My last two articles of this column dealt with empowering sales, e-commerce and marketplaces and how to use them to increase business resilience by offering additional channels for customers. It is interesting (well, and a bit pleasing, too) to see that some of my arguments have been put into solutions, e.g. by Salesforce with its Salesforce Meetings and Salesforce Anywhere products, or by Zoho with it’s Zoho Workplace suite that targets at enabling employees.
In my consulting business I also see a notable rise of commerce projects, which also points in the same direction.
In summary, it looks like corporations have acknowledged that they need to invest in business resilience and enabling sales.
But how about customer service? There are all sorts of customer service from low touch like FAQs to high touch like the visit of a service technician. Service centres get more distributed as agents shall not be crowded into confined workspaces, which makes software packages like the aforementioned even more important. Service agents need to be able to access knowledge and to communicate with each other. Else customer service suffers.
But there is more to it.
There is the matter of customer experience that needs to stay high enough while products and services get more and more commoditized, difficult times or not.
According to the peak end rule “people judge an experience largely based on how the felt at its peak and at its end, rather than based on the total sum or average of every moment of the experience”. This is where customer service comes to play. It can either avoid a low or create a high at the end after some failure that happened before.
This is also where the pyramid of customer expectations comes into the picture. This simple model says that, at minimum, things must work. If a customer doesn’t get her desired outcome, the experience is poor, even unsatisfactory. Next, the process must be easy and efficient, frictionless for the customer. From there on, once in a while, if there is some excitement added to the result, we have a memorable experience – which might add some friction to the process, where it matters.
Too many companies still have something like the following customer service process in place. A customer visits the web site in search of information, navigates it, and eventually searches for the search box to get the desired information. However, searches often deliver poor results.
In between a ‘helpful’ chatbot might pop up that offers additional, separate, services. This is followed by a live chat or a phone call with an actual agent who regularly does not know who the customer is and what she actually wants, as all these systems are not connected.
Add the aforementioned business challenges that are caused by Covid-19 and there is a serious case for increasing the speed of digitalization of customer service processes.
Thanks for asking!
It needs a Customer Experience Architecture
A customer experience architecture organizes data, services, applications, and touchpoints with the goal of enabling processes that help to solve customers’ jobs to be done with minimum customer effort.
At the bottom of it is data, structured and unstructured, in different databases, file systems, or other repositories. On the second layer, we have services that make the data accessible and operate on it; one of them being machine learning. On the applications layer, we have e.g. a good functioning search, a conversational platform, knowledge management, and service ticketing systems that are helpful for agents and customers. These systems collaborate seamlessly. Finally, there are the touchpoints that are used by customers, like the web site, mobile apps, chat/voice bots, assistants, etc.
The secret sauce is combining them with an outside-in view that keeps the customers’ goals in mind.
Hint: Having a platform helps.
Let’s start with the customer – and employee – experience. Touchpoints gather customer and employee interactions and intentions; they turn them into signals and return relevant responses. E.g., even having a perfectly designed website, its main touchpoint is likely the search. The search is also relevant for the employee. Bear with me for a minute regarding this.
The search needs to pop up in context when the customer seems to be searching for something and already then offers useful advice in human language. Text, for time being. Note, there is no separation between search and chatbot. The chatbot is the interface for the search.
The chatbot is also the conduit for escalating a session to a human operator. If it cannot answer the question it needs to be able to escalate the request to the service centre along with the session conversation. It also continues to assist the service agent who works using a ticketing system by delivering contextually relevant solutions and even questions and next best actions while continually assessing the and learning from the answers it provides.
The ticketing system, in turn, needs to be organized around customer conversations, not tickets. The ticket is more something like an annotation to the ongoing conversation.
The search engine gets used by both, the agent as well as the chatbot, to find results.
The conversational platform orchestrates the process that involves the ticketing system, the chatbot, and the search engine. It ensures that customer questions are routed to the right agents and also feeds data about the usefulness of search results back to the search service via the machine learning service, to improve the search results.
Services work across applications to ensure that customer inquiries can get answered. They make sure that information is available, organized, of sufficient quality, and that it is known where the data necessary to generate information is missing. Of crucial importance are the data access services that need to be modeled around business entities, rather than technical ones, because the relevant data does in all likelihood reside in different storages.
Data is the raw material that is converted to insight and needed to come to a result. It usually lies in different databases and other storages, like file systems, on-premise or in the cloud. Some companies have started to consolidate it into a data lake, which then is a copy of the original data.
Paving the Way in 5 Steps
Now, that we have seen a glimpse of the ideal world, one simple question arises.
How do we get there?
- Just to be sure … If you don’t have yet, you should really set up a ticketing system that your agents can work with.
- Connect the chatbot and conversational AI to the search and make sure that the chatbot is the only interface that is used by customers. There is really no need to have a bot and a search box! This way, there is only one single self-service. This way, its machine learning capabilities can also get leveraged for improving the search result by constantly learning about the helpfulness of pieces of knowledge for a purpose. Where possible make sure that you can identify the customer as part of the conversation.
- Train the Conversational AI on routing rules, so that it can route issues that it cannot solve to the right group of agents. This will improve the overall answer times and hence, customer experience.
- Provide the full conversation to the human agent If the chatbot needs to escalate; the agent needs to be provided with the full script to properly prepare for taking over. In parallel, a ticket is created in the service system or, as said above, the ticketing system uses the conversation as the leading entity and tags it with the ticket id. The bot also needs to continue to support the agent (and learn from the agent’s reactions). This ensures a smooth handover and provides the ability to improve service
- Improve the organization of data and knowledge by improving the existing taxonomy, and maybe ontology, and by making sure that data is accessed via a service layer that defines which attributes a relevant real-world object has and where these attributes lie in the various data storages. In essence, this creates a virtual golden record that improves reaction quality.
- Connect the conversational AI behind the chatbot to the knowledge management system, so that its machine learning capabilities can get leveraged for identification of improvement needs of the knowledge base itself by either improving existing knowledge or creating new knowledge. That makes sure that information provided to customers is at its best possible state.
Well, it is. But it is worthwhile the hardship. For your customers, hence for you.