Consumers are increasingly aware of the presence of bots, especially voice bots, in their interactions with companies. It is expected that within the next three years:
- The participation of bots in customer conversations will rise by up to 40%.
- Virtual assistants will revolutionize customer relationships.
- Twenty-five percent of customers will prefer to use a virtual assistant instead of a company website.
These trends are highlighted in a recent study by Capgemini, Conversational Commerce: Why Consumers Are Embracing Voice Assistants in Their Lives, conducted among 5,000 clients in Germany, France, Great Britain, and the United States.
Halo, Hello, Hola, Bonjour … artificial intelligence speaking!
Of course, this growing phenomenon hides artificial intelligence in the form of cognitive capabilities, for example the ability to recognize voice, understand the meaning of phrases in natural language (natural language understanding), and improve accuracy through machine learning. Thanks to progress in understanding natural language, chatbots have become very effective and sometimes even indispensable. These intelligent assistants can work non-stop, are never tired and react immediately, and gradually increase their role in enterprises. Over the last several months, most large companies have started developing their own bots.
However, the implementation of human–machine interaction is a rather complex process. Some people imagine that a bot is like a magical black box, “please install it and it will work.” But in fact, you first need to design the desired interaction. Think about what kind of experience the user is supposed to have with the bot, how will this experience help the user achieve their business goals, and how will it be a better experience than what the user is used to. Next, you need to develop dialogue flows, determine the bot’s personality, and integrate it into existing communication channels in the enterprise (for instance Skype, Slack, Teams). Then, you need to design and train natural language processing (NLP) models. This will enable the bot to understand the context of a dialogue and recognize user intent (for example, “I have a problem logging in,” “I want to order a pizza.”). It is necessary to supervise the machine learning process and make sure that the model understands the user. And, you need people to achieve this – UX managers, data scientists, AI developers, chatbot content designers, testers, project managers, and architects.
Do bots perform better than people?
Early experiences with machine learning were quite challenging. You had to train the model with hundreds of thousands of photos of an object (such as a chair) for it to be recognized at an acceptable level. However, the speed of machine learning grows exponentially. In 2018 for example, OpenAI defeated a team of former professional players in Dota 2. It learned by playing 180 years of games against itself for several months.
Machine learning algorithms have surpassed humans in the recognition of objects. As described in the artificial intelligence index 2017 Annual Report, these algorithms were able to achieve more than 97% accuracy, while humans remained statistically at 95%. Business tools such as Microsoft and Amazon are 95% accurate in speech recognition, in other words statistically at the same level as humans. At the I/O conference last year, Google presented a voice bot making an appointment with a hairdresser. The conversation between the bot and the hairdresser went quite smoothly, so much so that the hairdresser did not realize that she was talking to a machine. This April, Google released this functionality in the US for making restaurant reservations.
Today bots are available on mobile and desktop devices and via multiple channels, such as mobile apps, websites, social media platforms, instant messengers, and smart speakers. They are used in many industries. Large retail companies use them to help with shopping, check order statuses, proactively suggest other products, and automatically drive cross-selling opportunities. In the banking sector, where digital assistants already support the most frequent inquiries or customer complaints, bots are a hot topic. For example, the British bank Barclays allows its consumers in Africa to transfer funds via smartphone with simple voice commands: “Hey Siri, send Sara £15 with Barclays.” Bots are also present in offices. For example, L’Oréal employs bots in HR for recruiting, while Capgemini, includes them in digital workplace service offerings, such as the Connected Employee Experience, which links workplace, office, and employees.
Digital brand identity is what makes a caring bot
The real challenge is to build a bot with an identity that reflects the brand or the culture of the organization. Such a bot must have its own personality. It has to know how to communicate (using either formal or less formal language), look appealing look, and last but not least, it should have a name! The linguistic aspect, for instance company-specific terminology, is also important. In one organization, an employee can be referred to as an “employee,” while in another “colleague,” or “associate” is more appropriate. Such details have to be identified and embedded in the bot upfront. Also, it turns out that users don’t necessarily want to interact only with other humans. They’d rather have the option to contact a company with a question at midnight for instance, and get an answer straight away. They expect 24/7 availability and value from every interaction. What’s important is the choice, however. Users have to be informed that they’re speaking with a bot or that they’re being handed back over to a human operator.
Capgemini research shows that clients who have adopted bots in their organization report an increase in customer satisfaction. However, it’s important to remember that people don’t really like change. They need to be freed from preconceived notions and convinced of the change at hand. Some form of change management, for instance Capgemini’s Digital Adoption methodology which effectively engages customers in the use of bots, can facilitate this process. The expectation of being served by a human declines in the face of an always-available and “caring” bot whose personality and communication style aligns with the company brand and culture. It’s no wonder that more and more organizations, conscious of the expectations of their customers and of the potential savings to be gained, turn to customer service bots.
To learn more about it, contact me.
Lukasz Ulaniuk is head of cloud cognitive Solutions at Capgemini. Read more Capgemini blogs here.
