Hospital sector case study: how Reception AI improved switchboard performance

A healthcare client had the need to improve the performance of their call center by reducing missed calls, wait times, and optimizing call transfers to internal departments. With the help of AI Reception service, INGO was able to significantly enhance the outcomes.

Who is the customer

A major Italian hospital company, a benchmark in the healthcare industry for decades.

What were the client’s needs

The first intention was to improve the performance of the switchboard , decreasing the amount of missed calls, reducing waits before answering, and optimizing the transfer of calls to internal departments.

Within a hospital setting, in fact, the switchboard is a real gateway to which requests of various kinds are addressed: it is important to direct them to the most appropriate figures to provide the most accurate answers possible.

Thus, the client’s need was to retain the typical switchboard’s centralized role, while reducing wait times and improving the user experience, while seeking economic savings compared to the previous management done exclusively with operators.

What was accomplished

INGO, building on decades of experience with the service, conducted in-depth analyses of the knowledge base already established for the active service, dwelling on the various management case histories. This knowledge base was used to train the AI (artificial intelligence) reception service and make sure that it could start its operation reliably.

During the first day of operation, the voicebot autonomously absorbed 56% of requests, i.e., more than 5 out of 10 calls were completely answered by the automated switchboard without operator intervention. More than half of the traffic volume was transferred regularly and correctly under this new system, no longer burdening operators.

Bot trainers, data scientists, and data analysts constantly analyzed the various interactions so as to understand how requests were coming in and how they were being made by users, so that bot performance could be improved. Refinement was gradual over the first few weeks, and already after the first month of the service 80 percent of traffic was being handled automatically.

The benefits to the client and the results achieved

The adoption of an automated switchboard has not disrupted internal procedures, and service has improved because waiting times have been reduced to zero . Missed calls have been eliminated. The customer was also able to benefit from an economic saving of about 30 percent. The absorption rate, after the first month of operation, has reached and steadily maintained the 80% threshold (8 out of 10 calls are handled independently with Reception AI).

How Reception AI works

The service is based on an artificial intelligence system that analyzes the user’s natural speech, and the user expresses himself or herself verbally as if he or she were interacting with an operator, without having to press buttons on the phone’s keypad.

A complex algorithm disambiguates the requests and then the system transfers the call automatically. If the bot cannot understand what the caller reported, the call is transferred to an operator, who will also have available the text of the communication that has already taken place.

The system is fully configurable and allows certain business areas to be excluded from this management so that they are managed, for example, directly by the operator. A control dashboard is also provided to check service reporting. The bot can be continuously improved by subjecting it to newly encountered case histories and indicating how it should behave on those occasions.

How does the automatic switchboard operate? Examples of interaction

A patient called and asked if he could get in touch with the doctor treating him, including specifying the reason for the call. The system correctly directed him to the department where the physician of interest is active.

In another case, a person called the switchboard with the intention of making an appointment with a particular doctor, giving his name and specialty. The voicebot evaluated the most relevant words spoken by the patient and understood that the focus was on “requesting a reservation,” sorting the call to meet the need.

Sometimes it may happen that some words are not understood in the first instance, and in those cases the voicebot prompts the user to repeat. “I would like to make a station, is it possible?”-this is a clear example of an incorrectly dictated question. The automated switchboard failed to locate her in the case history it relies on and asked for clarification, receiving as an answer, “I want to make an appointment for a visit.”

It may happen that some users are not very comprehensive in their exposition, and it is in these cases that the voicebot must understand the real need of those on the other end of the phone. “I am waiting for an admission” is too general a statement, but the automated switchboard was able to correctly direct him to the admissions department.

Some people may have some reticence about interacting with a robot and on some occasions require talking to an operator. In these cases, the voicebot makes it clear that it too can be of help and expose its need, being able to respond effectively.

The case history includes not only interactions with medical and hospital implications, but can also involve the administrative and economic side. For example, a patient got in touch with the switchboard to receive an invoice copy of a service he had enjoyed.