Artificial intelligence improves the customer service

Replacing error prone tasks by automated tasks for enhanced customer support

Our customer, a leader in insurance in Switzerland, approached OWT with the need to review its multi-channel customer support's business processes. Aside from phone calls and forms, the customer receives each year tens of thousands of emails, which are handled by the customer support team. Among the emails, a lot of them could be fully automatized in order to re-focus customer support team on high-added value tasks.

Analyzing the client's business needs and potential improvement opportunities

Following a business assessment, OWT grouped all tens of thousands of emails into 15 use case categories representing each a different type of request. Among them:

• Change email address

• Request insurance documents 

• Inquire or contest invoices

Before being presented to the client for prioritization, OWT analyzed each use case's implementation complexity and potential impact on workload for the customer support team. This was then put in a decision matrix as seen below. As a result, OWT functionally and technically specified the requirements to implement the prioritized functionality: change address.

OWT Data Scientists and engineering team working together to deliver unprecedented process automation

The automation of the change address process required various components to be implemented:

1. Email triage: AI algorithm implemented by OWT that classifies the type of request based on the free text of every incoming email.

2. Email information extraction: matching of the database's information with the one present in the email. Also, the new addresses were matched with Swiss Post database to ensure the new address truly exists and does not contain any typo.

3. Email formatting: transform free text emails into structured information with a next step button to validate or perform actions directly.

4. Automation: update customers address in the system.

Automatizing flow to reduce errors and increase productivity

The new solution brought consistency and automation for the colleagues to be more focused on high level tasks; this is also reducing error thanks to the automation.

Proven Methodology

Based on its expertise in handling digital project, OWT applied its proven methodology to assess, analyze, design and implement the current and future solution for the customer's customer support. We applied the following methodology:

1. Blueprint: We analyzed the current process, tools & type of emails received by the customer team. This analysis led us to discover around 15 email topics among them 4 have been prioritized by the customers to automatize. After prioritizing the functionalities requested by the customer, OWT specified the functional and technical requirements.

2. Development: We applied Agile Scrum methodology during development in order to deliver software releases and receive continuous feedback from the customer every two weeks.

You can also read more about machine learning insight: AI Email classification.