CX Expert Talks

Anand Elangovan: Research confirms that chatbots will be involved in 85% of business and customer interactions. Another prediction states that the global impact of AI will result in $1.1 trillion in business revenues by 2021. 

Anand Elangovan: This chart is a Gartner hype cycle, and it illustrates the stages that a trending technology or software will go through:

• Technology Trigger

• Peak of Inflated Expectations

• Trough of Disillusionment

• Slope of Enlightenment

• Plateau of Productivity

Chatbots have already scaled the peak of expectations and crossed the trough of disillusionment. Now, they’re on the slope of enlightenment – businesses are rapidly seeing value in chatbots, and adopting them to simplify drudge work. They don’t replace jobs, but only eliminate and automate mundane tasks. Think about the time when ATMs first surfaced. People thought bank tellers would be out of work. But that’s not what happened. Tellers do exist today – they are now people who are capable of doing more than they were initially capable of. The moral of this anecdote? Humans are creative beings, and machines should be doing the mundane work. Post-2021, chatbots are expected to become a part of everyday life, answering every repetitive question that customers might have. Ultimately, this will enable customer service professionals to solve complex customer problems without being bogged down with menial tasks. Bottom line: I believe the hype around chatbots today is justified.

Anand Elangovan: There are three primary areas where chatbots deliver maximum impact:

• Leveraging automation: Businesses have a lot of different software used across business units, and chatbots can help tie them together by automating tasks. For instance, a bot that’s integrated with a billing system could simplify a support agent’s job by automatically fetching a customer’s bill from the billing software and presenting it to the agent within the chat window.

• Proactively engaging with customers: Bots can automate periodic outreach and instant replies that human agents may not be able to deliver on time, every time.

• Accelerating customer support: Bots are key to creating smooth, fast experiences for end-users. For instance, a bot could display relevant self-service articles based on what customers type, saving them several minutes that would otherwise have been spent on browsing the website to look for the right article.

We’re going to look at how all this is done, and why chatbots are so popular today. Customers prefer messaging over other modes of communication. Nobody wants to speak to an agent who doesn’t have the information they need to answer queries. Also, emails as a customer service channel are getting outdated. It’s only natural that messaging platforms like WhatsApp and Apple Business Chat will take over.

Anand Elangovan: How do chatbots actually improve CX? When the customer comes to you with a query, it’s probably not a new question. Consider it an 80-20 rule: 80% of the questions are repetitive and have already been answered by your agents in the past. Chatbots paired with a comprehensive knowledge base can pull articles from the said knowledge base to quickly provide answers to repetitive questions. Can you go deeper into the concept, Shubam?
Shubam Goyal: First, I'd like to say that there are two types of chatbots. One, the decision-flow-based bot. The other is the NLP-based bot. 

• Decision flow-based bots: There is a misconception that this is a static bot and that you need to predefine every conversation route. That is incorrect. Users will be given conversation options to select from, and there are certain rules that can produce the desired output. But, the options presented to the customer are dynamic and can draw insights from previous customer transactions.
• NLP-based bot: You don’t necessarily give closed or controlled input where users can select from a list of options. Here, the user can type in their query and the bot will try to understand what the user is typing, detect their intent, and give recommendations. This recommendation will be from your knowledge base, solution articles, or information the bot has collected from previous interactions. I believe in supervised learning: If a bot is not able to answer something correctly, the user will be able to give feedback that the bot can learn from.

Anand Elangovan: Which bot is better – NLP or decision-flow?

Shubam Goyal: It depends on the business case. It should serve the purpose for which the customer has come to the bot. Either bot should not be too open-ended. For example, an open-ended bot might result in a customer approaching the bot with: “Hey, can you help me fly a kite?” Here’s a good example of a well-configured bot: if you visit an eCommerce website, the bot asks you: “Hey, I see that you have placed an order, what would you like to do?”. It presents three options: Cancel, return, or modify the order. This is a decision flow-based bot. On the contrary, it could just have a text box that greets the customer and asks them to type their query, which is an NLP-based bot. If the question is beyond the limit of the bot, there should be a fallback – that is, passing the chat to an agent.

Chatbots can be used not only by end customers, but by agents as well. They allow agents to deliver faster resolutions. For instance, an agent could type a customer query into an NLP-based bot, which could prompt the bot to quickly bring up the related FAQ articles that the agent can then send to his/her customer. If they had to search for the article themselves, that would have taken quite a while, and resulted in customer frustration or drop-offs.

