Kaleyra
Providing Exemplary Customer Communication via AI-Powered Chatbots

CCustomer service expectations are evolving at an unprecedented rate, challenging companies to keep up. As modern-day consumers actively seek out information or contact customer service at a never-before-seen rate, providing prompt responses to their queries has become critical to maintaining good brand perception. This is where Kaleyra, a communications solutions provider, comes to the fore and provides real-time assistance to customers through their highly intuitive chatbots. The company helps businesses leverage conversational bots to automate customer interactions, which solves consumer problems in real-time, with fast, prompt, and effective responses, avoiding customer frustration. With their solutions, companies can effectively manage communication across many channels such as live chat, messenger apps, SMS and voice calls, to name a few. Kaleyra also brings an omnichannel business communication platform that enables its customers to reach, engage, and manage an integrated and multi-channel notification service. Their innovative communication tools help companies connect and communicate with the customer base effectively, and significantly increase customer retention and satisfaction. In a conversation with CIO Application’s editorial team, Kaleyra’s Chief Product Officer, Nicola Junior Vitto, discusses the company’s solutions and their efficacy at length.

Could you provide our readers a brief history of Kaleyra?

Kaleyra is a trusted Communication Platform as a Service (CPaaS). Our primary offering is an API-based platform, a single web interface that allows businesses to manage communications with their customers across various channels. Currently, we serve over 3,500 customers globally, including Hyundai, Uber, and MasterCard, to name some. Kaleyra was started back in 1999 as Ubiquity, in Milan, and the firm later merged with Solutions Infini, a Bangalore-based startup, in 2017. After this major acquisition, we rebranded the company as Kaleyra and we became a publicly-traded company in 2019 by getting listed on the NYSE. Last year, we had managed about 25 billion messages, and about four billion voice calls through our platform. In order to facilitate communication with customers, we collaborate with network operators in various countries. We work with more than 1600 mobile network operators worldwide that help us deliver our services efficiently and reliably.

Could you walk me through Kaleyra’s solutions?

Yes, sure. Kaleyra offers a single platform through which businesses can manage all their communication with their customers. We help companies engage their customers with personalized messages, chatbots, programmable voice services, push notifications, transactional emails, and other channels. Messaging and voice services are considered network-based channels, while every other service is IP or internet-based; we offer both kinds. Kaleyra has very robust APIs which allow developers to interconnect their systems easily with Kaleyra’s platform, and we also offer a web interface, an intuitive UI, where companies can simply create an account and log in to setup virtual business numbers, run marketing campaigns, build user communication flows, extract reports and such. As we cater to many organizations working in highly regulated sectors, especially in Italy, our primary differentiators are security, reliability and scalability. We offer authentication solutions to many financial institutions and help them validate and secure their users’ banking transactions. The solution enables businesses in other sectors as well to generate onetime passwords and send them to customers through SMS, voice calls, and email. We also provide chatbots that work across various channels. We also help companies implement actionable alerting systems, which enable them to send transactional notifications to customers. More importantly, we provide contact center services and improve companies’ operations efficiently.

What are some of the flaws of chatbots in the market space today, and how does Kaleyra address them?

Today, many companies in the market come with great capabilities in natural language processing, AI, and ML. However, to deploy such solutions, companies require extensive help from developers due to complications in the setup. Although many firms offer no-code or low-code solutions, which are easy to set up, such chatbots are rulebased and do not offer much in terms of NLP, AI and ML. For that reason, most bot offerings tend to work in one context and fail in another.

Moreover, these chatbots are not future-proofed, meaning they work with one or two communication channels today, but may not scale to other channels tomorrow. Kaleyra’s chatbots address all these pain points. For one: our solutions come with a natural language understanding engine, which we built leveraging our in-house capabilities. Our proprietary natural language engine also enables easy setup while customers try to build the chatbots. Clients should be able to set up a chatbot in a few minutes without any outside help. Two: our algorithm is capable of dealing with intent recognition efficiently. For example, our chatbots can differentiate between two sentences with the same number of words and tokens. We utilize the natural language understanding engine to recognize the intent of a sentence or a query raised by customer. We have built many features to help businesses improve performance by leveraging the data coming from chatbots. Smart detection is one such feature, which helps the bot ask only the necessary questions and avoids redundant questions, like if a user mentioned that they want a table for 7pm, the bot can avoid asking what time the user would like to reserve the table for. We primarily provide the service in English, but other languages, European, Indian, and others can be accomodated.

