It is important for us to maintain the quality of Bella’s operations and to quickly scale up based on the number of projects. Cloud powered by VMware is perfectly suited to addressing these tasks. I think that we have chosen a provider we can grow with.
Technical Director of Bellacall.me
Voice robots for calling a customer base
About the company
Bellacall.me automates business communication with customers. Our product is Bella, a smart telephone robot that is ideal for small, medium-sized and large businesses. Bella can perform 70% of call center tasks:
At the heart of Bella is our proprietary speech synthesis, recognition, and processing technology. The robot can start a conversation, maintain a lively dialog of any complexity and never interrupts the person on the other end of the call. While a human employee of the call center talks to one customer, Bella talks to 1500 at once.
The General Director of the Moscow Diagnostic Center approached us when he wanted to attract new customers. Previously, this job was handled by three call center operators who worked on it under a residual model. As a result, they hardly made more than 10 calls a day. At its first run, Bella screened out 15% of numbers as unavailable. Next, we ran the robot again for 2 hours and attracted 60 times more new customers than before.
Commenting on Bella’s work, Joseph Shley, the General Director of the Moscow Diagnostic Center, said:
“In two days, we managed to fully book up appointments with our specialists for two weeks ahead”.
Other cases are described on our website.
Problem: performance degradation and complexity of administration
We have been establishing call centers for medical institutions since the early 2000s. We worked mostly with people — we learned how to write quality call scripts, select a team, monitor customer experience, and find and correct communication gaps. With the development of technology, we realized that telephone operators can work less, but more effectively. We wanted them to deal only with non-standard situations involving customers where operators have to improvise. As for the rest of the routine work, we wanted to give it to machines.
Together with the development team, we automated the process of making appointments with doctors and voice messages for promotional offers. The next step was to make a robot that can understand human speech and react to responses.
The project started on the VDS of a Russian provider. We ran several virtual servers that were functionally separated:
When the load increased, the provider ran into the issue of sharing our resources with other customers, and the operation of services became unstable. In particular, this affected the storage subsystem. There were instances when productivity fell for several hours! Such downtime could reduce the number of leads in the following days, which meant losing up to 2% of monthly profit. Moreover, the provider refused to allocate additional computing resources. But quality of service plays a critical role when you are developing a new project on the market.
Our collaboration with Selectel began with renting dedicated servers. We needed a provider with servers that have optimal parameters at a good price. Positive customer feedback about Selectel gave us confidence in this company, and we decided to migrate. Selectel engineers provisioned the first server within a couple of days.
At first, we ran our virtual servers on dedicated Hyper-V physical machines. We eliminated regular performance degradation and, over time, experienced first-hand how Selectel supports a consistently high uptime.
If any issues arose, they were addressed 24/7 by technical support specialists.
We have been launching new projects on a regular basis. Now Bella could communicate with customers and bring in new leads without any failures. At that point, the number of projects began to increase and it became more difficult to manage a large number of physical servers. We felt that we had to deploy our new projects faster. Responding to our needs, Selectel specialists suggested their Cloud powered by VMware.
Each project in the cloud is a virtual data center with the same functionally-separated virtual servers, except for the router. It was successfully replaced by a VMware Edge router. This router is responsible for basic functions (firewall, address translation) and for organizing VPN networks between projects. So, we have a separate vDC running a central management server, which is connected to other projects with a star connection. This allows us to manage all projects at the application level.
Migration to the cloud provided a 4x increase in the speed of project deployment. Next, we started streamlining our internal processes. We wrote automation software using the vCloud Director® API and the Selectel API.
The Selectel API is used to create virtual data centers, while the vCloud Director® API automates the process further, from creation of virtual machines and network infrastructure to end router rules and application services at the business logic layer.
The software supports current configurations and workflows, making it easier to administer projects. Previously, we had to perform all kinds of service operations to configure peripheral equipment in terms of communications. For example, this included the already mentioned GSM gateways and virtual PBXs that we actively use in the telephony server. We manually reconfigured the equipment and the virtual PBXs. The automation significantly cut the time spent on such manual operations (about ten-fold).
It is important for us to maintain the quality of Bella’s operations and to quickly scale up.
The transition to the cloud and proprietary automation software, which uses the vCloud Director® API and the Selectel API, enabled us to reduce the time for launching new projects from 8 hours to 1 hour. Not only did the workload for launching the robot decrease; so did the cost of maintaining the infrastructure.
We were especially pleased to discover that Selectel has its own API for all services. In addition, we can order more services from a single panel. We rented a dedicated server for staging and use free DNS hosting, also with automation through the Selectel API.
Since we have sorted out the issues regarding the stability of the service and streamlined service processes, we now have the resources to further develop Bella in line with customer needs and market trends. Our immediate plans include improving the robot’s “intelligence” and raising the quality of speech recognition and algorithms for identifying the meaning of recognized answers. We also want to create a more versatile robot architecture and link it to Bellacall’s personal account.
I think that we have chosen a provider we can grow with.