Since our client was planning to implement a platform for organizing service in text channels anyway, the ability to simultaneously conduct written dialogues from a single interface and analyze recordings of telephone conversations seemed interesting to him. It was decided to connect the AI Contact Center in the beta version, without a fee for connection and use of the platform, only for Call Recording and Call Transcription, as it was before, plus text processing by a neural network. can be performed either based on ready-made instructions from the solution provider or on custom instructions for AI compiled by the client. For the tasks of identifying problematic dialogues and low operator competence, our prepared instructions with customer complaints and dissatisfaction, repeated calls, comments to operators, demands to involve management in resolving the issue, and the operator’s inability to help resolve the problem turned out to be sufficient - a total of more than 1,500 words and expressions.
Based on these instructions, the neural argentina cell phone number list network processes recorded and transcribed conversations. If the conversation contains words and phrases related to a specific instruction, the AI automatically classifies the dialogue as “customer dissatisfaction,” “repeat appeal,” and so on. Now the supervisor can filter operator dialogues, viewing only those that require his attention. In addition, the employees of the now contact center began to ask subscribers personally whether they were able to solve the problem, whether they still had questions, and how the clients rate the solution to the issue on a scale of 1 to 5. The neural network was instructed to save the stage of receiving feedback from the client as a short description. This way, the supervisor does not need to study the entire transcript of the conversation to find out the service quality assessment.
Finding a solution Initially, a special database was maintained for customer service, where all news from the freight industry, updates to ship and train schedules, changes in fees, duties and tariffs were loaded. Of course, using such a database was much more convenient than searching for all the information yourself, but the search was manual and took a lot of time, and clients wanted to receive answers to their requests as quickly as possible. The integration of a shipping cost calculator and the sending of changes to tariffs in the client's personal account on the website only partially relieved the support service. It was decided to try an AI-based chatbot so that it could access the information storage, extract the required data from there and process it according to the request.