As organizations grow larger, it becomes increasingly more challenging for them to handle their employee queries in a time-bound manner, and also maintain a personal touch to it at the same time. In order to tackle these concerns, Tech Mahindra recently launched the chatbot ‘UVO’, short for ‘Your Virtual Office assistant’, and positioned as a 24x7 workplace buddy. The chatbot will serve all 117,000 of Tech Mahindra’s employees, and will act as a consultative employee support tool.
In this conversation with People Matters, Kiran V S (who heads HR Transformation at Tech Mahindra) and Nikhil Malhotra (Head - Maker's Lab Innovation Center at Tech Mahindra) talk about the rationale behind introducing the home-grown chatbot, its key features and benefits, and the impact it is expected to drive, for the organization.
Q: What was the business imperative of automating employee query resolution? Which were the traditional methods used and how did they lead to the home-grown chatbot?
Kiran: Employees today expect the same kind of experience at the workplace, as what they are experiencing as customers, outside. Typically they look for convenience, personalization, and speedy response, while they are on the go. In order to achieve that, we thought of leveraging the advancement in digital technology to transform employee experience, and what better way to start, than moving our Human Capital Management (HCM) from a transactional interface to a conversational interface!
Currently, if the employees have any query, they either drop an email to the HR, or walk up to their desk; or alternatively, use the ticketing system, which is SLA driven, with a response time ranging between 4 and 48 hours. UVO will help address these queries instantly and 24x7, thus delivering consistent employee experience. Also due to the sheer number of emails and queries that come up, the HR team often gets bogged down with the time needed to respond to them, and is unable to invest the right amount of time for employee engagement activities. Similarly, the ticketing system has its own set of disadvantages, which includes delays at different stages of the query, the requirement of complete data to address the query, and a lack of personalization in the entire process. To overcome all these challenges, we decided to launch the chatbot UVO.
Q: What were the factors that led you to develop a home-grown chatbot over picking up an existing product from the market?
Nikhil: We generally do a lot of R&D on Artificial Intelligence (AI) and machine learning techniques. When we looked at the various bots that were available in the market, we found that these bots do well, but within a contextual domain, thereby not giving us the flexibility to work across domains. We also wanted to control the kind of user interface that we offered to our employees, and the vast amounts of data that was generated from the existing ticketing system. Therefore, we decided to go ahead with the internal implementation of our chatbot, rather than choosing a product externally.
Q: What was the solution implemented? Please elaborate on the design, framework, ways of working.
Nikhil: UVO is based on Tech Mahindra’s home-grown chatbot framework called Entellio, and is developed using Python technology. It is an open-source component of the Entellio framework, and utilizes a lot of contemporary machine learning and AI techniques like the algorithms on Jaccard, word2vec models of understand languages techniques which are together known as Natural Language Processing (NLP).
In order to build a conversational system, it is important to understand what the user intends to do.
Following from this, it is also important to ascertain the objects and entities on which the user’s intentions get applied. The current stack on which the system is built has got a brain of its own, leveraging NLP to find out the aforementioned intentions and entities. These inputs are semantically evaluated and mapped to what it really means, in the enterprise context. The word ‘red’ for example, would refer to the color in the universal context, but when we speak with a Vodafone virtual agent, it would mean the plan ‘Vodafone Red’, in his organization’s context. In this way, the meaning is contextually determined and the appropriate response is given to the user.
Q: What is the anticipated impact of the chatbot, and how are you tracking it?
Nikhil: The anticipated effect will essentially be reflected on the people management front. Whether it’s a query about training and learning services, or about R&D processes, the bot is expected to grow, and talk about all that will happen in the people management domain at Tech Mahindra.
As humans, we love to have conversations, and would probably be able to solve many more problems through conversations than through the other traditional methods. Based on that premise, from a technology perspective, we want UVO to become the de facto conversation interface for anybody and everything in Tech Mahindra.
Kiran: Our primary target is to eliminate all the level 1 queries that were previously being generated through the ticketing system, or through emails or walk-ins. In addition to that, we aim to get the level 2 queries down by about 30-40%. UVO's reporting system will enable the function owners to understand the process areas that need redesign, for better employee experience.
Going forward, we are planning to enable the chatbot to answer each and every employee query, right from induction to exit. We also want to do sentiment analysis based on the conversations that are happening in the chat, so that we can give the feedback on the employee mood in the organization. Finally, we want the employee well-being to also be addressed by the chatbot, so as to bolster employee engagement.