As we discussed in an earlier article (When and why do chatbots make a mess of customer service), chatbot deployments are in danger of damaging customer confidence and customer service levels because of inappropriate applications.
So what are the tasks best tackled by chatbots? Where in the customer journey should they ideally sit? And where are live agents more appropriate?
Let’s examine where chatbots circa 2017 can best be used.
A narrow purpose
Warnings by experts such as Forrester suggest that the majority of chatbot failures in 2017 will be a result of projects attempting to tackle too broad a domain. Forrester believes that the chatbots that will succeed this year are those that will stick to a narrow purpose, pointing to applications by USAA, Staples and Radisson Blue. So how are these brands using chatbots?
USAA, a provider of banking and insurance services to the US military personnel and their families, uses its chatbot to provide immediate online support and answers to its customers’ most critical needs, such as activating cards, changing PINs, adding travel notifications and reporting lost or stolen cards.
Content seriesView full content series
Office supplies retailer Staples has partnered with IBM Watson to fuse an AI chatbot with its renowned Easy Button. The Button – a way for organisations to quickly reorder supplies – now has an intelligent, voice-driven interface that interprets the speaker’s words, directs the conversation and gathers information to identify the correct products to order. If the system is unsure what to order, it suggests products based on past orders so that the correct items can be chosen. In the event that the system still doesn’t understand the request, the customer is connected to a human agent, with the conversation used to train the system to handle similar requests in the future.
Radisson blu, meanwhile, has introduced a virtual assistant at 12 of its hotels, using Interactive Text Response (ITR) mobile SMS to help customers to self-serve. The chatbot, called Edward, provides directions and tips, and allows guests to check and request hotel amenities – such as towels or room service, get information about local bars and restaurants and even express complaints – simply by sending a text message.
“One of the biggest appeals of chatbots is that they are a low cost option for helping customers get answers to their questions efficiently. However, it is worth remembering that chatbots are powered by artificial intelligence (AI) and, in their current form, perform best with narrow and tightly defined instructions,” notes Dave Campbell, vice president of customer engagement & support at LogMeIn.
“So while they are wonderful and cost-effective, they are not for every brand or every situation. Low-touch questions like resetting a password, for example, can be handled through AI. For more complex issues, human-to-human engagement is likely a better and more efficient choice. Chatbots are exciting and great for certain use cases, but they should be one tool in a customer engagement arsenal rather than a ubiquitous customer engagement tool.”
Tom Nickalls, product manager at Jadu Continuum CXM, adds: “Right now, text-based chat bots offer immediate gains by being able to carry out a plethora of simple and repeatable tasks. Therefore, when considering their role in customer journeys, organisations should be looking at the areas where requests and questions are being repeated time and time again or where issues are most prevalent.”
The human touch
But deployments also need to consider the touchpoints in the customer journey where ‘the human touch’ simply cannot be replaced.
Dr Nicola Millard, head of customer insight and futures in BT's Global Innovation team emphasises that as we automate more, we need to look at where the value of human conversation in relationship building comes in, and where it is appropriate to automate.
She explains: “Automation can effectively steer the customer through the “known knowns”, without the customer having to plough through lists of Frequently Asked Questions (FAQs), or lengthy IVR menus. This works when things are simple, but complexity, emotion, and complaints can throw a spanner in the conversational works. If we get angry, we tend to tell long, incoherent and rambling stories which are hard to parse into something meaningful by machines.
Automation can effectively steer the customer through the “known knowns”, without the customer having to plough through lists of FAQs, or lengthy IVR menus. But complexity, emotion, and complaints can throw a spanner in the conversational works.
“Sarcasm can also really confuse ‘bots. This is why sentiment analysis can fail horribly on some complaints coming through on digital channels, like social media and email. Comments that would probably get parsed as positive include: “Thanks, @TrainCo for my free sauna this morning”; “It was so good to see that your maintenance department hadn’t spoiled things by making unnecessary repairs”.”
Millard suggest that when faced with ambiguity, complexity, sarcasm or a dead end, the best ‘bots should effectively act as “IVR for digital” and rapidly triage the customer to the human advisor with the skills that are most likely to get the customer to their goal. The advisor has the advantage that they should be able to see or hear the conversation that the ‘bot has had with the customer, so the customer shouldn’t need to repeat themselves.
She continues: “Undoubtedly, the ability for machines to process natural language will improve, but they may never be able to become scintillating conversationalists. Advisors find it far easier to have a conversation – as long as they aren’t turned into the human equivalent of robots by too many rules or scripts. Ironically, the march of technology may make us re-evaluate the value of human connection on customer relationships.”
