Evolved switchboard: managing customers and requests with voice bots

Automated main call handling systems to a switchboard enable support and decongestion of URPs and internal switchboards. This solution is widely used in health care facilities. The evolved switchboard recognizes requests made by users and sorts them according to predefined procedures, relieving human operators and enabling reduced waiting times.

How does the evolved switchboard work? Voice bots and speech recognition

A user calls the switchboard number of a health care facility, the voice bot answers and handles the interaction using a speech recognition system. The evolved switchboard interprets the reason for the call, performs an intent check, and Routes the phone call to the department it believes can answer it correctly to the user’s request. Thus, there is a transfer of the call to the CUP or another extension in the health facility, and then there is a switch to contacting an operator.

The system is therefore able to Recognize user requests and handle them through predetermined procedures. The main case histories have been mapped, and in this way, when a question is asked to the voice bot, itidentifies the underlying message through the speech recognition software, extracts the text from the voice, and performs a match with the set case histories, then executes the corresponding expected procedure.

A bot refines its choice not only through the identification of a keyword but also through probabilistic calculations based on the relevance of multiple related keywords. This algorithm provides a “confidence” score at the detection stage to increase the probability of finding the right answer for the question asked.

The automatic switchboard is able to transform the voice into text, analyzes it, checks its intentionality, and based on the words spoken, evaluates its actual meaning. For example, a user calls to find out how his wife who has been admitted to the emergency room is doing. The bot transposes the most relevant keywords, performs disambiguation, recognizes the weight of the words spoken and identifies the main one in “emergency room,” connecting the interlocutor with that department.

Natural language recognition will allow the user to express himself directly as if he were conversing with an operator, and an intent database will allow recognition of multiple requests, leaving only those not yet present to human back-up.

The strengths and goals of the evolved switchboard

• Reduced waiting time. When you call the switchboard, you immediately talk to the voice bot, and you do not run into unproductive queues characterized by the usual musing. The automatic switchboard allows for the giving and receiving of information, thus saving user and staff time.
• The evolved switchboard aims to increase user satisfaction, directing them quickly, flexibly and practically to the best contact to receive the most relevant answers to their inquiries.
• The goal is to independently absorb at least 70 percent of calls. A human operator can also be had on the sidelines to handle any peculiarities and requests that fall outside the canonical case histories.
• Bots are constantly evolving through optimization work. The various requests received are checked, and on those that were not recognized, an analysis is performed, thinking about how a human operator would have responded and calibrating the robot to these metrics.