Discuss project
By clicking the button, you consent to the processing of personal data and agree to the privacy policy.
SHOWCASE
The virtual interlocutor answers questions about the profitability of programs, insurance cases, and the status of applications round the clock
A chatbot for AlfaStrakhovanie-Life's personal workplace
AlfaStrakhovanie-Life is a specialized life insurance company of AlfaStrakhovanie Group. It is among the TOP 3 life insurers in Russia in terms of fees. AlfaStrakhovanie-Life works in three main areas: investment life insurance, life savings insurance, and credit life insurance.
The task is to develop and launch the chatbot for AlfaStrakhovanie-Life's personal workplace.
Context

AlfaStrakhovanie-Life has several online customer contact points: electronic forms on the website and two forms of feedback in it a personal workplace.

To find out information about an application, the client submits an appeal and is waiting for a reply. A specialist is looking for information to answer in systems, databases and excel-files.

Before connecting the chatbot, clients asked questions about navigation in the personal cabinet through the feedback forms.
The purpose of creating a bot

Switch customer service to 24-hour mode. Take the load off the support team.


Additional requirements

Integrate service with company data sources. Connect an additional contact point for users — a bot in Telegram.
Implementation
1
Together with the client service team, we made a list of common questions from users.
2
We formed a bot knowledge base using the Dialogflow service. The service allows to recognize the natural speech of the user and select the appropriate response from the database.
3
We connected to several AlfaStrakhovanie-Life data sources. This allowed to obtain information on contracts and customer requests.
4
Set up an operator workplace for members of the support team. The client can start communicating with a specialist at any time. The specialist has access to correspondence between the bot and the client.
5
After building the project, we added the administrator web-interface, where specialists can view dialogues with users.
What can the chatbot do
Knowledge base
Menu
Document on request
Specification
Operator
Difficult questions
Telegram version
The knowledge base contains about 200 questions and answers pairs. For each pair, there are from 2 to 30 variations of question-wording.

The database is hierarchical, the bot can ask clarifying questions. The bot asks based on client information, which means there will be a minimum of questions.

The hierarchical menu helps to get a response from the bot in a few clicks.

The menu can be invoked using a command.

If the bot cannot recognize the user's question, it suggests using the menu itself.
The bot responds to the most popular user requests, such as where to see the yield.

If the user has one policy, the bot immediately gives a link to the yield report.
If the user has more than one insurance policy, the bot will ask clarifying questions and then provide a link.

We have programmed about 30 such dialogues using internal client databases.
The client can switch the dialog to an operator at any time.

For the user, this is a seamless transition to and from the operator.

The specialist will have access not only to the history of correspondence with the bot but also to data about the client from internal systems.
If the question involves operator assistance, the bot will suggest switching.


The dialogue with the bot can be continued in the smartphone, we integrated it into Telegram Messenger.
Results
Customers receive information at any time, regardless of the availability of specialists. This is especially important as AlfaStrakhovanie-Life customers are located in different time zones.

Specialists have access to the administrator web-interface. They can read correspondence with clients, coordinate the addition of new questions, generate reports with different parameters.
Statistics for the past year
900
users of the personal workplace contact the bot monthly
24%
of user requests came after hours
4%
of sessions with a request to switch to an operator
90%
of answers in the categories "Profitability" and "Survival" were assessed by specialists as correct
The number of new questions identified during the month decreased from 6% to 0,5%

The number of question versions identified for the month decreased from 16% to 6,5%

50% of the sessions involve no more than 5 issues
70% of dialogs consist of 10 or fewer messages (including service and welcome phrases)

37% of dialogs last no more than 5 minutes
Post-project service
Daily analysis of client correspondence with the chatbot
Study of the relevance of issues in the knowledge base.
Monitoring the correctness of the bot’s answers and identify new questions.
Remove questions that have not been used for several months.
Coordination of new questions for bot with the client