Artificial intelligence and machine learning roared on the scene in 2014 and 2015, with by mega-investments by tech leaders such as IBM and Microsoft, along with the obligatory accompanying hype. Now, the AI/ML market appears to be “backing off,” says Craig Le Claire, analyst with Forrester Research. Delivering his remarks at an event sponsored by Automation Anywhere, Le Claire points out that intelligent software bots are increasingly taking on the automation challenges of enterprises.
Indeed, the drive toward intelligent automation in all its forms is taking time to accept and adopt among enterprises. Only 17 percent of 600 CIOs responding to a KPMG survey acknowledge their firms have smart automation technologies operating at full scale, and 30 percent don’t even know where to begin. At issue are lack of staffing and technology resources, along with lack of organizational support and accountability.
The big-bang AI and machine learning projects may be stuck in pilots and proofs of concepts, but a less glamorous form of intelligent automation may be percolating its way through processes and channels across enterprises — software bots associated with robotic process automation (RPA). These bots usually take on single-purpose tasks, such as pulling data for a purchase order, or delivering an email confirming with a transaction. A majority of enterprises surveyed by Deloitte last year, 53 percent, report they have put RPA in place, with 72 percent expecting to do so within the next two years.
“Software running in obscure data centers that no one will ever see will transform or replace the jobs of cubicle workers, coordinators, and even knowledge workers,” Le Claire writes in his recent book, Invisible Robots in the Quiet of the Night. These “digital workers” are “the invisible robots that RPA platforms are building. Put simply, these platforms build bots that mimic what a human does on a computer. Repetitive tasks, such as posting data to a financial application, are easily programmed into the bot.”
Here’s the kicker: “Unlike many machine learning projects that may need data scientists, RPA is simple to deploy.”
Enter the software robots that RPA is now enabling. “The most disruptive form of AI will be software robots,” says Le Claire. “Some will combine robotic process automation and different AI components. Others will simplify the algorithms running in machines that make decisions that humans used to make. Still others will be chatbots that help solve customer problems. By 2025, most workers will have a personal robot, one they can configure by themselves through text, using emerging forms of self-configuring machine learning.”
While physical, mechanical robots, as well as artificial intelligence initiatives, receive the lion’s share of attention as the path to intelligent automation in all its forms, RPA and associated software bots have been gradually being adopted for tasks across enterprises, operating to a large degree under the radar. Some organizations may have hundreds, if not thousands of software bots operating behind the scenes. And they are increasingly adding intelligence as the amount of data generated and consumed by them grows.
The benefits of RPA are tangible, the Deloitte study found. Payback was reported at less than 12 months, with an average 20 percent of full-time equivalent (FTE) capacity provided by robots. Benefits cited include improved compliance (92 percent), improved quality/accuracy (90 percent), improved productivity (86 percent), and cost reduction (59 percent).
However, as Le Claire cautioned in his recent presentation, there are headwinds that may slow the progress of RPA and software bots. For example, building bots, especially in organizations with hundreds, or even thousands of these bots, may be incurring maintenance costs and headaches. “A bot with 30 decisions and 8,000 clicks may require two to three people to maintain,” he states. In addition, “companies today don’t know what automation is costing. A company may save 8,000 person-hours with automation, but what happens with those hours?”
Le Claire predicts that increasing use of AI in conjunction with software bots, but these efforts are still nascent. “It’s still hard to find really good implementations of RPA and AI,” he states. Still, progress is notable — RPA is moving beyond “simple recording of human clicks,” and bots are increasingly generating and consuming data around human-machine interactions.”
These software bots are, for all intents and purposes, part of an emerging “digital workforce” that will take on much of the grunt work of today’s office and knowledge workers — and should be treated accordingly. “Even today, experts in software robotics recommend we govern digital workers, software bots that perform work for humans, in the same fashion as human workers,” Le Claire writes in his book. “This means tracking their hiring date (software creation work), and assigning a boss (responsible for design, training and securing the bot’s password access). Each boss will even have a performance review and termination date (when they’re taken out of service). In this way, governance and management of digital and human workers are converging.”