In November of 2022, Open AI released the first public version of Chat GPT. It was also the first time most people had heard the terms “generative AI (Gen AI)” and “large language model (LLM).” Since then, a chaotic race to integrate LLMs—and the chat interface enabling users to communicate with them in natural language—into just about every tech product has ensued. As early as the Spring of 2023, intelligent automation providers were touting the addition of Gen AI into their platforms to make it easier for users to automate workflows.
In the intervening years, providers have honed the way Gen AI and LLMs are being used in their platforms enabling users to become more efficient builders of automation technology—in essence, automating automation. According to Dr. Lou Bachenheimer, CTO of the Americas at SS&C Blue Prism, Gen AI and other technologies available to automation providers have pushed the industry to an inflexion point as organizations move away from siloed technology deployments toward comprehensive platforms that are maximizing efficiency.
“The combination of artificial intelligence, orchestration and automation enable a broader, more scaled-out approach we call enterprise AI,” Bachenheimer told Automation Today in a recent interview. “It lets businesses shift away from focusing on cost cutting or process improvement and toward a scaled approach that enables improvement of key metrics or revenue growth. We’re at a point where we have flipped the automation question on its head. Where before organizations would say ‘what would we do if you could give every employee like an hour back in their day?’ Now it’s ‘what would you do if you had unlimited employees?’”
Scale Through Simplification
Bachenheimer, an AI veteran who worked on IBM’s cloud platform and on Watson before his move to SS&C Blue Prism, expresses excitement not only with the direction of the industry, but his company’s as well. The acquisition by SS&C more than two years ago has given Bachenheimer’s team access to resources that are now bearing fruit for the company that coined the term “RPA” in process automation’s infancy.
Generative AI, he says, has opened a broad range of automation use cases to SS&C Blue Prism customers. And it begins with the way the platform integrates the technology and enables users to engage with it. There are two ways Gen AI is available to the platform’s users. The first, he says, is the ability to connect to external LLMs through a digital exchange.
“We have a Generative AI library where we have pre-built connectors you can leverage to reach out to LLMs that are tailored to a range of specific industries,” he notes. “And it’s a nice, visual-based approach to call the tool you need. You don’t need to be a programming wizard and it opens up a host of use cases.”
The other half of the company’s Gen AI strategy is incorporating the technology into the platform itself to make all types of automations and business orchestration as simple as possible. According to Bachenheimer, Generative AI’s ability to translate natural language into automations is what will make the technology accessible to enough people within an organization to maximize ROI.
“By deploying our full platform internally at SS&C,” he explains, “one of the most important learnings we uncovered is the ability to scale the number of non-technical developers. Within the first 12 months, we scaled up more than 250 Blue Prism developers internally.”
While more traditional process automation focused on the backend, he says, leveraging the broader mindset inherent in enterprise AI and augmenting automation technology with Gen AI has provided non-technical users the ability to solve complex problems that can result in improved productivity and revenue growth.
“If you need to build out a calculation stage or complicated text manipulation, you can now, within the platform, describe in natural language what you want to do and it will at least provide you with a template, if not a custom automation,” he says. “Basically, Generative AI is being incorporated in every aspect of the platform to make citizen development easier and more intuitive.”
Mitigating Bias
While Gen AI undoubtedly will enable organizations to scale automation programs, Bachenheimer warns that with scale, risks begin to multiply and companies implementing intelligent automation technology must not only be aware of them, but also take steps to mitigate them.
Language—the second “L” in LLM—is inherently biased, he says, and the millions of words comprising the texts LLMs are trained on transfer those biases to the models, affecting their output in essentially unmeasurable ways (Bachenheimer calls them a black box when they make decisions).
“Since you cannot exactly tell how models are biased, you can’t audit the output easily,” he says. “But there are definitely ways you can investigate. We’ll audit all the information going in for each call and then audit everything going back. So, while the LLM itself might be a black box, we at least know what went in and what came out for each decision. We’ve got as much auditing as we can. We’ll also do some post-processing with the data that comes back. We’ll have automations examined for things like hallucinations or misinformation. We also have the ability to pull humans into that loop, if necessary, to look at the returns we’re getting and be confident they’re correct.”
In the end, Bachenheimer says, Gen AI is an incredibly powerful tool that, while important, is only part of the platform.
“The SS&C Blue Prism Enterprise AI platform we have now is comprehensive,” he concludes. “It comprises a world class RPA solution, a world class BPM solution, we can do intelligent document processing with some of the most accurate handwriting recognition in the world, there’s machine learning involved, there’s process mining, task mining, the ability to build up custom user interfaces and even human in the loop. There are all sorts of technologies that make up the platform, but the important piece is that it’s all unified. It’s all fully integrated. It all plays, plays nicely together at a level that you really can’t get when you’re connecting pieces from different vendors.”
