Professor Peter Thomas, COO of the Leasing Foundation, launches a regular column on new technologies that will impact the industry. This month he gives an overview of chatbots
What is it?
A chatbot is a piece of software that tries to talk to you in a more conversational manner. It uses text – and in some cases, speech recognition and voice responses – to help you get things done. You will have come across chatbots on sites like online banking, shopping and in the form of Apple’s Siri.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
How does it work?
Behind the scenes are complex software algorithms that process what you say to the chatbot – whether that is “Siri, call Jo at work” or typing “I’m interested in opening an account”. They use a database of phrases specific to a domain – no point asking a banking chatbot about groceries – plus knowledge of the structure of language, and in the case of Siri, digital representations of speech.
Collectively this set of resources – algorithms, domain and language knowledge – is artificial intelligence (AI), an area which has been researched for 50 years.
Researchers in all of these areas, and especially in processing speech and language, have built more complex and sophisticated software, facilitated by more computing power, to get to the point where the error rate – misinterpretations of what you type or say – are becoming much rarer, and if the software does not understand what you say or type, it can respond gracefully with a phrase like “I’m sorry, I didn’t understand that. Did you mean…?”
For example, the software behind Siri – which processes your requests in the cloud and so has access to a huge computing power – is smart enough to know that when you say the word “Call” it only needs to limit its search to your phone book; it also now knows that what is likely to come next is a person’s name; and that the response it will give will either be one of “Did you mean Jo Davis or Jo Williams?” or “Which number for Jo Davis would you like to call?”
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataWhy are they important?
There are much more complex and powerful ways that AI software can operate that will completely change the way we all interact with information, with businesses – and potentially with each other.
For example, software accounting company Sage is developing a chatbot that helps small business owners manage expenses. If you show the chatbot an image of a receipt, it is smart enough to figure out that it needs to be stored as an expense, and to ask you a series of questions such as: “Do you want to save this to the general expenses category? If I see this kind of receipt again shall I do the same thing?”
Companies like Sage, Amazon, Apple and Microsoft are putting resources into designing chatbots that make you feel that the interaction is natural, intuitive and not intrusive: if the chatbot does not respond enough you might feel you are being ignored; too much and you will be irritated. If the chatbot does not have the right domain knowledge you will think it does not know what it is doing.
There are some dangers here, of course. Recently, Microsoft’s Tay chatbot was taught by malicious users to start sending hate-filled posts using offensive words to describe ethnic minorities, and appearing to support totalitarian policies.
Tay was a very sophisticated chatbot, using machine learning – the ability to learn from previous interaction with people – to change its behaviour. Unfortunately, Tay did not have built-in filters, and so started to behave like users who goaded it into learning their offensive behaviour.
What the Tay episode tells us is that while technologies like machine learning are very powerful, they are not infallible, and so have to be used with caution. But implemented effectively they can add huge value to businesses and their interactions with customers.
Chatbots can replace the mandatory phone trees that businesses force customers to use; they can integrate directly into messaging systems – which are now being used by more people than use social networks, and so provide a more seamless experience of interaction with a service or brand.
They can make it easier to allow people to engage with a product or service before being handed to a human agent, and they are suited to mobile platforms, which is where people spend over 90% of their internet time.
Try using Facebook Messenger to interact with CNN’s chatbot at www.messenger.com/t/cnn and you will see how easily this experience integrates into how you use a messaging service: it uses a conversational style to encourage you to ask questions, taps into an extensive database of other users’ questions, and integrates with CNN’s application programming interface (API) that provides access to CNN’s live news stories.
What impact will it have?
Chatbots are here to stay. Even if their uses at the moment are limited, the knowledge being built up around algorithms, AI, what customers prefer and want and the growing effectiveness of computing resources will mean they will become ubiquitous.
There is also an ecosystem of third-party chatbots, native bots, distribution channels, and enabling technology companies that now allow businesses to create chatbots to respond to complex queries about products, accounts or services without the need for human intervention.
Just as automation decimated blue-collar manufacturing jobs, AI will decimate white-collar information worker jobs. An alternative view is that chatbots are a way to add capacity without the headcount, and to do so very quickly. <
