Everyone is familiar with them, but no one wants them – complaints are a burden for companies, but unfortunately they are part of business. But if they cannot be avoided altogether, they must at least be processed in a straightforward manner and handled efficiently. This is in the interest of the company, but also of the customers, who expect their complaints to be processed quickly and correctly. Automated processes serve this goal, and artificial intelligence (AI) provides important tools. What might this look like?
A neural network is trained based on existing, already classified complaint texts. In addition to the actual text contents and their evaluation by means of speech recognition, texts must be assigned to error codes and criticality levels. It cannot be assumed that existing training data would be directly usable; certainly mistakes have happened in the past: Complaints were incorrectly assigned or not properly prioritized. The data must therefore be prepared and cleaned prior to AI training. But then the AI application can be used in the complaints process, where it can automatically analyze and evaluate incoming messages. This can be done on the fly, which is of some importance, especially when the number of cases is high.
Now, this first stage in complaint handling is also the starting point for further possibilities for automation. After all, AI support in the complaint process need not be limited to the automatic classification of complaints, e.g., as “critical” or “non-critical.” In an extended scenario, follow-up actions are conceivable: In this way, the automatic evaluation of complaints can automatically create error messages in a digital management system, where the complaints are followed up and processed via the specialist workflows. It is also conceivable to forward evaluation results to an ERP system where, for example, the batch affected by the complaint is then blocked. And an initial pre-written response, appropriate to the complaint, can be sent to the person from whom the complaint message originated.
DHC has investigated these and other use cases in medical technology in more detail as part of a research project. The project results fundamentally demonstrate: AI-poweredcomplaint managementis possible and helpful. Complaints are classified in real time with AI support and can be assigned to the correct persons responsible or answered immediately if necessary. The added value of this approach results from improved efficiency in terms of higher processing speed and more security thanks to a “4-eyes principle” consisting of man and machine. It is also advantageous that critical complaints can be detected promptly by automatic means and prioritized automatically.
Results on the use of AI in the complaint process are published in a whitepaper “The New World of Complaints, AI-based – Processes, Application Potentials and Solutions“. There, the aspect of software validation (CSV), which is important for the regulated environment, is also addressed and discussed with regard to the special case of validation of AI-based systems. Another whitepaper outlines a “Technical Solution Architecture for AI-based Complaint Management“.