What drove the decision to develop an AI-based use case?

Especially when working with language (e.g. in the development of voicebots), AI offers many opportunities to transfer human skills, such as speaking or learning, to the technology. AI makes our voice technology capable of learning and helps us to develop voices and bots with natural human interaction skills.

Voicebots in customer service must be perfectly tuned, react quickly and correctly and ideally draw the right conclusions, as callers are much less inhibited about hanging up on a machine at the other end of the line than on a human. Customers have the same requirements, regardless of whether they are calling a service employee or a machine, which is why the voicebot must be one thing above all else: a digital agent.

 

What were your expectations or requirements for the AI application?

For companies with both a high call rate and high customer service quality standards, the requirement for an automation tool is always to act like a real person. This means that the voicebot must be reliable and as human-like as possible in the tasks it performs. One of the main cost factors in service is the agent's (working) time. This needs to be optimized through the sensible use of AI and many tasks need to be solved on a case-by-case basis.

We developed and successfully implemented a bot for a service provider in the automotive industry. The aim was to create a digital assistant that would answer the call and authenticate the caller according to four of the following criteria in compliance with data protection regulations: customer number, date of birth, zip code, house number and telephone number.

When dealing with customer data, data protection is of course the top priority. This also applies to us, which is why our solutions can be offered both on-premises and in the cloud or on-device.

 

What frameworks/methods did you rely on for development?

Since we develop in-house, we are usually not dependent on external partners. This gives our clients flexibility and us absolute customer proximity, as we can also respond to the client's wishes at short notice. The IT infrastructure in large companies is often complex, but this can usually be managed well through constructive, close cooperation. With 12 years of expertise, we develop state-of-the-art language technology solutions that we can integrate into any standard software environment.

 

The use of AI solutions must also be accompanied by a change in competences. How have you dealt with this challenge in respect of your employees?

At Aristech, we focus on the continuous development of the individual skills of each employee and the entire team. On the client side, competence development begins with our support through kick-off and project-accompanying workshops, in which we train the use of our self-developed tools, such as the Aristech Dialog Designer.

We also help the client to take an analytical look at their customer service, identify standard cases and select use cases that are easy to automate.

 

In what timeframe did they implement the AI solution in your company/your client‘s company?

The duration of the implementation varies depending on the solution and the client's infrastructure. In this use case, we implemented our voice automation within 16 project weeks. Starting with the requirements analysis and a concept workshop, the dialog was set up in the first phase and the interface connection and setup of a tracking tool were carried out. In phase two (project week 8), the technology was optimized after final overall testing of the pilot and on the basis of initial user tests. The pilot went live in week 16 of the project.

 

A conclusion to your AI solution:

The voicebot already achieved a higher authentication rate than estimated during the pilot phase. After agile optimization of the speech recognition models and the dialog flow, the voicebot achieved an authentication rate of 89 % even under increased load. This means that the voicebot is still slightly less performant than a human, but by reducing the call time by 1.5 minutes through authentication alone, resources can be allocated more sensibly. Instead of processing standard requests, agents can now use their time for individual requests. In addition, the client already achieves significant cost savings in the first year. All of this contributes to a significant increase in the quality of customer service.

Our goal is to offer our clients an outstanding solution that improves service, reduces costs and relieves the burden on employees. In the use case described, the solution has been continuously expanded for over a year now. After an initial on-premises implementation, the use case has now been moved to a secure cloud environment.

 

One final piece of advice/tip to other entrepreneurs looking to apply AI:

For clients who want or need to work exclusively in their own system environment and for whom data protection is a top priority, we naturally also offer our solutions on-premises. In our experience, the initial hurdle to introducing the technology is the biggest. An upstream pilot lowers the inhibition threshold and quickly convinces other managers. Agile optimization measures prove to be particularly important here.

 

 

www.aristech.de

 

Picture: Aristech