When it comes to many innovations across payments, hype can often outpace practical deployment.
But even against that backdrop, the emergence of agentic artificial intelligence might stand among the most transformative moments in enterprise automation, at least since the invention of APIs.
Unlike older AI systems, which require frequent human prompting and contextual resets, AI agents take the initiative. They perform tasks, update their own context, and communicate with other agents to complete complex workflows all with minimal human involvement.
“We’re no longer just orchestrating our actions,” Stax Chief Technology Officer Mark Sundt told PYMNTS during a conversation for the June edition of the “What’s Next in Payments” series, “What’s Next in Payments: Secret Agent.” “We can now orchestrate the actions of AI.”
Reflecting on his experience across legacy technology firms like IBM and Microsoft, Sundt admitted even he is stunned by the pace of transformation.
“The velocity of change is absolutely astounding,” he said. “The half-life of information in AI right now is probably measured in weeks.”
Agentic AI is evolving and deploying faster than cloud computing and client-server models, which took years to gain traction. This is not merely automation; it’s delegation. It represents a philosophical shift in how businesses think about AI in the workplace.
Orchestration Layers
The journey into agentic AI began like many others, with AI tools operating essentially as advanced search engines or static content creators, Sundt said. However, that changed quickly as enterprise users demanded more from their models.
“It’s always been creating content, but it hasn’t updated the content it created for us,” he said. “And now we have the capabilities of doing that.”
This difference is not semantic. It is structural and strategic.
“We can ask it to do something on our behalf, and we define what done looks like — and it works on it until it’s done,” Sundt said. “So, it’s like these parallel processes that can run and be working on our behalf, which is just mind-bending to see.”
Still, if agentic AI is the engine, orchestration is the transmission. Without a central conductor, even the most capable agents act in isolation.
“You’ve got agents to agents … but who’s driving the process?” Sundt said. “Who’s doing the orchestration?”
The solution is a specialized “orchestration agent” that coordinates other agents’ activities much like a project manager, he said.
“It might do things in a different sequence than I would’ve designed or thought about — but someone’s got to drive that process,” he said.
Orchestration also brings structure to flexibility. Using agentic AI, tasks don’t have to follow a rigid order, but when sequence matters — such as booking a car after confirming a flight — agents can reason about interdependencies and adapt in real time. The resulting shift is democratic. AI is no longer a tool just for technologists; it’s a business assistant, strategist and executor rolled into one.
Bridging the Risk-Adoption Gap
Despite this promise, not all leaders are ready to leap into the agentic future.
“There’s a split between, ‘Agentic AI is the future — let’s go now,’ and others who are like, ‘Eh, we need to pump the brakes on this,’” Sundt said.
He attributed much of the hesitation to questions of governance, infrastructure and control.
“People are struggling with how to actually have uniform context across an organization,” he said.
However, new developments are addressing that.
As third-party AI services proliferate, companies must demand transparency and accountability. Sundt pointed to the importance of having vendors who can show audit logs, confirm the masking of personally identifiable information (PII) and maintain standards like SOC 2 compliance.
“We’re just now starting to have these discussions,” Sundt said, but he added that preparing for AI compliance should be seen as an essential part of adopting the technology.
What are the strategic takeaways for companies still assessing their position on agentic AI?
First, treat context as the foundation, he said. Whether in customer interactions or compliance, well-structured context is the difference between generative AI and truly agentic AI.
Second, empower orchestration. If tasks are getting done in silos, businesses miss the value of end-to-end process management. Define a clear orchestrator — human or digital, he said.
Third, democratize usage.
“This isn’t just for developers,” Sundt said. “Anybody could write this markdown file.”
The power of agentic AI lies in its accessibility and scalability.
Sundt recommended beginning with low-risk, high-impact projects — especially in compliance-heavy industries like payments.
At Stax, a strong early win came from using agentic AI for merchant verification during onboarding. The process involved checking utility bills and phone records for consistency.
“We can very quickly automate [this],” he said. “In the past, those types of workflows would use OCR and required pixel perfection. … Now, AI can extract and reason over jumbled data and still produce accurate context.”
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
