I’m not talking about the Johnny Rivers theme for a late Sixties Saturday afternoon spy TV show. We’re talking software agents. Some may be secret, but none are men.
I once had an annual meeting with the CTO of a large automation company where I shared (non-privileged) information I’d gathered about the market while trying to learn what technologies I should be watching for.
With artificial intelligence (AI) and Large Language Models (LLMs) grabbing the spotlight at center stage, I’m watching for what technologies will make something useful from all the hype.
I’m looking for a return to the spotlight of these little pieces of software called agents. John Harrington, Co-Founder and Chief Product Officer at HighByte, an industrial software company, believes in 5 or so years from now, LLMs won’t be the game-changer in manufacturing that many expect. Instead, Agentic AI is set to have a far bigger impact.
So, John and I had a brief conversation just before my last trip. It was timely due to the nature of my trip—to a software conference where LLMs and AgenticAI would be important topics—and not just in theory.
From Harrington, “Agentic AI is revolutionizing the tech industry by addressing AI’s biggest limitation—making decisions that are more human like. AI agents are yet another application that analyzes and turns large amounts of data into actionable next steps, but this time they promise it will be different.”
He told me that Agentic AI will become more “human-like” going beyond LLMs. HighByte started up as an Industrial DataOps play at a time when I was just hearing about DataOps from the IT companies I followed. I told the startup team that they were entering a good niche. They have been doing well since then. They extended DataOps with Namespace work and now LLMs and agents.
“AI agents can enhance data operations by providing greater structure, but their success depends on analyzing contextualized data. Without proper context, the data they process lacks the depth needed for accurate insights and decision-making,” added Harrington.
Take an example. An agent can be a way to contextualize data, model an asset. Working with an LLM trained on data specific to the application, it can ask the LLM to scan the namespace to see if there are other assets in the database. HighByte’s can work through OPC, and also works with Ignition from Inductive Automation or the Pi database. It looks for patterns and can propose options as the engineer goes in to configure the application.
Not shy in his forecast, Harrington says the future is agents. They can affect and act on data. They can reach out to a control engineer, operator, quality group. It’s a targeted AI tool focused on one small thing. Perhaps there’s a maintenance agent, or one for OEE, or line quality on a work cell. Don’t think of a monolithic code in the cloud. Rather, think of smaller routines that could even work together helping business like Jarvis in Iron Man. Data is food for these agents, and HighByte’s business is data.
I’ve been impressed with HighByte’s growth and sustainability. Also that they’ve managed to remain independent for so long. Usually software companies want to build fast and sell fast. Watch for more progress as HighByte marries agentic AI with data.
