Stop automating blindly: Why human insight is key for intelligent business process automation

When many businesses first examine their processes in detail, they often discover something surprising: what they think is happening isn't what's actually happening. Consider a healthcare provider that set out to improve their billing operations. While data analysis revealed unexpected bottlenecks in the insurance pre-authorization process, it was the insights from healthcare staff that proved to be crucial in understanding the full picture.

Speaking with the front-line staff revealed an important truth: certain "delays" in processing—things that the data analysis flags as inefficiencies—were necessary moments of careful coordination between specialists, insurance providers and healthcare teams. These complex interactions created better patient outcomes but wouldn't be apparent from looking at the data alone.

This example illustrates a fundamental truth: successful business process automation requires both data-driven and human-driven insights. The gap between how processes work in theory and how they work in practice can be closed when organizations combine rigorous data analysis with deep understanding of the human role in workflows. When this synthesis is achieved, intelligent process automation (IPA) proves most transformative.

Intelligent process automation combines robotic process automation (RPA), artificial intelligence (AI), natural language processing (NLP), workflow automation and process management technologies to create an integrated framework capable of handling complex, end-to-end business processes.

This integrated approach is crucial because process improvement isn't just about the data or efficiency— it’s about better outcomes and a stronger understanding of how people work within their systems to achieve those outcomes. By combining the revealing power of process mining with the contextual richness of human insight, organizations can create better automation solutions.

Many organizations struggle with where to begin their automation journey. "Understanding where to start and what could be the most impactful area for automation to achieve maximum ROI is often the first hurdle," explains Parul Mishra, Vice President, Product Management, Digital Labor at IBM®. Process mining helps solve this challenge by offering organizations a data-driven lens into their operations.

"Process logs may be messy," notes Merve Unuvar, Director, AI Assistants for Business Automation at IBM Research®, "but studying them helps us find bottlenecks and understand where we can improve." Process mining tools analyze system logs and digital footprints to reveal actual process paths, measure variations and quantify the impact of different process paths on performance metrics.

The healthcare provider's experience in the aforementioned example demonstrates this revealing power of process mining. While the organization initially focused on billing inefficiencies, their analysis of process data uncovered a different story. The system logs showed the insurance pre-authorization process creating significant upstream delays, with certain paths taking up to four times longer than standard processing times.

The data revealed a clear pattern: when pre-authorizations were delayed, it triggered a cascade of missed deadlines, rescheduled procedures and delayed revenue collection. Process mining helped the organization quantify how delays in one part of the process rippled through the entire system, affecting both operational efficiency and patient care delivery. Most critically, these delays could postpone vital treatments and diagnostic procedures, potentially compromising patient care quality and outcomes.

Process mining reveals what is happening in an organization's workflows, but it can't always explain why these patterns emerge. To understand the full story behind these pre-authorization delays—and to design truly effective solutions—the organization needed to look beyond the data and turn to the people who knew these processes best.

"Humans have intuitions, experience and soft skills that computers do not have," explains Unuvar. These qualities were essential in understanding the healthcare provider's pre-authorization process. When the organization engaged with front line staff and insurance specialists, they uncovered crucial insights that the data alone couldn't reveal.

"Unless you dive in and take a look at the process, side by side with the knowledge workers, you may not be able to derive these conclusions," notes Unuvar. This proved true as what appeared in the data as processing delays often represented necessary clinical deliberation and coordination. Some cases required careful documentation from multiple specialists to guard patient safety. Others needed detailed coordination between insurance providers and healthcare teams to secure coverage for complex treatment plans.

The organization discovered that cases involving multiple chronic conditions, experimental treatments, or specialized medical equipment typically required longer review times. These cases needed thorough clinical documentation and often multiple rounds of insurance verification.

Armed with this understanding, they redesigned their workflow to create two distinct paths: a streamlined process for standard cases with clear documentation and coverage parameters and a more consultative approach for complex cases requiring specialist input and detailed insurance coordination.

This interplay between process mining and human insight proves equally valuable in another case study: an IT service desk's attempt to streamline their support processes. While process mining revealed repeated tickets for password resets, server access requests and software license requests, employee feedback uncovered crucial complexities. "When you look at the process holistically, you realize that some tasks are completely redundant and can be eliminated, or that there are things that you could optimize across the tasks," explains Unuvar.

The IT staff shared how users often mistyped or omitted critical information in their requests, leading to time-consuming back-and-forth communications. Overworked staff could take hours or days to respond with requests for missing details, creating a cycle of delays and frustration.

Not only did this slow down resolution times, but it also forced IT staff to spend valuable time on administrative follow-ups rather than addressing complex technical issues. These insights revealed that simple end-to-end automation wouldn't be enough—the solution needed to address both user guidance and information gathering upfront.

Drawing on this frontline experience, the team developed a hybrid approach—an intelligent assistant that combined automated information gathering with human oversight. The system was designed to automatically detect common omissions and guide users through a dynamic questionnaire based on their request type.

For instance, if a user needed software access, the system would automatically check their department, role and user access level while prompting for any missing details—checks that experienced IT staff knew were essential for proper access management. This real-time validation and information gathering meant that when tickets did reach IT staff, they were complete and properly categorized for quick resolution. This balanced solution not only improved efficiency but also enhanced customer and employee satisfaction by providing clearer guidance and faster ticket resolution.

The path to successful process automation reveals itself at the intersection of data analysis and human understanding. What might appear as inefficiencies in the data—whether longer pre-authorization reviews or repetitive support tickets—often mask crucial human elements that create value.

Rather than simply automating documented processes or eliminating apparent inefficiencies, successful organizations first seek to understand both what happens and why it happens. Through this deeper understanding, they create solutions that enhance rather than replace valuable human contributions, leading to more effective and sustainable process improvements.

The lessons from both cases underscore a crucial point: effective automation isn't about replacing human judgment—it's about enhancing it. "Organizations should prioritize open communication, thoroughly define their automation goals and actively engage employees throughout the process," says Mishra.

By incorporating both human insight and process mining data, organizations can develop solutions that both improve efficiency and preserve the valuable human elements that help promote quality outcomes.