A Danish unemployment fundReducing failure demand using speech analytics and Artificial Intelligence

The project: Utilising speech-to-text capabilities to assess the degree of failure demand in calls

A Danish unemployment fund wanted to reduce failure demand in their call centre and initiated a project with the purpose of utilising speech-to-text capabilities, storing spoken words as written text. Using this and existing background data, Implement assisted in generating insights about the client’s member base, the amount and extent of calls as well as other call-related data.

We also developed, operationalised and implemented a model to assess the degree of failure demand in calls. This made us able to score each call and determine whether it could be labelled as “failure demand”.

In the project, we defined calls as “failure demand” when they related to:

i) Problems that should not exist in the customer journey

ii) Processes that the members themselves should be able to solve in self-service

iii) Established communication that was not clear (e.g. on website, in mails or letters etc.)

The impact: Quantifying data, gaining insights and assessing the extent of failure demand

By organising and pairing the streamed text data with the client’s data from CRM systems, the project essentially helped across three tracks:

1. Quantifying data collection efforts related to the client’s call centre capabilities, ensuring continuous intelligent business development and robust basis for future data science projects.
2. Gaining insight into the overall state of the call centre. By displaying extent, topics and other relevant metrics, the client got relevant insight into the capabilities and capacity of their team.
3. Assessing the extent of failure demand in order to improve business processes and customer journeys, thus reducing the magnitude of these calls. Approximately 30% of calls were labelled as failure demand.