Bank AI Adoption Stalls on Human Fears, Not Technology

Bank employees raise the same three objections to artificial intelligence in every internal survey: "I'm afraid of losing my job. I don't trust AI and I don't have time to learn." Vikram Sivakumar, who works on change management for AI adoption in the banking sector, named those three fears during a panel discussion moderated by Chana Schoenberger, and said the ordering shifts from institution to institution but the list never does.

The stakes for banks are unusually high. Financial institutions operate on customer trust, face tightening scrutiny of third-party and model risk, and now confront a competitive penalty for moving too slowly on AI. The panel — Sivakumar and Kirti Choudhary, in conversation with Schoenberger — argued that the binding constraint on AI in banking is no longer the technology itself. It is the workforce.

How do banks weigh AI risk against the risk of standing still?

Sivakumar framed the risk question as a matter of ordering rather than appetite. "When you talk about AI, it's not about how do I operate at speed and safely, it's about how do I operate safely and at speed in that order," he said. Safe first, fast second.

He also pushed back on the notion that banks are uniquely exposed. "Banks are very good at managing risks by nature. I mean, that's why I bank with a bank," he said, adding that AI risk management is "just a new muscle to learn and completely expect them to do that." The counterweight, he argued, is competitive: "inaction is also a risk of some form."

Choudhary agreed that risk management is "first nature to banks" but flagged a posture problem — how to avoid being too conservative. She warned against holding AI models to a standard above the one applied to human employees doing the same job. "How do you not treat AI and models that it needs to sit above how a human would behave or an employee would behave in that same setting?" she asked.

Which risks worry banks most right now?

Choudhary named four specific areas receiving the most attention:

• Cyber threats driven by or targeting AI systems
• Fraud, which AI can accelerate as well as detect
• Third- and fourth-party risk (exposure through vendors and vendors' vendors as AI tools enter the supply chain)
• Operational risk from what models might do once they become customer-facing or bot-facing

That last category points to a structural shift. Schoenberger raised the scenario of a customer's AI agent negotiating with a bank's AI agent, with humans largely out of the loop. Choudhary acknowledged a liability question but argued against waiting for perfect answers: "we can either first try and solve every question in every edge case, or we can run some experiments, see how things shake out, learn from it and make some progress."

The takeaway: banks see the AI risk map widening from model performance to liability and supply-chain exposure.

Why is change management the hard part?

Sivakumar called the human side "the crux of the challenge." Deploying the tools, he said, is the easy half; the change must land "over a period of time," and the playbook "is not necessarily fully available for everyone to play out with."

His answer to the three recurring employee fears is two-sided. Banks need a structured environment in which teams can learn and work differently, and they need to show employees "that they have a job at the end of that transformation. It isn't about taking their job away."

He also assigned homework to leadership. Executives should "model the change and not mandate it" — down to being able to answer a town-hall question about a favorite AI use case "in authenticity and in detail." The change burden, he said, falls on frontline staff, middle management and top management alike.

Boards are now joining the conversation. Sivakuryar noted that directors increasingly ask management: what is the pathway for the workforce in the post-AI era, and how is the institution responsibly caring for its talent? He framed the mature version of that conversation as "an upskilling, re-skilling pathway rather than a, hey, we got here and we don't need people anymore." That workforce-transformation discussion, he said, needs to become far more explicit across the industry.

Does culture decide the outcome?

Choudhary's closing argument centered on organizational culture. Leaders must signal that they run "AI first institutions" that want progress — but "not at the cost of humans, but taking humans along on the journey."

The goal, in her framing, is a workforce that does not believe it is being replaced but "that they're being augmented and that there will be new jobs which are needed in the AI era." Culture, she argued, is what makes an organization rally behind that message rather than resist it.

What was left unsaid?

Neither panelist quantified adoption rates, spending, or headcount effects, and neither named specific banks or deployments. The transcript is a discussion of posture and process, not results. The claim that banks will simply learn the new risk muscle, as Sivakumar asserted, rests on analogy to past episodes rather than evidence from this one.

What the discussion does establish is where the friction sits. If the same three fears surface in every survey Sivakumar's team runs, the pace of AI adoption in banking will likely track how convincingly institutions can pair deployment with visible reskilling commitments.