With the advent of conversational bots, companies are investigating and experimenting with using them in various scenarios both in customer facing as well as internal scenarios.
Conversational bots started out with a decision tree approach to answer predetermined set of queries. These conversational bots have now moved up the value chain to morph into self-learning models utilizing natural language processing and natural language generation algorithms to understand and prepare responses to user queries. Most of the preliminary uses cases seen currently deal with bots answering queries and providing or guiding the users to the right resources.
Some specialized bots also do task level transactions which backed by domain understanding, language capabilities and integration channels provide a much better use case for companies to consider.
On the other end of the enterprise spectrum, enterprise workflows have been working within the context of an enterprise function. They have been around for quite some time and provide a robust mechanism for companies to standardize processes across their business functions and software systems.
Existing enterprise workflow have been built to work with structured data as they solve a problem statement which was contained within the enterprise boundaries. These systems have been doing this job for quite some time and do this job with a fair degree of quality and predictability.
The digital revolution in play over the past few years has created tremendous opportunities as well as challenges. It has generated a fire-hose of data which data analytics solutions are being equipped to give aggregated and actionable insights to companies. The challenges come in the form of information loss due to the unstructured nature of events, existing systems not being able to handle velocity of information flow and generally the multitudes of channels of data ingress and egress. Some enterprise workflows have reacted to this by extending their channels to the digital and social domain. The question remains on the ability of these legacy platforms which were built to handle structured data within the constraints of an enterprise data flow to cope with unstructured data rushing in from multitudes of channels.
It is a fine balancing act. Workflows need to get into the consumer space and conversational bots need to understand the context of the process which is happening in the background and perform the role of an agent rather than just be a content provider or a single task transactional channel.
The next iteration for conversational bots would be in the form of digital agents which are trained not just for conversational skills but understand a business process and take part in it. The workflow systems underlying the digital agents also need to be thought through grounds up to work at pace with the velocity of the events that they are supposed to react to. These two components work together to provide a true end-to-end engagement experience to the end user over non-traditional channels whilst enhancing the customer experience.
These digital agents would have a presence across an omni-channel model and engage with the end users on channels of their preference. The digital agents have conversations with the end user and work within the context of an underlying business process engine thus shadowing the enterprise process in the digital space and thereby having the capability to take it to completion or predetermined exit points.
Processes are critical for companies and are equally important as data. Data velocity is inherently reliant on the process which are responsible for ingesting the data into the enterprise infrastructure either on cloud or on-premise. To draw comparisons when a new person joins an organisation, they are trained for performing a certain job which basically means they understand a certain process either implicitly which is based on common understanding between the various participants or explicitly which is driven by proper workflow systems. Similarly, when a digital representation in the form of a conversational bot is made part of the organisation offering, they need to be “trained” for a business process apart from conversational skills associated with the domain. This leads to tangible SLA’s and measurable goals for the conversational bots. Companies may then start tracking how many loans were pre-approved through the conversational bots, or how many insurance quotations and claims were registered, or how many service tickets were logged, resolutions conveyed, and feedback sought as compared to maybe clicks on the bot or queries answered which really do not give a right picture of the return on investment/effort.
Natural language processing and Machine learning is an exciting new space and coupled with business process provide an offering which companies may not just experiment with but also measure.
