Speech routing gives customers the opportunity to state questions in their own words. 90% of all questions can be classified using speech recognition, which means a boost of self-care with 15% and redirected calls are recuded by up to 20%. AHT is reduced by up to 45 seconds per call as a result of routing with voice recognition and CTI. This has a positive effect on customer and employee satisfaction.

4 steps in the customer journey

With our speech routing solution, the availability of your customer service will improve and customer experience will be optimized. Customers are identified and questions classified using speech recognition. Calls are handled via self-service or routed to the right agent with the right skill. The agent will immediately see if the customer has already been in contact before and about what. We distinguish four steps in the customer contact journey.

Image

Who is the customer?

With the identity based on CLI, we query the CRM and other systems and collect as much information as we can. This information includes: recent orders, status of previous contacts, client profile, etc. We can anticipate questions; for example if we know that delivery of the customers order is delayed we can act accordingly.

Some callers will be directly routed to self-service or to a specific agent skill, for example if we suspect fraud or if there’s a problem with a payment. Repeat callers can also be routed directly to the previously selected destination.

Image

When you know who the customer is, the products and services of the customer are known and possible related questions can be anticipated for the next step: classification/p>


What is the question?

In order to help the customer properly, it is important to know the reason for their call. Often an IVR menu is used for this, but the most natural way is by using speech recognition.

Image

We ask an open question, like:
"Good morning welcome at Webhelp, how may i help you?".

We can do this in every (major) language and the exact phrasing of this question is determined throughout the data collection process and A/B-testing. We use (text) data from numerous sources, including your website to create a language model. In addition, we train the AI-system with tens of thousands of real-world examples. In a workshop with (experienced) contact centre agents, we determine the required classes and map all training data to the classes. As input to this exercise we need data over a long period to make sure that we covered seasonal variances.

In production, our system has a built-in feedback loop to continuously monitor and improve the classification process.

Image

Routing with CTI

Image

As soon as the agent receives the call, contact history and relevant data from the CRM, will be displayed on the screen using Computer Telephony Integration (CTI). Also, the transcript of the customer’s input and class/category of the call are displayed. The agent doesn’t need to find out the call reason but can focus directly on the answer. A link with the knowledge database can provide additional information on the topic without the agent having to search.

Therefore, routing with speech recognition and CTI can save 20 to 45 seconds per call.


Reporting & analytics

Real-time information about the status of the agents, service levels and customer satisfaction is essential for managing a contact centre.

Speech Routing provides real-time information on possible trends and TOP 5 frequently asked questions. In addition to the standard reports, Speech Routing provides insight into the handling of customer contacts per topic. It also provides a list of possible self-service applications.

Speech Analysis can provide even more information. Telephone calls from the contact centre are processed and analysed using speech technology.

As explained above, the system provides very accurate and real-time call dispositions and frequencies. The call dispositions and volumes provide a list of possible self-service applications. We can easily add informational (FAQ) self-service applications that require no integration with back office systems and route calls to existing transactional self-service applications.