Invoice Automation: Increasing Efficiency in the Office at Satherm GmbH Using Artificial Intelligence
Link zur Publikation:
https://link.springer.com/chapter/10.1007/978-3-030-80003-1_3
Abstract:
(a)
Situation faced: The application of artificial intelligence (AI) to automate business processes is becoming more important in small- and medium-sized enterprises (SMEs). This was also recognized by the company Satherm GmbH. With 20,000 invoices per year, which were all being processed manually, invoicing at Satherm was accompanied by high process costs and a high expenditure of time. Furthermore, Satherm struggled with regional socio-economic difficulties such as a shortage of skilled workers and recession for which a solution had to be found. In view of the situation at Satherm GmbH, automation using AI technology offered a great potential not only to reduce manual activities and costs but also to overcome regional problems.
(b)
Action taken: State-of-the-art technology consisting of seven established neural networks for superior, cognitive and automated document recognition and extraction was used in combination with a free-to-use robotic process automation (RPA) technology to automate the process from invoice receipt to payment. The implementation consisted of three steps. The first one dealt with requirement analysis, conception and identification of the necessary interfaces. To reliably extract valid data of invoices, the second step addressed the adaptation of the neural networks to the underlying documents at Satherm by means of transfer learning. The last step involved the setup of the holistic solution and the piloting in the company to test the solution parallel to the daily business followed by a smooth transition into final roll-out at Satherm.
(c)
Results achieved: Considering the set goal of an invoice automation degree of 50%, the results achieved exceeded the expectations at Satherm by 20%. With an overall invoice automation of 70%, the invoice processing time was reduced from 15 days to 2–3 h per invoice. Complementary to this the costs for processing an invoice were reduced from € 5.77 to € 1.93. This resulted in a cost saving of about 67% per invoice. In addition, the implementation was not only accompanied by a positive impact on the process efficiency itself but also by several positive side effects like countering the current as well as future shortage of skilled workers at Satherm.
(d)
Lessons learned: The use case of Satherm is a suitable lesson to understand how digitization and especially the application of almost ready-to-use artificial intelligence in German SMEs can increase efficiency, raise potentials and be the solution for different challenges. A sufficient pilot phase parallel to business operations can be recommended when introducing new technologies in order to build trust in the technology and ensure full integrity. With that an AI application can be the driver for the discovery of further potentials and an accelerator towards a more digitized, efficient company.