Verloop.io: Building an intelligent chatbot that enables end-to-end support in multiple languages
About Verloop.io
Verloop.io is an AI-powered customer support platform that enables global customers in the ecommerce, real estate, and banking sectors to deliver instantly scalable, 24-hour support across channels in the language of their choice. The platform enables end-to-end automated support in 92% of cases via AI algorithms built on specialist knowledge of each segment. Founded in 2015, the startup has expanded across India, Southeast Asia, and the Middle East and has plans to enter the US. It has been named one of India's 25 best SaaS companies and won Gartner's 2019 Cool Vendor Award for conversational marketing.
Tell us your challenge. We're here to help.
Contact usVerloop.io provides 24/7 fully automated online customer support for 15,000 companies and counting in more than 108 countries in more than 10+ languages such Arabic, English, Bahasa, etc., with multiple dialects, with an AI chatbot developed on Google Cloud.
Google Cloud results
- Guarantees seamless global customer support with sub-millisecond latency
- Supports end-to-end service by robots in 92% of cases with Google Cloud infrastructure, AI, and storage tools
- Enables instant delivery of AI model updates to client platforms with Cloud SQL and Cloud CDN
Enables up to 10x spike in customer query load
When Gaurav Singh founded Verloop.io in 2015, India's ecommerce firms were awakening to the possibility of full automation of online customer support and launching big teams to create the ideal AI chatbot. Gaurav saw an opportunity to create the chatbot for them, freeing them up to focus resources on core operations.
Gaurav's unique insight was to develop a solution for a small number of specific sectors instead of trying to create a generic platform for all industries. By limiting focus to ecommerce, banking, and real estate, Verloop.io would gain an end-to-end picture of the customer journey in each field. It believed it could thus minimize chat deflection, or the point the AI gives up and hands over a request to a human agent.
To succeed, Verloop.io sought powerful cloud infrastructure and world-class AI modeling tools to drive the startup's proprietary machine learning and natural language processing (NLP) algorithms. It turned to Google Cloud and found the tools it needed to invent a customer support automation platform that today enables end-to-end robotic support in 92% of cases.
Google Kubernetes Engine, AI Platform, and data storage solutions such as Cloud SQL and Memorystore combine to help Verloop.io guarantee 24-hour automated support to more than 15,000 companies in more than 108 countries.
"We started with a dream to build the ideal automated chat experience enabling global companies to provide customers delightful support without speaking to a human agent," says Gaurav. "Google Kubernetes Engine and Google AI and data storage tools give us the scaling power, low-latency, and creative freedom to turn our vision into reality."
"We started with a dream to build the ideal automated chat experience enabling global companies to provide customers delightful support without speaking to a human agent. Google Kubernetes Engine and Google AI and data storage tools give us the scaling power, low-latency, and creative freedom to turn our vision into reality."
—Gaurav Singh, CEO, Verloop.ioScalable, round-the-clock customer support automation with Google Kubernetes Engine
Verloop.io was launched on a different cloud provider using a Kubernetes-based architecture. It soon found the provider's orchestration support for Kubernetes to be too limited to guarantee the scalable computing power it needed to develop and run AI algorithms for global clients around the clock.
The startup moved to Google Cloud and quickly discovered it was able to meet customer expectations of uninterrupted, scalable service on infrastructure hosted on Google Kubernetes Engine (GKE). Deploying GKE also freed up Gaurav's team from mundane infrastructure maintenance to focus on the creative mission of developing AI solutions for customers.
"Google Cloud's native integration as the developer of Kubernetes helped assure our clients that we can scale horizontally and reliably according to their load distribution challenges," says Gaurav.
He says that since the beginning of the year, customer support queries have soared six- to ten-fold and the Verloop.io engineering team never even noticed the spike in load because it was all being scaled automatically by GKE.
