How To Get The Best Out Of Automation

Dr. Markus Noga and Sebastian Schroetel

In 2016, the management consulting firm McKinsey predicted that up to 70% of all tasks are potentially automatable with so-called next-generation technologies. Companies worldwide jumped on that bandwagon and invested heavily in one of the hottest innovations when it comes to automation – robotics.

Today, only a few months later, some experts claim that robotic process automation (RPA) has only been a fast-paced trend based on the excitement of industry leaders, and it is not the predicted “panacea” for all the challenges enterprises face regarding automation. Recently, another blog post by McKinsey tackled this topic and qualified the initial enthusiasm for bots and their supposed potential to incur all sorts of back-office processes. In fact, the rapid adaptation of robotization waived the consideration of its potential downsides.

According to McKinsey, “installing thousands of bots has taken a lot longer and is more complex than most had hoped it would be” and “not unlike humans, thousands of bots need care and attention – in the form of maintenance, upgrades, and cybersecurity protocols, introducing additional costs and demanding ongoing focus for executives.” All in all, the authors state that the economic results of RPA underperformed the estimations, especially with regard to cost reduction. The impression has been strengthened that “people do many different things, and bots may only address some of them.”

The next level of automation

Considering the latest trends in robotics that came along with unexpected complexity, little flexibility, and additional maintenance, automation is still pushing forward. The aim is to help enterprises realize their potential and switch focus from just “keeping the lights on” through human manpower to growth generation triggered by automation technology. A dedicated automation approach to achieve the intelligent enterprise involves three interacting levels:

  • Software components, or “engines,” that provide automation relying on highly specific process knowledge
  • Machine learning that involves teaching a computer how to spot patterns and make connections by showing it a massive volume of data – algorithms that can learn from experience without having to be explicitly programmed
  • Robotic process automation software that operates another application without the support of a human user, helping to run repetitive, rule-based monotonous tasks and bridging temporary gaps

In contrast to the mere RPA approach many companies have pursued in the past, only the integration of all three layers lifts the enterprise to the next automation level. Engines are the basis of enterprise automation activities. They enable companies to shape their processes by making decisions on where to direct incoming inquiries at subsequent steps. But engines have a fixed logic and limited configuration possibility. Therefore, they cannot cover all facets of the business processes and have the potential to only facilitate automation in up to 60% of all cases.

In credit management, for example, credit-rules engines can help evaluate personal creditworthiness and process credit limit applications in a structured way. This is done by automatically categorizing them based on defined scoring rules and assigning a specific credit limit to the customer after the examination is completed.

Applications for machine learning

But what happens when a scenario occurs that wasn’t encountered by the operator? By adding intelligent automation technologies to the automation portfolio, processes become noticeably intelligent. Machine learning can upgrade the automation level of a process up to 98%. How? By setting up general guidelines without telling the system exactly what to do. The underlying algorithm learns from the operator’s previous actions and takes all available data into account to deliver the most relevant response to an occurrence.

Applying this to credit management, machine learning is useful in those cases where a customer lacks a dedicated credit history. Here, machine learning fills in with more accurate forecasting models based on people’s overall payment history, on information related to the borrower’s interaction behavior on the lender’s website, and other unstructured data sets.

Robotics, as the third automation layer, can help automate the remaining two percent of repetitive, monotonous tasks in a process. But due to its lower integration level, RPA is limited in its reach and adds the percentage on top on much higher costs. In financial risk management processes like bank lending, robotics can deal with requests for overdraft protection or credit card approvals.

A genuine alternative to mere bot systems

Related to the downsides of bot systems, a multi-automation-layer approach is the way to set up a stable and holistic automation concept as an alternative to pure RPA to flatten or avoid the disadvantages bot systems entail.

The McKinsey authors support the thesis that robotics should be used in exceptional cases, instead of being applied as the universal remedy to deal with repetitive tasks.

All in all, enterprises are actively searching for ways to shape their processes and automate parts of their work. Robots are perceived as being too inflexible, expensive, and complex in their maintenance to accomplish these goals in a satisfactory manner. By expanding the automation portfolio with engines and machine learning, a meshing system of automation technology can address these concerns and force a holistic implementation of automation throughout the enterprise.

Currently, companies and CIOs are resetting their bot programs. Figuring out the desired goal of automation might help to steer it into the right direction.

SAP’s automation strategy in general, and our cloud-based machine learning portfolio and related services in particular, are ready to step in and to fill the automation gaps that bots leave.

Dr. Markus Noga

About Dr. Markus Noga

Dr. Markus Noga is vice president of Machine Learning at SAP. Machine Learning (ML) applies deep learning and advanced data science to solve business challenges. The ML team aspires to building SAP’s next growth business in intelligent solutions and works closely with existing product units and platform teams to deliver business value to their customers. Part of the SAP Innovation Center Network (ICN), the Machine Learning team operates as a lean startup within SAP with sites in Germany, Israel, Singapore, and the United States.

Sebastian Schroetel

About Sebastian Schroetel

Sebastian Schroetel is a director at SAP for machine learning in the digital core. In this role, Sebastian and his global team shape and create machine learning solutions for SAP's core ERP products. Sebastian has 10 years of experience in innovation software development, with focus on automation, analytics, and data processing.