Automation’s latest arms race may be focused on agents, but Adi Kuruganti thinks the real battle is happening on less glamorous ground: orchestration, governance and process reliability.
As chief AI and development officer at Automation Anywhere, Kuruganti oversees product engineering, AI initiatives and developer strategy at a company that spent years rooted in robotic process automation but has repositioned itself around what it calls “agentic process automation.” In practice, he argues, enterprises are moving beyond isolated copilots and experimenting with systems that combine deterministic workflows with cognitive agents capable of making contextual decisions.
The market conversation around AI agents has shifted noticeably over the past year, according to Kuruganti. He says enterprises that were cautiously testing concepts in 2025 are now putting production workloads into live environments, particularly around operational business processes.
“I definitely see a big difference compared to last year,” says Kuruganti. “Last year customers were still dipping their toes. We had about 1,500 deployments at that point in time, but they were early deployments, not actually in production,” he says. “Since then, it’s grown. We have about six million agent executions now in production. Not pilots.”
Kuruganti explains that many of those deployments are tied to narrowly defined operational outcomes in industries such as healthcare and finance, where organizations are pairing AI agents with more structured automation layers rather than replacing existing workflows outright. He argues that the market is moving toward hybrid automation architectures instead of purely autonomous systems.
Kuruganti sat down with Automation Today at Automation Anywhere’s recent Imagine event in Dallas as the software provider made a raft of product announcements. Notably, the company launched several of the industry-focused agentic solutions Kuruganti touts as the future of process automation along with a claw-style system enabling organizations to deploy agents that can execute securely and reliably across varied systems and networks.
Reliability Is Becoming the Real Enterprise AI Battleground
While much of the public AI discussion still centers on model intelligence and reasoning benchmarks, Kuruganti says enterprise customers are increasingly focused on something more practical: whether agents can consistently complete business processes without introducing operational risk.
According to Kuruganti, enterprises are paying close attention to issues such as drift, governance and process adherence, particularly in heavily regulated industries. To that end, Automation Anywhere has been building evaluation frameworks that monitor model and agent behavior in production environments. Those controls, he says, are becoming necessary as enterprises move AI deeper into essential systems.
“We’re talking about mission-critical areas,” says Kuruganti. “AI is touching healthcare systems and financial systems. It must work.”
Kuruganti also describes the company’s broader effort to improve agent reliability through what it calls a process reasoning engine. The platform, he explains, uses metadata derived from hundreds of millions of automated processes to improve contextual decision-making over time. He claims the system helps agents reuse prior process context rather than treating every workflow execution as an isolated task.
Betting on Interoperability
One of the recurring concerns in the enterprise AI market is the growing number of vendor-specific agents embedded inside individual software ecosystems. Organizations adopting AI tools from CRM, ERP and IT vendors increasingly face fragmented automation environments where systems operate independently.
Kuruganti argues that enterprises do not want another disconnected AI layer added to existing technology silos. Most enterprise business processes already span multiple systems, making orchestration a more important capability than building standalone agents tied to a single application stack.
“No process lives in just one application,” says Kuruganti. “Pretty much 100 percent of the processes that Automation Anywhere executes run across applications.”
He says the company’s strategy is centered on becoming an orchestration layer capable of coordinating both deterministic automations and third-party AI agents regardless of origin.
“We obviously have our own agents where customers can build agents on our platform, but we also know our customers might be using an agent built on AWS or Salesforce or ServiceNow,” says Kuruganti. “We welcome that.”
Kuruganti frames that interoperability approach as increasingly important for enterprises adopting multiple AI platforms simultaneously. Rather than attempting to standardize around a single vendor, he says organizations are assembling collections of specialized tools and agents across departments.
That dynamic, he explains, is influencing both product architecture and release strategy inside Automation Anywhere itself.
Faster Release Cycles Are Reshaping Enterprise Automation Development
Kuruganti says the accelerating pace of AI development is forcing enterprise software vendors to rethink traditional release schedules and experimentation models.
“We used to be on a quarterly release,” says Kuruganti. “We’re now on a monthly release.”
Customer expectations around AI innovation, he notes, now require more continuous iteration, even for enterprise platforms traditionally associated with longer deployment timelines. At the same time, however, Kuruganti says the company is trying to balance rapid experimentation with the operational caution expected by large enterprises and regulated industries.
One example of that experimentation cycle, according to Kuruganti, was the company’s rollout of Enterprise Claw, intended to operationalize autonomous agents within governed enterprise environments. He says the initiative emerged after the company saw growing interest in autonomous desktop-style agents but recognized that enterprises needed additional security and governance controls.
“The agent has to go where the data is,” says Kuruganti. “Sometimes that means interacting within a secure system. They don’t just want an agent; they want to pull protected data from their healthcare systems into the cloud so that an agent can run.”
The company collaborated with Cisco, NVIDIA, Okta and OpenAI on the project that enables businesses to build AI agents that work across enterprise teams and workflows while maintaining centralized control over access, activity, and observability.
Looking ahead, Kuruganti suggests the broader enterprise market may need to become more disciplined about separating operational outcomes from technology trends. He says organizations are still being flooded with new AI announcements, model releases and agent frameworks, but he believes the more important question is whether enterprises can translate those tools into measurable business processes.
“I think about outcomes,” says Kuruganti. “The technology is there and the technology continue to evolve superfast, but we have to have the outcomes in place.”