Could you walk me through the stepby- step approach that you take while building a chatbot?

The first step in building a chatbot is making sure it can understand the user intent. The purpose of a chatbot can be to fulfill any need, but to fulfill that purpose, the bot should have some idea of what a user may say to express their needs. Our algorithm looks for utterances that express this intent. For instance, when someone comes in and says, “please book a cab for me,” this statement indicates a particular intent. After understanding the intent, the chatbot can follow up the user’s query with questions required to fulfill the same. Such follow-up questions are generally entered into what we call slots. We make the slots smarter by leveraging AI, so the bot knows the right followup questions to ask. The final step in building a chatbot is fulfillment. Once the chatbot captures all the information from the user, it can perform the required actions, such as creating a cab booking or booking a doctor’s appointment, and respond with a confirmation message of some sort.

There can be many instances when the chatbot fails to answer the user’s query without the help of a service executive. For those instances, we use a human-plus-AI concept, which facilitates a seamless handover to an agent when the chatbot cannot fully recognize the user intent. The most common instance is when the customer asks for an agent—we can hand over the interaction to an agent swiftly. In other cases, we can also analyze the customer’s sentiment and hand over the conversation to an agent, especially in the case of the bot conversation taking a negative turn. The nice thing about the humans-plus-AI concept is that as the chatbots learn various from human interactions, they can handle customers who express the same intent in the future without the help of a human agent.

Could you narrate an instance that highlights the benefits brought to one of your clients after approaching Kaleyra?

Yes, of course. We completed a project with the government of India where we helped them effectively manage the COVID-19 pandemic through chatbot triage by helping the people book medical appointments in the event that they test positive. We built a chatbot that performed a preliminary screening of users which enabled the government to understand the patient’s medical condition before assigning any limited resources. The chatbot provided instant appointments to patients who were screened as being at high risk. Doctors reached out through video calls to provide medical care and expertise to patients in the early stages of the disease. We completed this project in just two days and launched the service in three different languages. With the help of the chatbot, the government screened more than 2,500 patients who tested positive for COVID-19.

In another instance, we collaborated with the government of Karnataka—one of the southern Indian states—to help them manage the lockdown imposed during the initial waves of the COVID-19 pandemic. We partnered with last-mile delivery services operating in the state to provide essential goods to people. We built a chatbot on WhatsApp where users could place orders for essential groceries and medicines. The bot would pass the orders over to the delivery partners via APIs, and they would deliver these essential groceries at the user’s doorsteps. The bot was intended for delivery of essential goods only, so we leveraged AI to screen the orders for prohibited items such as alcohol and cigarettes. We helped more than 25,000 customers just within Bangalore city through over 55,000 chat sessions.

What does the future hold for Kaleyra?

It’s pretty obvious that chatbots can be used in a number of different applications to interface with customers. One application that we see in particular is the use of bots to solve the issue of fragmented customer interactions across channels. Part of this is employing bots to obtain the correct information from the users based on intent, but bots could also be used to surface the correct information in front of support agents to provide them with better context about the customer, which in turn would help them provide the customer with a better and more seamless experience. This has a lot to do with improving organizational processes with AI. The future of conversational interfaces will be to address some of the fundamental flaws in such processes. In the future, chatbots will emulate the material world through a variety of virtual interfaces, not just text like we mostly see now. In this regard, we can expect to see the market evolve by combining AL and ML capabilities with a variety of channels, including voice, which a few players have already started experimenting with. We believe that conversational AI will become integral to business processes in the days to come, and we’ll augment our solutions with new, innovative capabilities to help our clients communicate better with their clients.