Editorial
The evolution of human intelligence is a domain of knowledge that has undergone intensive study over the years, and for which a large body of theory exists as a result.
Now, questions have turned to whether artificial intelligence (AI) might follow the same type of evolutionary trajectory.
A 2012 essay on Alan Turing’s work excerpted in The Atlantic explores the evolution of intelligence in computing, drawing parallels with the concept of Darwinian gradualism — the idea that evolution occurs in tiny increments over a period of time.
It theorized a type of gradualism emerging in AI’s evolution, arguing that structures for AI will become larger and “more competent” as time passes, reaching a point where the assembly is intelligent — or at least “intelligent enough” to power “competences that deserve to be called ‘comprehending’.”
Of course, AI is a broad church with many subdomains that vary incapability and complexity.
Most of the attention today is focused on one specific strain, generative AI, and the way it might evolve to augment current processes, workers and roles.
Gradualism in this space, it could be argued, is occurring on a significantly truncated timeframe at present, as the initial surge in interest and usage drives a considerable amount of experimentation. However, there are already suggestions that peak growth rates may have been reached and are in the process of leveling out.
The Broader View
To understand what generative AI brings to the table, we need to take a broader view of how “intelligence” has evolved more generally in business, and particularly intelligence focused on processes.
The evolution of capabilities in this space has improved over a period of years as opposed to months.
Businesses first introduced intelligence to process automation in a significant manner around six years ago. In this context, intelligence comprised rules-based workflows that were applied to automation. Business rules logic was used to handle repetitive tasks, recognize and deal with exceptions, or for simple decision-making that, up until that point, had been performed by people.
For example, when an internal business unit received a request, either from a different business unit or perhaps an external customer, intelligent process automation became the first point of call or triage, determining whether the request met conditions that required it to be routed to an executive for approval, or whether it was simple enough to be approved on-the-spot. Both paths ran automatically. This was considered intelligent for the time period, and interest and adoption rates reflected the efficacy of the approach, as well as the general enthusiasm in business for intelligence adoption.
In the proceeding six years, gradualism has redefined what businesses expect when applying intelligence to their processes.
When leaders talk about intelligent process automation today, they more than likely expect some form of AI to be providing the intelligence component.
As this occurs to a greater extent, the ”art of the possible” is evolving. All eyes and attention are presently focused on what step change in capabilities artificially intelligent process automation will produce, and where to apply them.
Related Article: Forge a Path of Least Resistance to AI and Automation
The Addition of Learning
A missing element from earlier iterations of intelligent process automation was the ability of the algorithm or engine to “learn” from every task it performed, and to iteratively improve based on those learnings.
By leveraging the historical knowledge base that’s been built up over time as a training dataset, and taking into account each additional job it handles in real-time, AI-augmented intelligent process automation creates a path to process improvement that has what a lot of people would understand today as “intelligence”.
“Learned” intelligence can be applied to ensure that the process is always in its best possible shape, and that there is a pipeline of improvements. This will naturally lead to better automation and the emergence of adaptive workflows.
With these continuously learning processes, the business units that rely on them will become more efficient, and decision-making will be more transparent, documentable and auditable.
Our Current Stage
The question that remains is: over what period will this occur? Gradualism is driving experimentation and improvement of the underlying technology, but we’re still in a relatively early age of understanding.
At present, AI is mostly being used to finesse existing processes and automations. This makes sense: these are the processes that are perhaps the best understood internally, have been mapped to some extent, are core to operations and/or have the highest rates of usage. The business units and domain experts that live and breathe these processes know how they run, and their current challenges and pitfalls.
The high level of internal knowledge around these processes marks them as strong candidates for additional improvement, because the business has a baseline for how they perform today, and so they can better understand whether AI augmentation creates an incremental improvement or a step change. But this will evolve and change with technology.
Conference
Sep
16
HR Tech Conference Las Vegas 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
Once the efficacy of AI-augmented intelligent process automation is better understood, application of the technology will naturally shift to more “greenfield” processes, which promises to bring the significant improvements in automation efficiency that many businesses want to reach.
Learn how you can join our contributor community.
