Maximum automation starts with the right mindset

It is widely known manufacturing was an early adopter of automation. Ford engineers began work in 1949 on the first factories built for extensive use of automation, though that was just the beginning. Today, we are experiencing an automation revolution. AI components that use computer vision, machine learning and natural language processing have stepped up the game. Robotic process automation (RPA), for example, is surging for low-complexity tasks and virtual agents are handling more customer self-service.

The business case is overwhelming: enable a conversational interface to advance customer experiences (voice will be the main input/output method within five years), drive the top line by advancing existing products and inventing new ones, and create a margin expansion engine.

According to Gartner, AI-derived business value is predicted to reach $3.9 trillion in 2022, more than double that in 2018.

Businesses face a radically different competitive market. To offset shrinking margins, technology product managers must invest in intelligent automation to reduce labor costs and improve service quality, otherwise risking obsolescence.

To succeed under the imminent change, we need an automation framework. Technology leaders need to control a diverse, pervasive and (sometimes) disruptive roll out of automation across the enterprise — capturing the expected upside without hurting customer and employee trust and experience. The automation framework needs to encompass not only the technology elements to automate most IT and business processes, but also be able to address the people and business impact.

Successful automation initiatives involve not only having the technical tools, but also a cultural shift towards an automation-first mindset among the people of the organization. This could be enabled by using the best practices of Organizational Change Management (OCM). One must get past the “whatever I do is too complex to automate” culture.

This article outlines some of the pragmatic ideas that help maximize the benefits of automation.

Embrace the transition, not the outcome

We must change our mindset from asking the question of whether we can automate to why we need manual effort in the first place. In the beginning, we should exploit the best of available automation tools with possible augmentation by humans to achieve the efficiencies, which can then possibly fund the more involved machine learning and AI use cases. You don’t need to use AI techniques for a lot of the current automation needs. It’s like Warren Buffet's money; you don’t need to spend it just because you have it.

Start small and capitalize on the positive momentum

At Softtek, we believe in the common wisdom that one should start with a few use cases or processes that are repetitive and follow the logical flow (low complexity) to automate—some might call them “low-hanging fruit.” As long as the expectations are correctly set—that this is the start of a journey—then this is a great way to begin. It creates excitement and gathers momentum when the users realize the benefits. This can be done as end-to-end automation without any manual involvement. Once the initial roll out is complete, you continue to iterate.

Don’t give up                                         

Once we get past this simple automation, some complex process workflows might require human involvement at specific steps before continuing. We should build an automation framework that targets incremental autonomy rather than start from the (usually unattainable) goal to create full autonomy. Typically, our initial reaction is to give up on automating the process as it cannot be fully automated and requires numerous manual steps. However, to maximize automation, we can’t look at this as a dead end. We should automate as many steps as we can in the process and provide a way to obtain manual input easily – e.g., a convenient mobile interface either as an application input or text response. This is especially applicable in use cases in which approval is required at certain stages before continuing. Other examples include steps that involve human judgment as automation cannot be trusted with the definite action (which can be enhanced over time through training the machine learning algorithms).

Automation mindset in the real world

As an example, consider the SAP ERP refresh process. This process typically has more than 120 steps. After every few steps, the process needs approval or the specific input of a SAP (Basis) Administrator to continue. With the automation mindset, we automate most of these steps (~95%), but if we can send a text message with the context and an easy Approve/Deny button choices using a mobile interface, then we have a winner.

Unmanned aerial vehicles (UAV) are another great example of augmented automation. Someone in a far-off location is monitoring the movement and guiding it, as fully autonomous mission completion is not yet possible for UAVs.

Taking automatic cars as another example, the current technology is good enough that given relatively benign conditions, automated cars can traverse. The problem with mass adoption pertains to the number of exceptions that can occur, such as extreme weather elements like rain and ice, or someone jumping from an alleyway. The currently available AI solutions are not adept at handling such exceptions with the confidence that is required in the real world. That is why cars like Tesla have augmented automation (however, Tesla and other vendors are collecting data which will help them in full autonomy in the not-so-distant future using AI).

Automation mindset: appendix

Some other considerations for your automation journey:

• At Softtek, the concepts mentioned above help us execute a mindset of maximizing automation in all the services we provide for our customers. No process is too complex to automate; we might start with 30% steps automated, but over time we continue to increase the level of automation.
• In the corporate IT world, often, there is not enough data to apply data science techniques, so the use cases are limited for applying AI. However, we found a tremendous opportunity to apply Natural Language Understanding (NLU) and machine learning algorithms in building solutions for QA and DevOps. Another important application of machine learning algorithms is in the AIOps (Artificial Intelligence for IT Operations) area by automating actions for alerts from monitoring tools. Also, we have been using NLU and ChatBots in the IT Service Desk area to provide automation and self-service. For business processes, we find many more applications of data science as there is usually a lot more available data and larger Return on Investment (ROI).
• With augmented automation, the required skill level is lower and temporary, making sourcing from low-cost countries an attractive option.

In summary, we must apply an automation-first mindset to best leverage the benefits of currently available automation technologies. This includes building an automation framework focused on incremental autonomy vs. full autonomy. AI solutions should be used in specific use cases where it makes sense, e.g., when we have the availability of large, quality data sets for predicting the behavior or for self-healing. We should run the marathon, not the sprint.