Artificial intelligence is everywhere. Generative AI (genAI), of course, is supposed to change everything for everybody all the time. Yes, there is definitely a wave of almost unprecedented hype. But the gains are real, although the journey can be a rough one.

Amardeep Modi, vice president of Everest Group, said AI has become intelligent automation. Functions such as the user experience, form development, conversational interface development, API automation, intelligent document processing, process mining, task mining, process design and execution, business rules management, and workforce management can all be enhanced with AI. AI is going to be implemented across every aspect of the network and across every IT system in the near future.

“AI makes each function smarter,” Modi said. “AI can be viewed as the brain and RPA [robotic process automation] as the arms and legs that get things done.”

Particularly for industrial companies that focus on equipment manufacturing and servicing, the transition to AI can be far from easy. Here is how one company is dealing with AI and genAI deployment in its operations, and how it is rolling it out across its network.

Founded in 1874 in Milwaukee, A. O. Smith Corporation is one of the largest manufacturers of residential and commercial water heaters, boilers and water treatment solutions. While its roots are traditionally based in hardware and engineering, the company has spent many years developing software and systems to enhance product value. Most recently, it has gone all in with AI-powered RPA.

While most people associate industrial automation with the physical robots used in automotive assembly lines and other manufacturing processes, RPA harnesses software bots that operate on the network to integrate many different actions. RPA software makes it easy to build, deploy, and manage software robots that emulate human actions and interact with digital systems and software. These bots can do things like understand what’s on a screen, complete the right keystrokes, navigate systems, identify and extract data, and perform defined actions.

RPA robots handle routine and repetitive tasks. Recently, they have been paired with AI's cognitive capabilities to streamline workflows, increase efficiency, and enhance productivity.

Diana Swain, business process optimization manager at A. O. Smith, said the company took its time rolling out AI in the enterprise. The company first identified three proof of concept (POC) projects in the finance department in February of 2021. One was low risk, one medium risk and one high risk. It took until the beginning of 2022 to go live on all three.

“The low- and medium-risk projects succeeded, and the highest risk one did not, but we learned from all of them,” said Swain.

Those lessons were applied to the next three projects, again in finance, and again split into low, medium, and high categories. They begin April of 2022. A team was created within IT to evangelize and manage the RPA program at the company. This was a necessary step to ensure that the operational technology (OT) side integrated well with IT, that network resources were being used correctly, and enough compute power was available for AI. By March of 2023, the results of better organization became clear. All three new projects were completed, this time with each surpassing expectations, said Swain.

Based on those successes, the program was expanded beyond finance to take in more divisions of the organization. In coordination with IT, customer engagement was prioritized using new standards and processes and applied to document understanding.

Software robots are used to extract, interpret and process data from forms, PDFs, images, handwriting, scans, checkboxes and more using UiPath Document Understanding. Whether they’re rotated skewed, or low-resolution and contain signatures or checkboxes, UiPath Robots can recognize their content. Pretrained machine learning models add a higher level of intelligence and interpretation as well as the ability to trigger actions based on content identification.

A.O. Smith began with 3,600 documents annually (3% of annual invoice volume) and is currently up to 61,200 docs (48%). In the next two years, we expect as much as 95% of our invoice document volume to be automated.

There have been many lessons learned from the AI automation work done to date. It quickly became clear that AI had to be part of every document project due to its ability to differentiate meaning and identify obscure characters.

“We were having lots of document problems due to bad handwriting, documents written in different formats and platforms, and legacy and new applications being used for invoicing,” said Swain. “We measured the success of our document understanding project based on time saved in invoice processing.”

Bots now transact invoice processes without human interaction. If someone intervened due to errors or inaccuracies, the bots are tweaked to eliminate the problem.

“Document understanding has already freed up 7,200 hours per year that can be used for more strategic and fulfilling work,” said Swain.

Another lesson learned concerned maintenance. It is one thing to design bots and automate processes and quite another to maintain them. Network connection and security changes may interfere with bot operations. The company turned to Ashling Partners as internal resources couldn’t cope with maintenance demands.

“We grossly underestimated how many changes would be needed for our bots and AI only added further to our maintenance needs,” said Swain.

Storing output files posed yet another challenge. Early decisions on storage proved problematic later and were hard to undo. Swain explained that decisions on where to store output files, what data to save and how to set up access worked well in finance but didn’t scale outside of that department. That led to having to go back and fix some things.

“We didn’t bring our IT team along early enough to address important issues such as security accesses and ERP system interactions,” said Swain. “While it is easy for business units to set up bots to automate things, problems will arise if you don’t work closely with the IT team.”

She called attention to executive buy in. Those in finance were fully on board and that proved a wise decision on where to begin. Based on results in finance, other areas came onboard.

There were hard calls to make on standardization to make automation work well. This was achieved by getting all the right people in the room for process mapping exercises.

“We managed to get everyone’s perspective on what the process actually was, not what it was supposed to be,” said Swain.

An Everest Group survey revealed that the key challenges hindering adoption of AI in the enterprise were the complexity of solutions and the scarcity of skilled AI talent. Those issues placed ahead of ethical implications (bias, hallucination, security, data privacy), reliability, and changing regulations.

It makes sense, therefore, for IT departments to engage early with whoever is working on the AI and gen AI rollout within the organization and ensure that the network and storage side are up to the task.

“Each industry and each company must develop their own AI strategy taking into account the various challenges and pitfalls and select the right partners to help them on the journey,” said Modi. “Tailor AI and RPA to you, not what others are doing.”