'What if?' Imagining the future with generative AI tools

According to FreedomPay's performance guru, GenAI will let IT ops pros experiment with ideas for performance optimization and disaster recovery.

From where one enterprise observability leader sits, generative AI tools and agentic AI represent a hypothetical future -- and that's a good thing.

IT ops teams at FreedomPay aren't strangers to AI -- the digital payments provider already uses Dynatrace's built-in Davis AI for predictive analytics and root cause analysis. But as GenAI tools and agentic AI emerge, they have the potential to open up new IT automation scenarios based on what-if questions, according to Mark Tomlinson, senior director of performance and observability at FreedomPay.

Mark Tomlinson, senior director of performance and observability, FreedomPayMark Tomlinson

"Traditional predictive analytics ... is based on known data, not potentially fictitiously rendered, according-to-a-model type of data -- a what-if scenario," Tomlinson said during an interview on Informa TechTarget's IT Ops Query podcast. "[Generative AI] really changes the definition of what it means to say, 'What if' -- like [for] disaster recovery, 'What if this machine got turned off?'"

For now, it's still very early in that story for large language models -- Dynatrace engineers are working on refining LLMs for just such a purpose, to understand what production code does and how changes to it will affect systems built with it. At FreedomPay, the strict compliance rules that come with being a financial services company limit generative AI use to a copilot for novice observability users so far.

[Generative AI] really changes the definition of what it means to say, 'What if' -- like [for] disaster recovery, 'What if this machine got turned off?'
Mark TomlinsonSenior director of performance and observability, FreedomPay

It will also take time for human workflows to adapt to GenAI, Tomlinson said.

"I don't think we're quite ready for the [effect on] human processes," he said. "The mitigation to that is less of a soft skill coaching or understanding of an organizational issue, and more, 'What concrete things do I do, given the fact that, just like every other piece of code on the planet, this stuff is fallible?'"

Still, Tomlinson said, even if they are confined to analyzing system metadata, generative AI tools will lead to advances in business analytics.

"For us, there's a lot of growth [coming] in that metadata space -- reporting, understanding the business, understanding globally what a merchant company is doing," he said. "Observability tools give us this incredible, robust, rich feature data to build [business analytics] models ... I can see that happening before we ever go deep into the core of what we do."

Even longer-term, Tomlinson expects further advances in auto-remediation as AI agents emerge that can work together to automatically execute multistep workflows in response to events, such as an IT outage.

"Agentic AI is ... almost like an organism that learns from things that you didn't even ask it to remediate," he said. "It might say, 'You might not see that as a problem, but I'm seeing ... a problem. What if ... we bypass a component this way and send [traffic] in this direction?' And now we start talking about automatically releasing features for [A/B] testing ... so auto-remediation becomes auto-identification of problems that you didn't even think you had."

Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.

Dig Deeper on IT systems management and monitoring