Why Trust, Not Tooling, Is The Constraint
Getting Beyond Pilots: From Embedded Analytics To Agentic Workflows
experimental tools to trusted infrastruc- ture. “Leadership in the AI era requires a delicate balance,” he reflects, “cham- pioning innovation while managing risk, developing AI literacy while maintaining human judgment, driving transforma- tion while building trust.” That human- centric discipline shows up in several ways that matter for different parts of the C-suite: • For CFOs and treasurers: AI is po- sitioned as a partner that escalates stra- tegic decisions, not a replacement for judgment; success is measured by how effectively AI augments human exper- tise, not how much can be fully auto- mated. • For CTOs and CIOs: transparency and auditability are first-class design cri- teria; Kyriba customers can trace how an agent used their data to arrive at a rec- ommendation, supporting compliance and internal audit. • Broadly across finance and trea- sury teams: skills development focuses on prompt engineering, scenario design, and critical review of outputs, moving teams from passive users of “magic but- tons” to active designers of AI-enabled workflows. Callway is candid that this skill-build- ing takes time, “It took me a good year to build the confidence to really make the LLM sweat,” he says. “The quality of the questions you ask an LLM is di- rectly related to the quality of the out- put and that’s a skill finance teams have to build.” Kyriba’s roadmap also anticipates an ecosystem of agents. A world where or- ganizations and partners capture best-
For finance and treasury teams, trust is the gating factor between small pilots and enterprise-scale AI. Leaders are wary of handing critical workflows to opaque models when their core data, liquidity positions, risk exposures, and bank connectivity - sit at the center of organizational resilience. Callway’s definition of trust is con- crete: data must stay within the platform boundary, behavior must be observ- able, and outcomes must be auditable by humans who own the risk. Rather than invoking AI through a generic API to a public LLM, Kyriba embeds its LLM within its own environment, so customer
Most organizations’ initial AI steps have centered on what Kyriba calls “embed- ded AI”: pattern-matching and machine learning that quietly improve forecast- ing or anomaly detection inside existing workflows. That phase delivers value, but it rarely changes how work is organ- ized or how decisions move through the enterprise. The next stage is agentic AI - configur- able agents that execute bounded work- flows such as cash-flow management, reconciliation, or liquidity analysis, with human validation at each step. Kyriba’s trusted agentic AI: TAI, began with preb- uilt agents that function as a treasury as- sistant, collapsing multi-click processes into conversational interactions. “We’ve seen six-click workflows become five- second experiences,” Callway notes, “and that’s where CFOs really start to feel AI in their day-to-day work.” For finance and treasury leaders, this is the practical path beyond pilots: start with narrowly scoped, high-value use cases where agents augment exist- ing workflows and every action can be inspected and approved. For IT lead- ers, the agentic model offers a way to encode domain expertise into reusable components, then govern them through the same lifecycle and controls applied to other critical applications. Human-Centric AI: Building Oversight, Skills, and Ecosystems A striking theme in Kyriba’s journey is how human-centric the successful AI adoption story must be. Kyriba invested early in a co-innovation lab with custom- ers, organizations such as Koch, Sodexo, and Mews tested model AI scenarios in real treasury environments before it was broadly rolled out. Callway emphasizes that human over- sight, domain experts validating outputs, challenging edge cases, and shaping how agents behave is essential to move from
“To build trust, we had to architect AI so treasury leaders could see exactly how their data is being used.”
data does not leave the Kyriba platform and customers can see how their data is processed. “We knew our custom- ers wouldn’t be happy with their highly proprietary data going out to third par- ties and being processed in ways they couldn’t see,” he explains, “so architect- ing AI inside Kyriba was foundational to building trust.” For CFOs and treasurers, that architec- tural choice speaks directly to fiduciary responsibility and risk appetite: the AI must be treated like a 24x7 assistant that operates inside existing control frame- works, not an uncontrolled external ser- vice. For CTOs and CIOs, it reframes AI adoption as a platform decision about where models run and how data sover- eignty, observability, and compliance are enforced end to end.
“Turning a six-click workflow into a five-second experience is where CFOs start to feel AI in their day- to-day work.”
44
Made with FlippingBook - Share PDF online