How Hudson Valley Credit Union's CAIO Trades Point Solutions For Unified Intelligence Layer

Enterprises no longer have an AI adoption problem. They have an AI fragmentation problem. After two years of pilots and experimentation, most organizations have proven AI can create value. The question now is whether that value can scale and sustain beyond isolated use cases. For AI leaders, the next mandate is to move from point-solution sprawl toward a unified intelligence layer that connects AI to how the business actually operates.

Preetha Sekharan is the Chief AI Officer at Hudson Valley Credit Union. She joined in early 2026, bringing a track record of translating AI into measurable business outcomes at Unum, where her applied AI and digital incubator teams have delivered real business impact, driving operational efficiencies and business growth. Earlier in her career, she led large-scale healthcare and financial services transformation programs at McKinsey & Company and Booz & Company. Her current focus is building the operating model and execution discipline that allows AI to scale across the enterprise.

"Point solutions can prove AI works, but they don’t scale value. Without a unified intelligence layer, the impact remains fragmented and short-lived. Organizations that win will architect AI as platform capability, not deploy it as a series of disconnected tools," Sekharan said. "You have to think about a unified intelligence layer that compounds over time."

That distinction frames the central challenge for enterprise AI leaders. Point solutions can prove that AI works, but they don’t automatically create the connective tissue needed to change organization at scale. A tactical chatbot, workflow automation, or decision-support tool may improve a sub-process, but the compounding value comes when intelligence is organized as a shared layer across systems, individuals, teams and enterprise. That is where AI shifts from isolated productivity lift to transformative change.

• Blaming the mirror: Building that unified intelligence layer first requires an honest look at what the experimentation phase left behind. Sekharan noted that AI doesn’t create new problems; it just exposes existing ones. Organizations that struggle to measure value, enforce discipline, or drive change will see those same gaps amplified with AI. Without operational rigor, AI doesn’t fail quietly. It makes the failure visible. "AI puts a spotlight on what is not working already in your organization," Sekharan said. "So if there is a debt around data, value realization, processes, AI will showcase that gap." The same discipline gaps that slowed previous transformation efforts show up again, accelerated, when AI is layered on top. 

• Building the muscle: "AI is not just a technology shift. It’s a workforce transformation. Roles change, decision-making shifts, and new capabilities must be built. Organizations that treat this as a tooling exercise will stall; those that invest in change management will scale," she added. "People's roles will be different. Processes will be different. New muscles will have to be built. And those are critical to realize value at the other end." At Hudson Valley, that meant recognizing early that the credit union would need to build the change muscle.

Governance is the foundation. Rather than treating governance as a brake on experimentation, Sekharan is standing up a governed, federated execution model through a cross-functional team spanning audit, information security, enterprise risk management, legal, and compliance, framing it as the structure that lets employees act with confidence. Effective AI governance functions as a growth enabler when it gives teams clear rules of engagement rather than reasons to hesitate. A low-stakes chatbot does not need the same scrutiny as a high-risk financial model, and in regulated environments like financial services, that distinction matters.

• Safety net, not brake: "Not all AI carries the same risk, and governance must reflect that. Low risk use cases should move with speed; high-risk applications demand full rigor. The key is a unified inventory and risk classification model that ensures consistency across the enterprise," she said. "Governance is not a control mechanism. It is a scaling mechanism. At Hudson Valley, we are building a federated, risk-based governance model that allows teams to move quickly within clear guardrails. The goal is consistency, not constraints”

The principle carries over directly from her time at Unum, where deploying machine learning for live claims processing taught her that lab environments miss the edge cases that only appear in production. Her rules have not changed: pilots must run with real operators on real transactions, or the results cannot be trusted. Organizations that skip this step fall victim to the last mile problem, where controlled successes fail to survive contact with actual operations.

• Escaping the lab: "Pilots that run outside production don’t generate real signal. AI only proves its value under real operating conditions, where edge cases and complexity emerge. If it hasn’t been tested in production, it hasn’t been tested," she said. At Hudson Valley, that means a strict rule: pilots run in production with real operators. Proof of concept shows something can work. A production pilot shows it will.

All of it connects back to a single economic constraint: Growth without operational scale creates bottlenecks. Sekharan noted that the only sustainable path is to decouple growth from headcount by scaling through technology. That requires focusing on a few high-impact value streams—not dozens of disconnected automations, which is why she is concentrating on a handful of end-to-end value streams rather than chasing dozens of small task automations. For financial services leaders like Sekharan, lending is at the top of that list.

"Whether it's lending operations, mortgages or home equity loans, there is an opportunity to look at that value stream end to end, compress the decision times, make the process efficient and effective and remove friction for the members," Sekharan explained.