How AI + Human Partnerships are Driving Process Innovation
By Greg Council on July 13, 2023
As learning technologies have gone mainstream, there has been a lot of focus on nearing 100% automation and eliminating manual intervention. Robotic Process Automation (RPA) initially seemed poised to accomplish “lights out” automation in repetitive tasks. More recently, Generative AI has exponentially increased interest and concern over AI completely taking over everything from individual jobs to entire fields of work. But the reality, as anyone attempting to use any form of AI has discovered, is that even with ChatGPT demonstrating value in many different scenarios, we’re still very much in an age of AI + humans, commonly known as “assistive automation.”
Just as we equip modern self-driving cars with human overrides, there’s an ongoing need for AI and human partnerships in processes involving many distinct tasks. While RPA reliably automates simple tasks with binary yes/no options, more complex processes involving multiple decisions are yet to achieve 100% automation with 100% accuracy. That doesn’t mean organizations can’t realize significant efficiency boosts with assistive automation.
My favorite example of human/automation partnerships involves a complicated contracts process between hospitals and insurers. Complex contract terms must be identified and transposed into a system that enables quick reconciliation of payments to claims. Depending on terms and conditions, the conversion can take days. Today’s Generative AI technology can’t completely automate this complex conversion process – but 100% automation isn’t necessary to realize significant improvement. By combining automation with targeted effort, we can make the process much more efficient. Automating just part of the work can lead to a 50% reduction in manual tasks, translating into thousands of reclaimed hours of staff time – a major win for the organization.
Incorporating Generative AI into Process Innovation
When we shift the focus to Generative AI’s core ability to generate responses based on complex, human-fed text prompts, we see exciting opportunities to incorporate it into real process innovation.
For instance, consider parsing log files from various systems. Log files are foundational for monitoring the status of various processes, both at the business process and systems level. The ability to analyze status is critical, but it’s difficult to achieve actionable insights using traditional processes. One challenge is that log files are not standardized into a common format. Each system has its own structure, requiring significant understanding and mapping of each format to reliably use the data for functions such as system alerts.
These log files are text-based – meaning that humans can read them. And if humans can read them, so can Generative AI large language models such as GPT-4.
Imagine the ability to collect a series of system log files and ask for a summary of actions. Even though a Generative AI does not know the log file format, it can read the text within and make some sense of it. With a little more context provided, the AI can provide even more meaningful responses.
For example, I can input the log files from a Z/OS mainframe into ChatGPT. Working from the source information, ChatGPT generates a summary, including a list of the major steps involved and all relevant warnings and issues. Working from this summary, it’s straightforward to create notifications for system admins without first needing a specialized application to parse log files.
Generative AI also allows us to automate process modeling itself. Working from the log file summary, we can ask for a visual representation of all the key steps within a process. This process map can then be exported to a modeling tool (plug-in) to create a simulation. Using Generative AI to interpret and summarize system logs gives us a comprehensive view of an end-to-end business process. The human partnership with Generative AI enables us to map the simulation of a future state.
Applying generative AI to process summarization, process analysis, and process modeling enables a new, adaptive, and user-friendly way to create process innovation. And that’s just the start.
This post is adapted and condensed from two pieces originally published in the ILINX User Community. Want early access to more content like this? Join the group.