Digital customer service: can chatbots help?
What was the real agenda of omnichannel implementations, whereby customers could manage services and make purchases using the touchpoint of their choice, moving seamlessly between one and another? Was it to enable the digital customer to choose any sales or marketing channel of their preference, or to drive the digitization of all operations? In a sense, the underlying motive now seems unimportant, as the ability to deliver purely digital services has turned out to be a winner for all parties. Customers no longer have to visit a physical shop or spend a long time waiting for call center help, as they have all the tools they need in their pockets. At the same time, service providers benefit from churn reduction (thanks to better service and fewer frustrations for customers), and can count on OPEX savings of around EUR 1-3 per minute compared to just a few years ago.
Most service providers have already launched their online portals, helping customers to purchase services and products, and to resolve any issues that may arise, all on their own. But the process still requires some patience, as customers navigate through the portal to reach the information they require. There is of course the option to use a chat box or the traditional method of talking to a call center agent, but with operations becoming increasingly global and contact center staffing no easier to scale up than in previous times, there is a better option.
Enter the bots – services such as Siri, Cortana, Alexa and Google Assistant. They may differ cosmetically, but all have the distinction of being cloud-based and easily scalable enhancements to chatbot customer experience. In truth, they have not yet found wide deployment in the telco market, and where they are in place (Vodafone’s Tobi, for example) their capabilities are somewhat limited. Yet the bot approach is growing in popularity.
With bots, customers can express their intent (using voice or text) using language that comes most naturally to them. Chatbots, using their natural language recognition capability, evaluate that intent to guide the customer to the relevant topic information, or propose appropriate action.
There are three groups of bot use cases that bring immediate value and are a good starting point for the implementation of chatbot programs.
The most obvious and theoretically the simplest use cases are related to informing customers about offers, offering comparisons, checking invoices and providing updates about the latest orders or invoices. Most providers already possess wide knowledge bases containing FAQs, topics and social forums, or already have relevant information somewhere on their online portals. So, with the availability of structured data, a chatbot can easily demonstrate its wisdom by responding accurately and appropriately to a direct question from a customer.
Self-installation and problem-resolution scripts are the natural domain of conversation-based support. Handling issues starts with a single text box, with questions delivered at each point of the conversation between customer and bot. Chatbot text boxes should be easily actionable – especially on smartphones where it’s more likely that the user will want to click instead of type. But the execution of actions arising from a chatbot conversation relies heavily on underlying systems which must be capable of delivering industry-specific solutions (for example, order a SIM, pay an invoice, block a phone, add roaming, etc.).
Commerce and Marketing
Compared to Information and Support, use cases related to Commerce or Marketing have their own challenges. Primarily, there is a need to set up bots with more knowledge about a given customer, and which are capable of delivering relevant content that does not alienate or irritate the user.
Chatbot implementation offers real opportunities and potential value. However, even relatively simple use cases related to providing information have not yet been implemented on any real large scale. Even when bots are available, as is the case with providers such as Windows, the default seems to be a “question click” approach rather than a real “how can I help?” welcome for users.
This just shows how early we are in the adoption of this technology, and how long a journey we have in front of us. Implementations will certainly become more widespread, but the question is – when?