Don't automate immediately, think carefully first
Robotic Process Automation (RPA), Large Language Models (LLMs) and artificial intelligence (AI) make it easier to automate existing (legacy) processes. A common example is the use of RPA to process incoming PDF invoices from a mailbox in an ERP system. But isn't it more efficient to place these invoices directly in the ERP via an XML message?
In the financial sector, we see that Know Your Customer (KYC) departments, sometimes consisting of more than 3,000 employees, struggle with identity verification processes and try to speed this up with algorithms. By applying digital identity techniques, you can indisputably determine who the person is at the first customer contact. This makes KYC departments largely redundant.
In field services, we also see inspection apps being used by technicians to automate data collection processes. However, with real-time monitoring and predictive maintenance, many of these field service operations can become redundant.
At the same time, marketing departments of large multi-brand companies are using Large Language models and chatbots to improve their customer service. However, the number of customer service inquiries can be significantly reduced by simplifying the product offering and improving the user-friendliness of the services.
Start with the process, then automate
In short, we automate what we know and forget to repeatedly scrutinize the processes. Therefore, start with a comprehensive analysis of the processes and only then automate.
Don't get me wrong: technology can certainly help to improve our world and automate repetitive, mind-numbing work. But it is essential to first determine which processes actually need to be automated. The greatest cost savings are achieved by completely eliminating certain processes, as digital techniques make them obsolete.
