Banks are no longer treating artificial intelligence as an experiment. In ProSight’s 2026 CRO Outlook Survey—just out this week—54% of respondents said their institutions have AI in production, and 48% expect to have AI deployed in risk in the next two years. The focus has shifted from testing to scaling—and from curiosity to control.
So far, the most successful use cases are practical ones. CROs are applying AI to automate anti-financial crime monitoring, improve report generation, strengthen quality assurance, and spot emerging risks earlier. Most banks are taking a cautious approach: only 16% want to be on the leading edge, while more than half describe themselves as “fast followers,” looking to expand what’s already working.
But moving from pilots to production comes with growing pains. Thirty percent of respondents said limited staff capabilities and training are holding them back from scaling AI, while another 27% pointed to problems with data quality and availability. Twenty-six percent said their AI risk frameworks remain too immature to support wider use. Governance, in particular, lags behind capability. Only 12% of CROs described their AI governance and approvals framework as “highly developed,” though large institutions tend to be further along than smaller peers (17% of larger banks vs. 10% of banks with less than $50 billion in assets).
That gap is a key reason banks are emphasizing human oversight. “We want to make sure everyone in the company understands the risks that may be present in using AI and not doing it blindly, ensuring there is a human in the loop,” one large-bank CRO said. About half of survey respondents view using AI without adequate verification as a top AI-related risk—a reminder that model performance is only part of the challenge. As another large-bank CRO put it, “I believe the way we manage risk today will fundamentally change by 2029. … You have to completely rebuild your risk framework.”
The takeaway: Scaling AI responsibly means building the infrastructure around it—trained teams, clean data, and clear governance—with care, like the models themselves. Do that well, and curiosity begins to give way to control.
For a deeper look at the full findings—including CRO perspectives on credit, regulation, and the macroeconomic outlook—read the complete report on the 2026 ProSight CRO Outlook Survey.
