Celonis banking head: Banks want proof before scaling agents

As industry leaders gathered in Miami last week for Sibos, a global financial services event, Celonis' SVP and Head of Global Banking, Chris Johnston, said the industry's launch of AI pilot projects is outpacing its ability to prove the technology works at scale.

He said the gap stems from three questions banks want answered before committing further investment: whether an AI agent genuinely improves a process, whether it reduces risk, and whether it saves money rather than adding cost.

He said those questions have become central to conversations with banking clients as institutions weigh moving AI pilots into production.

"Before we see things seamlessly operating at scale in regulated industries there needs to be more certainty of outcome," said Johnston. "It may be something more along the lines of 'what's the standard operating procedure for this type of escalation,' and 'how do we handle that in this new world of a hybrid human and agentic workers potentially trying to resolve this issue together?' We're providing context to help answer these questions for the banks."

As KPMG's June AI Quarterly Pulse Survey reports, banks are projected to spend USD $170 million on enterprise AI this year through the first quarter of 2027 on average in America. The question Celonis challenges is whether the large language models financial institutions use are specialised enough to handle the complex understanding required by banking regulations in a data-sensitive sector.

Johnston said a standard AI assistant might offer only generic guidance, such as checking service-level agreement timers or prioritising high-value transactions. He argued that a system combining process intelligence with AI can produce more specific, actionable output.

He added that Celonis positions itself as "agnostic" of any particular AI vendor or infrastructure provider, working across different agentic strategies and hyperscalers depending on what banks have already adopted. He said the company aims to build on a bank's existing systems and expertise rather than requiring wholesale changes to existing infrastructure.

The banking head explained that Celonis's banking work has traditionally centred on four areas: customer onboarding, servicing and middle-office operations, risk and regulatory compliance, and capital-markets trade operations, where it helps free up capital stuck in exception queues.

More recently, he said, Celonis has seen growing interest from Chief Information Officer functions in what the company terms AI readiness, covering how software development processes and AI-driven "agent mining" are converging as developers work within increasingly hybrid human and automated environments.

From there, the firm supports servicing and middle-office operations, including lending and account management, as well as regulatory compliance, where Celonis maps internal control frameworks against how a business actually operates to identify where processes break down.

"When you think of banking they always have to worry about risk so oftentimes regulatory risk and compliance is something that is hard for institutions to get their arms around because the rules change often," said Johnston, adding that while they are not set in stone and risk is always evolving, banks have to manage risk across multiple dimensions. "[With] Celonis, you can lay the control maps of these institutions over the way their business actually operates - having a much more foolproof and complete view of where breaks occur, and then the ability to mitigate and fix those breaks in real time."

 For Celonis, the pitch is that context is what turns generic AI output into something banks can act on and defend.