Artificial intelligence (AI) can significantly relieve the ticket process at bank service desks and make processes more efficient and more consistent. And employees gain freedom for more demanding tasks. A best-practice report.
Digitalisation in the banking environment is leading to ever-increasing demands on IT service management (ITSM): more and more internal and external requests arrive at the service desk through different channels and have to be handled quickly, correctly and in a regulation-compliant manner. For employees, the many internal (“My laptop doesn’t work”) and external tickets (“Error at an ATM”) mean a lot of manual effort. Even the correct initial categorisation and routing of a ticket can take a great deal of time – especially when service desk colleagues first have to look for the right category or responsible group.
That is because the underlying category systems for ticket management are often very granular. Depending on the department, there are thousands of categories and many different handler groups. In individual cases, this initial assignment used to take considerably longer than solving the actual problem. Against this background, a bank decided to examine the use of artificial intelligence to support its service desk.
Support instead of forced automation
The aim of the project was not to fully automate the service desk, but to relieve it in a targeted way – to achieve faster and more consistent initial categorisation of tickets, more reliable assignment to the appropriate handler groups, and less manual searching and decision-making.
To achieve this, the existing USU Service Management (USM) was extended with the AI component USU AI and seamlessly connected to the existing ITSM system via a REST interface. The AI-based classification was built on historical ticket data that had already been correctly categorised and assigned. This legacy data was used to pre-train the AI model to recognise typical patterns in short texts and descriptions.
AI support in practice
Today the solution supports the service desk’s ticket process in two key steps in particular: the functional categorisation of a ticket and its assignment to the responsible staff. The artificial intelligence is not used fully automatically, but serves as a controlled assistant for service desk employees.
The following example shows what this looks like in daily work: when a customer reports that an ATM at a particular branch is not dispensing cash, the service desk employee first creates a ticket, for instance with the short text “ATM not working” and a description with location and additional information. Within the ticket, the AI can then be activated deliberately when the employee clicks the corresponding button. The relevant content is then sent to the AI, which analyses the text, compares it with similar cases from historical ticket data and, on that basis, suggests both a suitable category and a likely handler group. The employee reviews this suggestion and either accepts it in full or adjusts it as needed.
If the match probability appears too low or the assignment is not plausible, the ticket can still be processed manually in the classic way. If the AI suggestion is accepted, the ticket is forwarded automatically; otherwise it follows the usual process. There is no obligation to use the AI.
Noticeable relief and more consistent processes
By using the AI component, the manual effort at the bank’s service desk – particularly for initial assignment – was significantly reduced. The solution also ensures more consistency in categorisation, because it is guided by historical patterns, regardless of the experience or detailed knowledge of individual employees. Especially in the banking environment, where incorrect assignments can quickly lead to escalations, this is a key advantage. The quality of the initial assignment improved, which in turn enabled faster processing in the specialist and technical teams.
Implementation and operation
The technical implementation – from training through integration to go-live – took only around three months, thanks in part to the close cooperation between the “handz.on” team and the USU experts, as well as access to existing interfaces and functions.
An important success factor for the new solution is also the continuous learning effect the software enables: new ticket data generated in day-to-day operations regularly feeds back into the model’s training. This keeps the AI accurate even when system landscapes change, new services are introduced or terminology is altered.
Conclusion: freedom for more demanding tasks
For the bank, the greatest benefit of AI in IT service management lies in its intelligent support of employees. Today the solution relieves the bank’s service desk exactly where it can handle routine tasks and save staff time. It gives employees room for more demanding tasks without destabilising existing processes. Particularly in the regulated, process-driven banking environment, this approach is decisive for acceptance and lasting success in day-to-day work.
