.
Many of these responses are very limiting. They will only get us so far in dealing with the need to protect staff, serve customers and reduce costs as we deal with the impact of Covid-19 and, quite possibly, head into a longer-term recession.
Fortunately, the technology is there to do what we need. Chatbots, in particular, are perfect for reducing the load on call centre staff. They have endless capacity, are always connected, always available, consistent and responsive, and have low running costs.
So why aren’t companies making greater use of chatbots in this crisis?
There are probably two reasons. The first is the shear pressure they are under; it’s hard to look up and see the solution when you’re head down fighting the battle. The second is a mismatch between expectation and reality.
Our expectations of chatbots have been set by portrayals of all-seeing, all-knowing AI oracles. Whereas, most chatbots today are relatively simple and capable of answering only a few questions really well. And there are lots of really bad examples of chatbots that struggle with dialects or suffer from poor conversation design (e.g. the PayPal chatbot that responded to a customer who said “I got scammed” with the answer “Great, ..”).
This has led us to perceive chatbots as being at the “Peak of Inflated Expectations” stage of the Gartner Hype Cycle. When the reality is, measured against what we actually need right now, chatbots are really on the “Plateau of Productivity”.
What we need now is a solution that can handle the standard questions: When is my delivery? What are your hours? How are you handling Covid-19? What is your telephone number? When will you have new delivery slots? These are what are called “short-tail” questions, i.e. they are high frequency, with a small number of unique intents, and relatively easy to handle.
A good example of this type of chatbot is HSBC’s AiDA, which deals with relatively simple queries (account balances, password resets etc.). It handled 25,000 conversations in one month, reduced customer contact costs from $7.50 to $0.62 and saved the equivalent of 15FTE’s in productivity, while saving customers an average of 12 minutes per conversation.
Chatbots of this nature are not only useful in the context of call centre interactions. They can also handle ecommerce transactions, execute booking requests (via APIs), provide status monitoring (where’s my delivery?), etc. In fact, the more seamlessly they’re integrated to relevant systems, the more value they can add.
It’s also important not to think in terms of just one channel. By analysing the types of users you have and how they are interacting, you can understand where your chatbot needs to be present, in order to reduce the load. The content you produce can, and should, be delivered as pervasively as possible, across all the channels your customers use: website, Facebook, WhatsApp, Alexa, etc.
So, the key message is: chatbots are proven technology which you can apply now, to take the pressure off your call centres. You don’t need to do a proof-of-concept or develop an MPV. Just do it. Your competitors probably are.
If you do decide to take this leap, there are a few things to get right:
This pandemic is changing the way we work and the way companies, customers and employees interact. It’s accelerating the adoption of many innovative solutions, including chatbots. We should start looking at how we leverage this technology to enable the virtual, remote ways of doing business that are becoming the new normal and which may well persist beyond this immediate crisis.
We are all used to the accelerating pace of change. But that acceleration has been kicked up a notch by Covid-19. Even if you were able to iterate quickly before, you need to be able to iterate even faster now. The good news is you can get a chatbot up and running in about a week.
If you would like to see an example of a chatbot Filament developed in just one weekend, please visit https://coronavirushelper.com/
If you would like to learn more about the tools Filament can provide to help you develop your own chatbot please visit: https://enterprisebotmanager.com/
If you have found this write-up interesting and would like to register for future Virtual Coffee Breaks or learn more about our other events, lease visit www.clustre.net/category/events/