Q&A with Sourav Banerjee: How Customer Service Automation is Benefitting Businesses Across Industries


by Shamia Mohamed

It is an undeniable fact that analytics and automation are essential to gain a competitive edge in any industry and the customer service industry is no exception. As more and more customers expect quicker and more personalized resolution of queries, and while support volumes are on an exponential rise [1], it is of no surprise that many service leaders believe investments should increase in self-service channels like self-help portals to enhance support capabilities [2].

However, there are several recommendations for this process of automating customer service. Below, Sourav Banerjee, Head of Innovation, TheMathCompany, discusses the specifics of the process and how you can optimize your business operations by adopting customer service automation.


Question 1: The increasing popularity of support automation is now transforming customer service in enterprises. What do you think are the biggest benefits that come with integrating AI into customer support?

Sourav: Customer service can be enhanced significantly by AI using data that is generated from a combination of sources, like customer service conversations, transactional data, and community forums, to name a few. AI is already adding significant value across industries in a few areas such as

  • Agent assist (which helps customer service agents in resolving queries)
  • Customer self-service
  • Product return automation
  • Automated case routing

Ultimately, these solutions lead to better customer satisfaction and quicker resolution of customer issues, which is a win-win for both the customer and the enterprise. Virtual agents also helped enterprises handle a sudden surge in call volumes due to the pandemic by being online 24/7 and resolving common queries without the involvement of human agents.


Question 2: Which industries could benefit the most from customer support automation in your opinion?

Sourav: Any industry that generates a lot of data through interactions with customers will benefit from utilizing AI to enhance customer service. Today, people use chatbots to book flights, place food orders, buy clothes, schedule appointments, and more. Here are some instances of major industrial firms successfully implementing automated customer service:

  • In 2018, a multinational investment bank launched a voice-enabled virtual assistant that helps customers manage finances. The assistant helps customers monitor their finances, resolve a wide variety of queries, and even track investments. The launch was a tremendous success, with the chatbot now boasting close to 1 billion interactions [3].
  • In the hospitality industry, a major American hotel company saved over $4 million using a virtual reservation assistant, which automated several steps in the reservation process, helping boost live agents’ efficiency and improve customer experience [4].
  • A messenger chatbot launched by a Dutch airline company has assisted over 500,000 customers with ticket bookings and query resolutions [5].


Question 3: Could you provide some practical examples of customer service automation in order to describe the benefits of this process better?

Sourav: There are various practical uses for AI in customer service, and they vary by industry. The following are a few common applications:

  1. Self service
    • Bots and interactive voice response (IVR) automations can resolve customer queries using natural language processing (NLP) and natural language understanding (NLU) on data from past cases, community forums, and in-house transactions.
    • Knowledge-graph–based intelligent searches can quickly direct customers to relevant content, thereby resolving issues without the need for live support.
    • Autocomplete speeds up customer query resolution by displaying matching issues based on historical data and suggesting resolutions in real time.
  2. Agent assist
    • Virtual agents can quickly collect relevant data from customers about an issue, which can help the human agent diagnose and resolve issues faster.
    • AI can be used to recommend solutions to customer issues by parsing through similar historical issues, thereby reducing time to resolution.
    • An additional benefit is the auto-population of entries for case closure, which boosts agent productivity and enhances data quality.
  3. Operational efficiency
    • AI is used to automate product returns/refunds by diagnosing the customer issue and predicting if it warrants a return. This leads to substantial increase in customer satisfaction.
    • NLP is very effective in categorizing cases and routing it to the agent who is best equipped to resolve the case.
    • AI is used to reduce case escalations by predicting emotional sentiment in customer interactions and detecting chances of a case escalating.


Question 4: We now understand how AI-integrated customer service is highly beneficial in improving company performance. What then are the steps to take to automate customer support?

Sourav: A successful customer service automation initiative requires multiple cross-functional teams to work together and bring their expertise to the program. Additionally, since data generally resides in siloes in organizations, any AI solution has to be founded on data from multiple cross-functional sources in order to provide an integrated experience to a customer. For instance, in a large tech company, collaboration across customer care, sales, engineering, and the supply chain is required to automate product returns seamlessly. Data from the customer relationship management (CRM) system, internal engineering systems, and community forums have to be combined with product return data from the supply chain to provide an end-to-end case automation experience to customers.

AI in customer experience usually involves a deep understanding of speech or text data generated from customer interactions. As a result, scaled domain-specific models are required for improved customer experience. While industry-agnostic models will get the program started, each and every industry has its own peculiarities and terminologies that the AI models will need to be well versed in. Eventually, it will be imperative to have scalable domain-specific models that have been trained to understand the complexities of the industry in order to address complex customer queries.

Another important aspect of successful customer service automation is shifting agent focus from cost reduction to greater customer satisfaction. While customer service will naturally become more efficient as bots take over simple support cases, human agents will then be required to handle more difficult and emotionally charged customer conversations. Therefore, in such a hybrid future, where human and virtual agents co-exist, the goal for human agents should shift from fulfilling efficiency-driven metrics to driving customer satisfaction.


Question 5: Which 3 communication channels would benefit the most from integrating AI tech in your opinion?