Thomas Rødseth, VP of product and marketing at Intelecom, sees two specific tasks that are particularly promising for businesses to introduce chatbots into service.
- Improving the use of existing FAQs. “Here, the content and often the linguistic analyses are in place and by changing the user experience from a defined number of predefined questions to a more open dialogue supported by a chatbot has proven powerful and an achievable first step into the use of bots,” says Rødseth.
- Collecting information in the initial part of the conversation. “Start with existing web chat offerings, or social media messaging applications like Facebook Messenger, and then introduce bots to automate the repetitive tasks of collecting information at the beginning of a dialogue,” suggests Rødseth. “Rather than just greeting the user with a standardised welcome message the chatbot can start a human-like dialogue with the user collecting information such as name, customer number, and an initial classification of the problem. From these simple answers, references or a transfer to existing self-service solutions can be given. An alternative option is a handover to a “colleague” or an agent to assist with the processes further.”
Beyond these tasks, organisations should examine the journeys that their specific customers take, identifying where opportunities exist for chatbots to add value. For instance, points in the customer journey where chatbots can be a quicker, simpler option than a call centre, or where enquiries may be of a sensitive nature so the sense of anonymity provided by a chatbot may be valuable. While there will be common tasks across businesses, these opportunities will often differ from organisation to organisation.
Ultimately, there is no ‘one size fits all’ solution to servicing customers - it really comes down to the nature of the customer journey,” says Bill Safran, CEO at Vizolution. “Organisations need to identify the possible breaks and stalling points in each journey and then choose the solution that can address them most suitably, whether that’s a simple commonly asked question tackled by automation, or a more complex scenario more suited to human flexibility and adaptability.
“The role of chatbots within customer experience really depends on the type of customer service being provided. Typically, chatbots work well in simple, self-serve customer journeys, where common questions can be easily answered and the customer journey is quite linear, an example of this would be going online to check a bank balance or making a simple purchase online.
“When it comes to more complex journeys, where the journey is not typically linear or there are a range of stalling points or ‘uncommon questions’, an element of human intervention is needed. A mortgage customer journey, for example, involves a number of stages, plus the exchange of documents, which means a chatbot is unlikely to be able to provide a sufficient level of customer support. If a chatbot was to be used in this type of scenario, it would most likely result in a permanent break in the process and a lost customer as customers would not be able to get the level of personal support they desire. Also customer journeys that go down an unhappy path - such as failed credit checks or failed proof of identity - would most likely need human intervention to get the customer journey back on track and guided through to completion.”
One final point to consider is the readiness of your own customer base to use chatbots, which again can vary from organisation to organisation.
As Rødseth notes: “Firstly, know what your customers want now and feel comfortable with. It’s no use introducing chatbots if the majority of your customers still prefer to interact with you by telephone or email, and you are happy with this, too. However, if you have already successfully used self-service to improve customer satisfaction levels then it could be a good time to improve upon this and to mix bots into the offering.”
Jim Bowes, CEO and co-founder of Manifesto, adds: “My advice would be: understand your audience, carry out user research and test a prototype of a bot with real users before rolling it out to customers.”
Understand your audience, carry out user research and test a prototype of a bot with real users before rolling it out to customers.
Nickalls agrees: “It’s got to be well tested, and not just at the beginning of an implementation - continuous testing all through the life-cycle of the service. In particular, that testing should be user-driven, with real customers.”
What emerges here is a picture of chatbots and live agents working in tandem, where chatbots are applied to simple journeys and processes which require low engagement, but with the ability to transition into live agent interactions for higher engagement when the customer journey demands it. And if this is accurately defined, customer service levels and efficiencies can be significantly elevated.
“Smart people plus smart technologies are better together than as adversaries,” concludes Millard. “Smart technologies can take away the mundane stuff, precision route complex issues with the right agents, allow agents to switch channels seamlessly and make things easier for customers to get to their goal.
“This is about using AI to help the customer get to their goal by handholding them through a process or personalising things according to channel, preferences, profile, location, etc. If this fails, then it needs to seamlessly triage the customer though to an expert agent. AI can help the agent to stick the fork into the complex spaghetti of internal process whilst they use their knowledge, ingenuity and skills to help the customer get to their goal.”
About Neil Davey
Neil Davey is the managing editor of MyCustomer. An experienced business journalist and editor, Neil has worked on a variety of newspapers, magazines and websites over the past 15 years, including Internet Works, CXO magazine and Business Management. He joined Sift Media in 2007.