Meanwhile, Google data storage solutions such as Cloud SQL and Memorystore combine with Cloud CDN to enable the sub-millisecond latency delivery Verloop.io relies on to provide seamless customer support automation around the world. Powerful, instant-access data storage also enables the Verloop.io development team to constantly update its AI models and serve them instantly to client platforms. Google Cloud enables it to give its clients control and flexibility over machine learning model updates too. During the early stages of client adoption, Verloop.io's clients update models with new training data more than 20 times a day with the confidence that these training updates will work as expected. This is due to the reliable GKE infrastructure, which can adjust variable loads and training frequency.
"Having large volumes of storage available at the blink of an eye lets us transit between training models and upgrade them seamlessly to provide cutting-edge ML services for our customer," says Gaurav.
The result is a multilingual automated chatbot that increases the rate of end-to-end customer support automation without human intervention from 60% to 90%. Verloop.io's industry-specific chatbot yields a 50% improvement in completing customer queries compared to generic solutions designed to handle requests from any industry.
Fixing problems before they happen
In any global support solution that lives or dies on the reliability of its 24/7 scalable support, it's critical that potential bugs and glitches, especially during service updates, are quickly identified and resolved.
Verloop.io depends on Cloud Logging and Cloud Monitoring as a behind-the-scenes helper that enables customers to offer uninterrupted customer support even under conditions of major system upgrades and sudden spikes in demand.
"Google Cloud's operations suite for logging and monitoring deserves special mention," says Gaurav. "Having an integrated logging, monitoring, and alerting system allows our engineers to trace issues across the microservice landscape and identify and resolve problems quickly."
Cloud Build, meanwhile, ensures the reliability of Verloop.io's product release pipelines by allowing developers to keep releasing bug fixes and new features at high speed.
Cloud Build allows Verloop.io’s developers to keep releasing bug fixes and new features at high velocity, producing releases as and when new code is pushed to GitHub. The horizontal scaling support from Kubernetes enables functional teams like machine learning to focus on the challenges which are closer to the customer, while scaling is managed by GKE. For instance, Verloop.io saw peak loads of six to 10 times its usual load over two quarters, but the ML team never had to worry about it as this was already taken care of by Google Cloud.
"Google Cloud's native integration as the developer of Kubernetes helped assure our clients that we can scale horizontally and reliably according to their load distribution challenges."
—Gaurav Singh, CEO, Verloop.ioPowerful AI natural language tools to invent the future of online customer support
Beyond end-to-end automation, Verloop.io also needed its chatbot to speak multiple languages, since it's mission is to serve both a global market and the diverse and multilingual communities in India. It developed proprietary natural language processing (NLP) algorithms for more than 10+ languages such Arabic, English, Bahasa, etc., with multiple dialects on GKE infrastructure.
Verloop.io's favorite implementation has been AI Platform for ML Ops use cases. It uses a wide variety of Google technologies within its NLP stack as well, including TensorFlow Hub and the AI Platform. “More than 80% of our natural language queries are answered using transformer-based models, a key contribution from Google Research,” says Gaurav. “The Google Cloud AI Platform enables seamless integration and makes our AI pipelines robust, scalable, and easier to maintain.”
Verloop.io now plans to take its AI solution a level higher by incorporating speech-enabled automated support into its chatbot. To achieve that goal, it is exploring Google AI features such as Speech-to-Text to power the next phase of its evolution.
Tell us your challenge. We're here to help.
Contact us
About Verloop.io
Verloop.io is an AI-powered customer support platform that enables global customers in the ecommerce, real estate, and banking sectors to deliver instantly scalable, 24-hour support across channels in the language of their choice. The platform enables end-to-end automated support in 92% of cases via AI algorithms built on specialist knowledge of each segment. Founded in 2015, the startup has expanded across India, Southeast Asia, and the Middle East and has plans to enter the US. It has been named one of India's 25 best SaaS companies and won Gartner's 2019 Cool Vendor Award for conversational marketing.