Task
In 2020, due to the coronavirus pandemic, the load on the Helix Laboratory Service call center almost tripled and the operators physically could not respond to all requests.
The subsidiary IT company Medindex on behalf of the Helix company turned to Metabot for a chatbot for the web portal https://helix.ru with a focus on automated communication with customers.
The company provided more than 10 interaction scenarios, which they wanted to switch from the call center to the chatbot in the first place, as well as functional and non-functional requirements for the solution, including:
- Integrate the chatbot with the omnichannel platform Webim.
- Integrate the chatbot with the Helix web portal API for checking order status, sending PCR test references by email, etc.
- Provide a chatbot as a containerized application Kubernetes (K8s).
Result
The Metabot team was able to automate popular routine requests, relieve call center operators, and improve service. We also managed to satisfy the functional and non-functional requirements of the customer.
The chatbot was launched in the summer of 2021 in conjunction with the call center, which was in dire need of automation, and the marketing department, which participated in building the sales communication in the chatbot.
Immediately after the launch, the call center operators noted that the load during the day fell 5-6 times, and the processing of requests was reduced from 15-30 minutes to 5-10 minutes.
With the help of the chatbot, Helix employees were able to automate not only the provision of answers to frequently asked questions, but also checking order statuses based on customer data and information about tests online without human assistance.
“The load on operators when the chatbot is connected drops significantly – from 5-12 simultaneous dialogues per person to 1-2”.
Julia Bychenko, Head of Contact Center and Customer Service.
Every day the chatbot processes more than 400 dialogues on average, of which 70% are processed by the chatbot automatically. The other 30% are requests related to the complex specificity of the questions asked and the constant need to further train the NLP model with new requests. At peak times, the chatbot processed more than 2000 dialogues per day.
During the second half of 2021, the chatbot was improved by the team of Medindex and Helix specialists, who underwent training at our private webinars and mastered development with the help of Metabot and Dialogflow. New information sections were added and speech recognition was improved. Thus, contact center operators received a useful virtual assistant in the existing active communication channel.
You can interact with the Helix chatbot right now by visiting the company’s web portal and clicking on the small chat icon in the lower right corner of the screen: helix.ru
Emotions of the Performer
Together with the Helix and Medindex teams, we developed a clear scenario for providing information about tests, analyzing the best cases in this niche, and developing our own solution; both teams of IT specialists were involved in the development of API integration points: from Medindex and Metabot. As a result, we collectively created a unique product that conveys the corporate spirit of the company, selected understandable symbols for scenarios used by chatbot designers, and to this day, this project has allowed us all to gain experience and keep pace with digital communicative development.
Leading chatbot designer-architect at Metabot & Co. – Alyona Rvacheva
Tools
The chatbot used a standard support chatbot configuration: a self-service menu with natural language recognition. In addition, unique API points were integrated:
- Checking the user’s city and based on that, providing information about orders that can be fulfilled in a specific city (brief info, price, availability of types of delivery).
- Checking the readiness of results by the order number and the user’s full name, if the order is ready, automatic sending of the results to the user’s mail using an SMS gateway for number verification.
- The possibility to cancel/transfer appointments for tests by the phone number and the user’s full name.
- Providing a link to tests from the site’s knowledge base with detailed information after entering the test code.
The Dialogflow platform from Google was used as the technology for recognizing user requests in free form (Natural Language Processing \ NLP), for which there is ready-made integration on the Metabot platform. NLP was used to transition between scenarios and provide answers to frequently asked questions. There are about 100 intents in the bot, based on real client requests.
Almost all scenarios at the end of the dialogue can be transferred to a call center operator or another scenario. Before transferring to an operator, the chatbot automatically indicates the topic of the request and asks for users’ contact information: phone/email. The call center works in conjunction with Metabot + Webim.
For more detailed information about the functionality, you can study the document: Overview of the Helix Chatbot.
The image below shows part of the visual Miro scheme for the Helix chatbot, which clearly demonstrates the complexity of the project.
