Digitalization of processes is the dominant trend in business today. Increasingly better technical capabilities are fostering the automation of larger and more complex areas. This trend is also forced by the labor market situation. Difficult access to specialists is a good motivator to relieve them of repetitive and predictable tasks by assigning them to machines. On top of that, there are of course financial and working time savings.
In addition to the aforementioned repeatability, other features that foster process automation are high frequency (the process must be mass-scale for the preparation and “hiring" of a bot to be cost-effective), unambiguous rules governing the process and its high standardization.
Typically, automation also involves tasks that require actions in two or more applications (time-consuming for an employee who has to switch between different solutions). The widespread adoption of artificial intelligence and machine learning solutions allows for supporting automation and expanding its application in business to include processes that were previously impossible for bots to perform due to their complexity, and which take place at the contact point between humans (employees, customers) and the organization’s systems that are, however, not used by them.
Artificial intelligence has long been no longer futurology, but reality. We encounter it as customers – using automated hotlines, chatbots or other means of communication in B2C relationships. We can also deal with it as employees.
AI is widely used in IT systems across various business domains. However, AI is not just a machine. It is also a data analysis and inference tool that helps identify trends, patterns and relationships to make better business decisions. An increasing number of companies are planning to apply artificial intelligence in their systems, and at All For One, we strive to provide a qualitatively new approach to this topic in the Polish market in order to meet these expectations.
There are various tools available on the market to support the application of AI in integration with business solutions. They include both open source tools and tools from renowned software vendors. When choosing the appropriate solutions, companies consider not only the license cost but also the implemented security policy and the adopted rules for the selection of AI solutions.
AI from SAP
SAP also has such solutions in its portfolio. One of them is SAP Service Cloud Ticket Intelligence Machine Learning for service support, which allows you to shorten the response time to a ticket thanks to the use of accurate and automatic classification of tickets based on predictive models.
SAP Service Ticket Intelligence (SAP STI) is a tool that is part of the SAP AI Business Services portfolio. The main goal of SAP STI is to shorten the response time to customer tickets, increase service efficiency and improve customer satisfaction. The system analyzes the submitted ticket information, such as content and topic. Then, based on the collected data, it assigns it to the appropriate category, group, and finally to the appropriate team or employee who will be responsible for the processing of the ticket.
STI uses machine learning (ML), which is a subset of the broader concept of artificial intelligence (AI). ML focuses on using data and algorithms to mimic the way humans learn, gradually improving its accuracy. As a result, instead of having programmers write instructions on how to solve a problem, the ML model learns based on information from large data sets.
The SAP STI service offers the following ways of operation:
- Text classification e. the prediction of categories of new tickets. Predictions are based on historical data of previously classified tickets.
- SolutionRecommendation: the service returns a proposed solution to the ticket based on previously submitted data;
- TextClustering: the service helps search for trends and patterns in submitted service tickets. They are then used as a basis for creating clusters and keywords.
The process of launching and implementing the service can be divided into several stages:
Data preparation. In order to effectively classify tickets, SAP STI requires properly prepared training data. This data includes a set of tickets that were previously manually assigned to the appropriate categories. The training data is used to train the classification models.
Learning stage. After collecting training data, SAP STI proceeds to machine learning. The machine learning algorithms analyze the content of the tickets, including text, topics, keywords, etc. Based on these features, the models are trained to recognize and assign appropriate categories to tickets.
Testing and optimization. After the machine learning process is complete, the models are tested against a test set to assess their performance and accuracy. In case of poor performance, it may be necessary to adjust the model parameters and repeat the learning process to improve results.
Implementation and integration. Once the SAP STI models are successfully tested, they are deployed in a production environment. This may include integration with existing ServiceDesk customer support systems to automatically classify incoming customer tickets.
The SAP STI service provides a convenient REST API used during the phase of implementation and integration with the customer’s system.
SAP STI’s expandability and flexibility are also worth mentioning. The service can be tailored to specific requirements and industry contexts. Customized ticket classification rules can be defined, taking into account unique terminologies, an industry language and even preferences of specific customers.
A smart bot for HR
We are often approached by organizations that want or even need to automate their processes. And even though the individual reasons vary, it is still most often a matter of optimizing the time needed to handle employee matters. And for this purpose, we used the SAP STI solution for one of our clients from the automotive industry. The project involved digitalizing and automating HR processes and supporting employee self-service at the BeeOffice Service Desk.
One of the main goals of this project was to prepare a solution for handling employee requests (HR requests) in the ticket system for more than 10,000 employees. The tool supports the processing of several dozen types of HR requests.
The challenge was the need to prepare several channels for submitting requests to BeeOffice Service Desk and their further processing, including assigning different types of requests to the appropriate teams in the HR department. In addition to being able to set up a ticket directly in the system, employees can submit a ticket by phone or send an e-mail. All these communication channels set up the tickets in one place in the system. Then, in the classic approach, the responsible person should distribute tasks to departments, service groups, specific people.
But what if we are handling such a sensitive area as HR data? If we want certain tickets to be delivered directly to authorized persons without being sent to third parties on their way? The second very important problem our client faced was the number of tickets that were received by the Service Desk every day and needed to be classified into the appropriate service group. On an average day, more than 200 tickets were registered in the system. It took a very long time to read each one and direct it to the right place.
The solution to these problems was to automate assignments using machine learning.
At our client’s company, AI was trained to recognize to whom the request should be directed based on the content of the ticket. The learning process itself was based on already classified tickets from the Service Desk. We used more than 22,000 tickets as training data for SAP STI.
Then we tested the learned model on new tickets and compared it with actual assignments made by the Service Desk operator. When the effectiveness of AI reached 75 percent, the solution was transferred to the production system.
If new categories of tickets or customer service groups appear in the system, the artificial intelligence is additionally “trained" and can operate in the new reality without any obstacles.
The effects of introducing artificial intelligence in the organization have been evaluated very positively. First of all, the process has been accelerated and we have eliminated the need to perform routine, repetitive tasks. As a result, work efficiency has increased and considerable time savings have been gained. The system works 24 hours a day, 7 days a week. This means that it performs tasks even outside working hours, which increases the availability of services to users.
The AI-based system is scalable and can handle large amounts of data and tasks simultaneously. This allows the organization to continue to grow and adapt to changing needs without having to significantly increase human resources.
It’s already happening
The described solution is one of the examples of using AI in integration with business systems. Automation of service support, whether for employees or customers (e.g., support for processes carried out in broadly understood CRM and other systems) is our reality. Chatbots, reporting defects and complaints via websites or applications, or adapting the offer to customer preferences are services that we use to a greater or lesser extent on a daily basis in various areas of life. Behind them, there are often self-learning solutions that, replacing a human, enable quick and almost 100% effective execution of the request.