Sourav: While numerous companies have successfully integrated AI into their communication channels, it should be noted that customers do not interact with each channel separately; rather, they do so with the brand as a whole. The key differentiator in customer service experience, therefore, will be collecting the data generated from customer interactions across multiple channels and utilizing it to enhance AI systems for a truly omnichannel customer experience. Since existing technology can support this, AI can and should be integrated into all major communication channels. AI chatbots are already in widespread use across the internet, particularly in messaging applications like WhatsApp and Facebook Messenger, due to the numerous benefits offered. For instance, voice bots are being increasingly used across industries to personalize customer interactions. Moreover, social listening can detect customer issues and route them to the right team for resolution. Help/FAQ pages experience can be also improved significantly by automatically categorizing queries and assisting customers by autocompleting their search.


Question 6: Which key customer service metrics can be improved through the use of AI (and how)?

Sourav: AI-enabled customer service will lead to better customer satisfaction as well as lower customer service costs. The following customer service metrics are the ones that are usually affected by AI solutions:

  1. Average case resolution time
  2. Net CSAT score
  3. Agent satisfaction score
  4. First-contact resolution rate
  5. Case deflection rate
  6. Number of case escalation requests


However, some of the benefits of AI success in customer service can be difficult to quantify or measure in a straightforward manner. The following are a few instances of such cases:

  • The number of cases that were not counted because consumers were successful in using the self-help resources to resolve their problem
  • CSAT improvement when all customers may not provide explicit feedback

Due to these complexities, organizational alignment is required to define and track the appropriate KPIs to monitor the success of AI investments.


Question 7: What should one look for when selecting an automated customer support software for one’s organization?

Sourav: Deciding which software to use for customer support automation is highly specific to each organization’s unique requirements. Moreover, with cloud providers significantly expanding their capabilities, there is a compelling case to build custom software that is tailored to the organization's needs. Some of the key considerations that need to be made while evaluating customer support software are as follows:

  • How unique is a specific process in your organization's customer support workflow?
    Every customer support organization needs virtual agents and there are now plenty of such products available on the market. However, the case routing process may be unique to an organization; it is not unusual for multiple internal teams collaborating to resolve an issue. Consequently, a custom solution may be required, keeping in mind the unique nuances of the organizational workflow.
  • How unique is your domain?
    AI software in customer service often need to process customer conversations (text or speech). AI models trained on general conversations may not perform well in specialized domains like health care or networking technology. In all of these circumstances, the best course of action is to build specialized software in that domain or a bespoke solution suited to your needs.
  • What integrations are provided by the software?
    Integrating AI models into multiple communication channels as well as CRM systems used by the organization are key to achieving success.


Question 8: What are some common issues that you have noticed companies face while automating customer support? How do you think these issues can be prevented or resolved?

Sourav: Like any other feature that has the objective of enhancing and optimizing business activities, customer success automation comes with its fair share of challenges. Here are some of the more common ones, along with tips on how to mitigate them, should they arise.

1. Developing and sustaining AI systems can be expensive

Multiple steps are involved in real-time data analytics for customer service automation.

  • AI models must be trained on large volumes of data and supported by the appropriate infrastructure.
  • Data from disparate sources like CRM systems and community forums must be collected and stored in a data lake for it to be used.
  • A centralized customer service AI team needs to maintain all data, infrastructure, and models to ensure operating leverage across multiple AI initiatives.
  • Fragmented teams also tend to work on their own use cases,

All these steps result in extra expenses that render the AI initiatives unsustainable. The best approach to combat these costs when implementing an AI-based custom service management system is to use shared infrastructure with reusable language models and knowledge graphs.

2. AI has its limitations, which need to be augmented with human intervention

While AI models are becoming more and more efficient, they still have their limitations and will not be able to handle customer service edge cases. It is, therefore, imperative to avoid over-automating and to ensure that the process is designed in such a way that human agents can take over and minimize any negative customer experiences. Before launching AI systems to customers, thorough internal testing needs to be conducted wherever possible. For instance, conversational assistants can be tested internally before use to analyze dialog quality and solution relevance.

3. The KPIs used to measure the success of customer service AI initiatives are not aligned

Before launching an AI-enabled customer service program, it is crucial to define its objectives. The program design should include aligning the KPIs and collecting the appropriate data from AI systems in order to measure those KPIs. Often, there is an over-emphasis on the cost reduction of customer service operations. However, that should not be the only criterion for success; for instance, what if there is high sales growth in a product category but the number of customer issues has also increased, as was the case with ecommerce during the COVID-19 pandemic? Even though operational costs may not decrease in such a scenario, AI can still be the best tool to manage large customer volumes and provide a consistent support experience.


Question 9: Although fully automated customer service is still a ways off, many in the sector are concerned that AI may eventually take over all jobs. How likely do you think this to be the case? What do you think lies in the future for automated customer support?

Sourav: There is some validity to the worries that automation could replace human labor, not just in customer support but also across many other work sectors. However, in the foreseeable future, AI and humans will still need to work together to address customer issues. While AI will take over some of the more tedious, repetitive issue resolution tasks, which will free up humans for more complex problem-solving and customer relationship-building, AI does not yet possess the capabilities to establish an emotional connect with customers during conversations and handle all customer problems. It will be up to humans to bridge this gap and adapt to the ever-evolving demands of the customer. This also means that in a world where AI handles the repetitive tasks, humans will be required to improve their technical and soft skills in order to be able to provide the best service experience to customers. In fact, rather than taking up jobs, AI can actually be quite effective in helping humans become more efficient in resolving issues of greater complexity, by making suggestions and enabling users to search through large content bases.