As technological innovations keep emerging, Shared Service Centers (SSC) need to constantly rethink their strategy to adopt and leverage new technologies. In previous articles we focused on current trends of SSCs in terms of process excellence, value-added service portfolio as well as the next maturity level of Digital Shared Service Centers: Virtual Delivery Centers. In this article, we will concentrate on further expanding the digital toolkit for SSCs by complementing Artificial Intelligence (AI) with RPA and explore AI’s possibilities and limitations in the context of SSCs.
While expanding SSC Service Portfolios with value-added services, which create additional benefits to the SSC’s clients, process excellence is required to ensure high levels of quality, cost efficiency and standardization. To become “digital” and reach the highest possible maturity stage of a Virtual Delivery Center, SSCs continuously expand their functional scope and automate services and processes. Thereby, standardized processes are an imperative for increasing process efficiency and implementing new technologies. Two prominent examples of emerging technologies are Robotic Process Automation (RPA) and Artificial Intelligence (AI). RPA is generally a great lever to increase process efficiency, robustness, costs and quality. Even more it finds applicability in SSC where processes are repetitive, transaction volumes are high and the costs are pivotal. But as RPA is focused on rule-based, structured processes, complexity is a natural delimiter of possibilities. Value-added services tend to be complex and versatile, making them resource intensive as it is harder to automate them. Think of processing unstructured data, recognizing images or even interacting with customers. Implementing AI with RPA allows to extend benefits of RPA to more complex and value-added services. Capgemini’s Business Service Centers already leveraged on both technologies in their daily operations which enabled us to gather significant insights and experiences.
AI technologies can be clustered into 5 different categories: Natural Language Processing (NLP), Natural Language Generation (NLG), Machine Learning/ Deep Learning, Predictive Analytics and Image Analysis.
Figure 1: RPA and AI technology categories
AI offers a huge digitization and optimization potential as it can answer complex questions within seconds which takes humans several hours to answer. But still, for value-added use cases AI is in its infancy. To start small, SSCs are a great spot because they tend to have a high level of process standardization and many (sub-)processes can be automated e.g. with RPA or ERP extensions. AI can serve to close gaps in semi-automated workflows to achieve fully-automated workflows. Therefore, we present two examples that illustrate the potentials of AI using classical SSC processes such as invoicing and ticketing.
Image Analysis applied during invoicing
Our first example focuses on Optical Character Recognition (OCR), which is no new technology and is already being used in SSCs. The OCR core functionality lies in recognizing characters and numbers in standard languages from pixeled images. Image Analysis complements existing OCR engines and improves quality and speed. With the renaissance of neural networks, the quality of OCR has emerged to Intelligent Character Recognition (ICR). A good practical example to use ICR in SSC is the P2P process. To automate the process, inStream, a machine-learning platform with built-in OCR tool, can be used in combination with RPA.
Figure 2: Invoicing workflow with inStream and RPA
As invoice images from a scanning provider or via email flow into inStream, the content is recognized and processed automatically. To do so, the tool has an engine which needs to be trained and filled with business rules. Therefore, it reads and learns to understand content, to extract and to verify key data. It can also enforce checks and measures to ensure compliance. After running standard OCR on scanned invoices, inStream starts recognizing document content based on patterns and applies defined business rules to decide on required next steps. The generated output is then picked up by a robot which validates the data and transfers accurate, relevant and structured data into business systems. Once the content of the invoice has been automatically inserted into the ERP system using RPA, the data is ready for further processing. Manual indexing is only required by exception.
Natural Language Processing applied on ticketing systems
Our second example focuses on Natural Language systems which are known as chatbots or voicebots and it shows how those can optimize a classical SSC support process such as ticketing. Both, chat- and voicebots revolutionize the ways of SSC’s communication and automation. In a similar manner like leveraging OCR/ICR for invoice reading, chatbots can be leveraged for Frequently Asked Questions (FAQs) during a ticketing process. In our practical example, a client works with a ticket system for HR processes that allows employees to submit requests to the SSC team. Tickets are answered by the SSC team within 48 hours. The large number of requests currently overwhelms the team and leads to delays and additional efforts. Indeed, most of the requests are repetitive basic HR questions (e.g. “Where can I find my pay slip?” or “Where can I change my private address?”). Many of these questions could be answered much faster if the employees had access to the right systems and information. Applying AI the solution lies in the implementation of a chatbot in the HR Portal as a first level support to answer first line questions. The chatbot has initially been trained with FAQs but it can further increase the process robustness and expand the level of service as it is continuously learning based on each employee-chatbot interaction. Depending on the complexity of the questions, further capabilities can be added to the chatbot. Common extensions are for example the integration of the chatbot with the ERP system e.g. in case the employee wishes to change personal data. A similar use case has been described in “How a chatbot can combine RPA, AI and ERP”.
The following illustration outlines the different stages of chatbot developments. In our example the FAQ have been trained via short tail. With increasing complexity of questions long tail trainings enable a chatbot to reply to more unique intents.
Figure 3: Complexity and capabilities matrix for chatbots and learning trails
Climbing up the complexity and capabilities matrix, will bring new challenges and possibilities when applying AI in SSCs. The use cases demonstrated in this article can already create positive business impacts after 7 to 8 weeks, which is easy to handle and implement. Thinking big when applying AI on value-added services should be translated into the strategic agenda. Setting up a realistic agenda and time horizon for adoption and application in SSC is key to realize the existing potentials. Sophisticated solutions will create awareness and open a range of new possibilities to further expand and optimize the SSC service and product portfolio and its operations, but business implementations need to follow the market maturity of technologies to leverage on their existing capacities. During a time when more and more customers expect the use of AI and new automation tools, the SSC is the right spot to advance with the application of automation and AI for its core operations.
This article was written together with my colleague Pavel Herrmann.