With the exponential increase in B2C (Business to Consumer) interactions, many brands and enterprises have turned to Conversational Bot Platforms (CBPs) as a solution for improving conversations with their customers. We have already established that
Robotic Process Automation (RPA) and chatbot integrations are a match made in heaven, so we will now explore the finer details including the inner workings and functionality of this dynamic.
A CBP can be built using multiple technologies, hosted on many platforms, or bought off the shelf.
Lets breakdown some of the key components of a CBP within the entire conversational life-cycle and look at how they can be integrated with NICE's Robotic Process Automation software.
How CBP and Robotic Automation Software Components Work Together during a Customer Conversation
1. Data Acquisition
Humans interact with CBP's using text and voice channels, in the form of:
- Text: (Social media, messaging apps or a brand's website)
- Voice: (Alexa, Google home and dedicated apps)
2. Data Structuring
This component of the CBP will attempt to analyze and understand the meaning of each interaction within the customer conversation. It has the capability to convert the unstructured data flow into a more logical data structure by utilizing a variety of models and algorithms for Natural Language Processing (NLP) or voice/audio analytics, to name a few.
So how exactly does the chatbot produce highly structured output data from unstructured input data? When the chatbot deduces that the conversation is reaching its end, it simply produces a form for the user to fill out. This ensures that the most critical data following on from the unstructured conversation is captured optimally, therefore allowing for easier analysis and verification of data entry integrity.
3. Insights Generation
Now that the structured data has been well received, the CBP will attempt to understand the meaning of the data in the same way that a human would. Here, Natural Language Understanding (NLU) models and algorithms are applied, as the system works towards understanding the context of the conversation and ultimately the customer's intent.
In the event of the system failing to deduce the customer's intent, then a human employee will need to intervene by continuing the conversation with the customer. Human employee intervention has many terms, some of which includes: Conversation Routing, Transitioning, and Hand off. They all face the following challenges:
- The customer interactions need to be analyzed, understood and completed in real-time.
- The human employee, needs to understand the call context up until the point at which he/she intervenes.
- In order to successfully close the customer interaction, the employee will need to have the right skills and systems permissions.
This is where NICE Desktop Automation (attended automation) comes into play. With a robust human-robotic interface and ability to communicate in real-time with multiple front and back-end systems – it supports any human employee to execute multiple tasks in order to conclude the customer interaction in a quick, efficient, and error-free fashion.
Should a human be required to intervene, it is advisable to notify the customer that their conversation is now being escalated to a Customer Service Representative.
4. Decision Making
During a customer interaction scenario without any human intervention, the CBP will need to apply machine learning capabilities, historical information and training to all of the data received in order to make a decision about the next best action for the customer.
It is imperative that any CPB has the capabilities to enable developers to navigate the interaction in order to effectively retrieve all pieces of info, whilst also taking real-time changes into account, as a result of the customers' input in each and every moment.
5. Taking Action
At this stage of the process, the CBP now understands which business processes it needs to execute in order to successfully complete the interaction.
Since CBPs today are constrained by APIs from the systems they interact with, executing more complex customer interactions may be a challenge. Take for example a simple process of executing a web service call to enter a pizza order. This can be actioned with ease, but when we bring in more complex business processes involving many disparate systems from different parts of an organization (some of which are not exposed to the outside world), what now?
This is where NICE Robotic Process Automation comes in. The CBP has the capabilities to action its decision by connecting to a menu of pre-defined and smart automated workflows. It then simply triggers one of the workflows to be executed by Robotic Process Automation software, with the added capabilities to communicate the results back to the CBP.
6. Replying to User
In the event of RPA software being utilized to take action, its important to note that preconfiguring an SLA (Service Level Agreement) is also an option. In this instance, the preconfigured SLA will define whether the CBP should wait for a sync result i.e. should the customer wait until the completion of the process before obtaining the result of the conversation? In sync mode, however, there is a pool of RPA software clients dedicated to handling conversations, requiring the execution of complex back-end processes. Take for instance a process which takes six minutes to execute. Instead of keeping the customer waiting, the CBP will send a message to the customer saying "Your request is being performed and an e-mail confirmation will be sent to you on completion." This saves the customer time that would otherwise be spent waiting on the phone for the next Customer Service Representative to continue the process. Instead, the customer is assured that his/her request is being taken care of, while the RPA software client simultaneously wraps up the process execution.
The CBP then replies to the customer using Natural Language Generation (NLG), by constructing a sentence in a manner that a human is accustomed to reading or hearing (in the event of a voice conversation). The NLG process is very much a part of the CBP's conversational flow i.e. the component in which developers design different options and mechanisms for the conversation to evolve. It is essential to ensure that there is also an option to transfer the conversation to a human employee.
7. Monitoring / Dashboards
This component tracks and assists operators to run user and conversation analytics. These insights can be applied to shape and optimize the flow of future conversations, based on what previously worked and did not work, in addition to how customers react to different CBP generated responses.
NICE's Advanced Robotic Process Automation platform has built in dashboards and a dedicated control room, through which CBP and RPA interactions can be analyzed.
Key Take Out
We have now assessed some of the key components making up a CBP, and many options exist to build holistic and comprehensive CBP solutions from these different, yet complementary elements. However, not every platform or off-the-shelf product has all the capabilities needed to execute an end to end process following on from human conversations.
NICE's comprehensive Cognitive RPA platform has an open framework, enabling the integration of desktop and robotic automation with any of the key cognitive components making up a CBP. This type of process automation platform could be the silver bullet supporting chatbot solutions for complex enterprise requirements.