Navigate to 21:10 in the recorded webinar for a demonstration of:

• An NLP-based chatbot

• A decision flow chatbot

• An agent-assist chatbot

Shubam Goyal: I’ve picked two use cases to explore. As a business, your product has two sides. One is the pre-sales side. The other is on the post-sales side. 

• Pre-sales bots: Sales-side bots can be proactive and personalized while reaching out to customers. For example:

  • “Looks like you’re looking for this product, let me help you out with a recommendation” 

  • “Here are some key features of this particular product”

  • “Why don’t you give me your number so that our experts can get back to you on this?” 

Bots can also help with gathering information, making reservations, or placing an order. 

How does this help the customer? Customers want conversational guidance. At every step, you ought to be making use of some past information based on an interaction with the customer. The bot can say something like, “Hey, looks like you bought this desk a few months ago. Why don’t you buy a table lamp along with the chair you’re buying now?” This sales process can be automated with the help of a bot.

• Post-sales bots: Let’s take a look at support-side bots as well. For example, a customer might want to return an order. With a bot, no manual intervention is required to process the request. If you have the chatbot connected to your order management software via APIs, the entire return process can be automated. On the other hand, if they have queries related to your product, and those queries are FAQs, the bot can provide an immediate response by pulling information from the knowledge base. A human agent would certainly take more time to manually execute these tasks. This way, your human agents can work on more complex and high-value tasks, while the bot fields mundane ones.

Shubham Goyal: Let’s talk about proactive engagement. For example, if the customer is idling on a page, a bot can pop up and ask if they need help. Then, if a customer has a query, a well-trained bot can answer that query from its knowledge base – that’s deflection. Here, we might see cases where the bot is unable to answer all the queries. At some point, manual intervention will be required. 
There are many use-cases where you would want your experts to take over. Rather than transferring this to a random agent, the bot will understand the nature of the query, gather all the information necessary, and intelligently route it to the right team. This gives you a seamless and quick resolution.

Anand Elangovan: Let’s talk about the return on investment of a chatbot. 

• Sales Opportunity Generation: Let’s say a B2B business wants to increase its revenue using a chatbot. Using proactive customer engagement, if they can convert a casual website visitor into a prospect who provides their contact details, the business has added a sales opportunity to their pipeline. The sales bots outlined in the previous section are a great example of this, especially if they suggest products to visitors based on their order history.
• Agent Effort Reduction: Agent assistance via automation is something we’ve already discussed, and that provides great RoI, especially if it eliminates manual effort for either the customer or the agent.
• Cost Optimization for 24x7 Support: Another big benefit is reduced costs. For instance, think about round-the-clock support. Agent involvement can be greatly reduced during non-working hours by using bots as deflecting tools, allowing the agents to be more productive during their daytime work hours.

We recently ran a customer survey to measure the effectiveness of a chatbot, and the biggest benefits we uncovered were coverage, accuracy, and deflection.

 

1) Can a chatbot pick up keywords, create tickets, and assign them accordingly? 

Shubam Goyal: Yes, it can route tickets based on the keywords it picks up from analyzing the customer query.

 

2) Can you configure the bot to answer specific questions?

Shubham Goyal: Yes. Also, we can train the bot to follow certain steps to answer the question. Or, we can use the knowledge base to pull answers from there, if you already have one.

 

3)What other languages does the Freshdesk bot support other than English? 

Shubham Goyal: For the decision flow-based bot, we support 36 different languages. For the NLP-based bot, we support English, but are building support for French, German, and Dutch. 

 

4) How compliant are your chatbots with GDPR?

Shubam Goyal: All our products, not just our chatbots, are GDPR compliant.

 

5) Is a chatbot capable of getting user information from pre-saved data? 

Shubham Goyal: Yes, the chatbot will be able to do that. So if someone has entered information once, they will not be asked to do so again. Also, if the user is already logged into your website, the bot will auto-detect their sign-in information.

 

6) In the RoI section, you mentioned deflection. Please define the term.

Anand Elangovan: When someone interacts with a bot, gets their answer, and leaves, that’s called a deflection. If the bot hadn’t been there, this query would have been presented in the form of a call or an email. The bot ‘deflected’ it away from an agent.