Intelligent Automation: Hyperautomation or Just Hype

Many terms describe the automation trend used by various organizations: Intelligent Automation, Cognitive Automation, Cognitive RPA, and even Hyperautomation.  These are all variations on a theme, even if each source promoting them tries to distinguish them from each other. The bottom line is that Intelligent Automation is the combination of the process automation with some type of decision (inference) capability, typically Artificial Intelligence (AI). AI could be machine learning, deep learning, cognitive vision, etc. that infers a decision using data versus traditional programming.  We will use Hyperautomation interchangeably with Intelligent Automation for practical purposes.  

Let’s start with a definition of Intelligent Automation.  

Intelligent Automation is the combination of process automation software and artificial intelligence models to automate business processes. These technologies drive enterprise benefits by improving quality (fewer human errors), shortening process times, and reducing operational costs.  

Gartner’s publication “Top 10 Strategic Technology Trends for 2020” identifies Hyperautomation as one of the top trends defined as  

“Automation uses technology to automate tasks that once required humans. 

Hyperautomation deals with the application of advanced technologies, including artificial intelligence (AI) and machine learning (ML), to increasingly automate processes and augment humans. Hyperautomation extends across a range of tools that can be automated, but also refers to the sophistication of the automation (i.e., discover, analyze, design, automate, measure, monitor, reassess.)” 

An organization’s process can be complex. Intelligent Automation can involve a combination of technologies. These technologies include Robotic Process Automation (RPA) and Business Process Management software (BPM). The addition of some form of AI can allow for data extraction or decision processes to be automated.  

The best-case scenario is that a full business process is automation from start to finish. In this case, “finish” means that all cross-functional tasks are completed, and no work is left to be performed. For example, for AP, this means ingestion of an invoice, extracting the information into an AP system, validation checks, reviews/approvals, scheduling payment, and finally, the payment is disbursed.    

However, this is a strategic journey requiring the adoption of new technologies and digital skills. Often, an organization will start with automation of various sub-processes or tasks. They expand as they mature, add automation and decisioning to create an increasing level of value. (Figure 1) 

Process automation provides the foundation for internal processes, whether it is a BPM or RPA style automation. It is not uncommon to review processes to clarify the “approved” process path. This review lays the groundwork for automation, i.e., what is done when and by whom. Process automation relies on well-defined rules to be repeatable. The absence of a clear rule at a given step indicates then the automation of the process, sub-process, or task ends because it relies on a human. This reliance is also the opportunity step to investigate if an AI model can model that human decision.  One can now ask the following questions: 

• Does the task rely on a third-party outside of your organization?
  • If yes, then identify if there is a way to extend the automation workflow to this person (e.g. document e-signature). 
  • If no, then this is a logical breakpoint. Automation could pick up after the third-party acts. For example, receiving a file, using a form email to extract data, checking for a data update in a system such as in a partner portal or database, etc.  

• Is the lack of a clear rule due to a lack of clear guidelines or policy? 
  • If yes, then clarification with a rule or policy enables further automation. Identify exceptions and how to handle them. 

• Does the step rely on data in a structured (e.g., a form) or unstructured document?
  • If yes, then there is an opportunity to use AI/ML to extract this information via automation removing humans, improving quality, and decreasing cycle time.  

• Is the decision is based on tribal knowledge?
  • This institutional knowledge is where it can get interesting, as the question arises as to how the decision is made and validated as a “good” decision. This question is one way to identify AI opportunities.  
    
    When decisions are based on data, it may be automated. Given the right data, an AI model can be created to infer the right decision (input to the process). Humans often rely on history to make these types of decisions. However, humans are limited to what they can remember and infer. AI uses thousands of data points and results to identify the decision, often better than a human. As before in the above step, identify exceptions (or low-quality model output) and what steps need to happen.  

This process seems easy. The process automation portion tends to be more straight forward. Integrating an AI model gets quite complex, as bespoke models have their own development and lifecycle. These models require new skills and maturity levels. In some cases, a third-party AI solution tends to move that risk/cost externally, and yet beware that you may need to validate the results are within your tolerance level. Therefore, Intelligent Automation is a journey. Start small, expand, and measure the value at each improvement.  

Hyperautomation/Intelligent Automation Benefits: 

• Efficiency. Automations can run around the clock every day. Throughput increases without increasing staff. The automated steps are also performed faster than humans, so more work is completed in each time frame. Thus efficiency enables scalability.  

• Data Quality. Human error is a given and has a cost. Each time a human is required to get data from one source and input it into another source an error is possible. One of the rules-of-thumb in quality management is the 1-10-100 rule. It states that if prevention has a cost of 1 unit (e.g., $1), correcting an error costs 10 units, and failure costs 100 units. The impact could be measured in dollars, hour effort, lost opportunity, etc. The cost escalates the farther down the path this error is identified.  Removing human error is a unit of prevention by mitigating costly data errors.  

• Repeatability and Consistency. Automated workflows perform each time reliably. In compliance scenarios, this is important. Given an appropriate audit trail (i.e., log the steps), this can be proven.  

Automation is a strategic decision, often part of an overall digital transformation. This requires vision, leadership buy-in, commitment to new technology, and cross-team involvement. There should be governance processes to identify new opportunities and engage. 

The benefits are real. Once a commitment is made, the ROI increases with each use of automation. Start your journey today.