A Bank Just Built a Team Whose Whole Job Is Deciding Which AI Tools Are Worth Adopting
Valley National Bank announced this week the launch of Valley Foundry, a new internal unit built for one specific job: evaluating emerging technology, mostly AI, cybersecurity, and data tools, faster and more consistently than the bank's existing structure allowed. Industry veteran Michael Rappaport was named to lead it, reporting up through the bank's technology and innovation functions.
The reasoning behind it, as chief operating officer Russ Barrett described it, is worth sitting with regardless of your institution's size. Valley already had several channels bringing in outside technology and startup relationships: a tech and international banking team, a corporate venture capital arm, and a partner banking group working with fintechs directly. Those channels were surfacing plenty of vendor opportunities. What Valley didn't have was a consistent structure to actually evaluate them once they showed up. Barrett said the bank assesses 12 to 15 outside opportunities a year currently and expects to roughly double that pace with Foundry in place.
The distinction worth borrowing: evaluation is a different job than investment
Barrett was specific about how Valley Foundry differs from Valley Ventures, the bank's venture capital arm. Ventures invests in and supports early-stage companies. Foundry brings what Barrett called a technology-based lens instead, staffed by engineers and practitioners who actually test the technology inside the bank, not investors deciding whether to fund it. That's a distinction most smaller institutions collapse into one informal process, usually whoever in IT or operations happens to take the vendor call.
Why this matters well below Valley's size
No credit union needs a dedicated innovation unit with a named executive to take the real lesson here. The lesson is that vendor evaluation is its own discipline, separate from the decision to buy, and treating it as an afterthought is exactly how institutions end up with AI tools that got approved because a rep gave a good demo rather than because anyone consistently checked what the tool actually does with member data, how it was validated, and what happens when it's wrong.
A smaller-scale version of what Valley built doesn't require new hires. It requires a short, repeatable evaluation process applied every time, not just when someone remembers to ask hard questions: what specific problem does this tool solve, what data does it touch, how would you know if it started performing worse than expected, and who internally is accountable for the decision to bring it in. Write that down once, as a one-page checklist your team runs before any AI vendor gets approved, and you've captured the actual value of what Valley just built a whole department around.
Sources