RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2
Project

RT Labs received the mega-contract of Sberbank for creation of voice identification of clients

Customers: Sberbank of the Russian Federation

Moscow; Financial services, investments and audit

Contractors: RT Labs
Product: CST: VoiceKey.IVR
На базе: CST: VoiceKey.Platform
Second product: CST: VoiceKey.Platform
Third product: CST: VoiceKey.Agent

Project date: 2017/06
Project's budget: 301.4 million руб.

Content

Contractor Selection

In June, 2017 according to the results of consideration of bids Sberbank selected RT Labs the supplier of the solution of voice biometrics for the Uniform Distribution Contact Center (UDCC).

Total amount of agreements with the company will be about 256 million rubles, follows from the published protocol. From this amount about 236 million rubles are the share of licenses softwares, developed by CST company: Voicekey.Platform, Voicekey.Agent, Voicekey.Fraud and Voicekey.IVR.

Sberbank reserves the right further to sign with the winner the agreement on support of the purchased software for two years of 86.7 million rubles, it is specified in the protocol.

The contact center of Sberbank will identify clients on a voice

Sberbank announced competition at the choice of the supplier of the solution of authentication of clients in the Uniform Distribution Contact Center (UDCC) using technologies of voice biometrics in February, 2017[1]. The starting price of the contract was 301.4 million rubles.

In addition to RT Labs the companies also applied for the contract with Sberbank:

Technologies and the expected effect

The solution of CST implemented in Sberbank is based on technology of deep neural networks.

Sberbank expects that implementation of the additional method of authentication in contact center should optimize labor costs of employees at determination of clients, reduce the number of the provided data during this procedure, reduce time of processing of a call and minimize fraudulent risks.

Work with the system of voice biometrics provides several scenarios. So, client authentication can happen at self-service in the IVR system. In this case a system suggests the client to say a phrase, for example "My voice – my password".

Additional biometric verification of the authenticated client can be applied in the course of the conversation with the operator. If the client contacted the operator without preliminary authentication, additional verification of a voice in the course of the conversation can serve as client authentication if the system of voice biometrics displays high reliability of a voice. In case of low reliability the operator applies additional methods of client authentication.

One more scenario is a check according to the black list of a voice of the client in the course of work with the operator.

For primary accumulation of biometric voice molds the possibility of use of the existing records of the negotiations of the client with the operator (marked as undergone strict authentication) or removal of a voice mold directly in the course of the conversation with the operator is provided.

The existing methods of authentication do not stop the work for cases when the system of voice biometrics is not able to recognize with guarantee the client because of changes in a voice or bad quality of communication channels. Biometric authentication and verification is entered as the additional security level. Further, after assessment of results of system operation of voice biometrics and level of recognition, the decision on failure from old methods of authentication of clients can be made.

As of February, 2017 the personalized service in Uniform contact center is carried out only after successful client authentication. At the same time 3 options of authentication are available to the client: using the specialist of CC (the specialist asks the client questions, verifies the received answers with data of the assorted client of bank, and depending on result, the client is considered authenticated or is not present), using the automated Code of the Client service (or T PIN) or control information, available in IVR ERKTS. Depending on the client's question/problem, even in case of his successful authentication using Code of the Client service, the specialist can request the additional information for gain of the authentication scheme. At the same time authentication is required more than in 60% of the calls coming to ERKTs; authentication duration through the specialist averages 40 seconds.

The main problems which the client at authentication faces it is unavailability in the mode of dialog to provide this or that information (the card number, a control word according to the card, the passport number, the address on a registration provided in bank when obtaining the map).

As a result of project implementation process of biometric authentication which substantially will reduce costs of the ERKTs resources and will increase for clients convenience of interaction with contact center should be implemented and also will reduce risk of fraudulent activity.

Implementation of a new system will in addition allow the staff of department of work with fraud to identify malefactors not only on the questions registered in the corresponding scripts but also using technologies of determination of a voice. Thus, if the client has a voice mold, then the swindler will not be able to obtain this or that information on the client. The functionality of "black list" will not allow swindlers to obtain information if the voice mold of the swindler is in emergency.

Notes