What Filene’s latest research and our own AI BASECAMP data both suggest
Summary: Filene Research Institute recently published research arguing that AI agents are already comparing and selecting financial products on behalf of consumers, and that most credit unions aren’t structured to be part of that comparison. Filene’s recommendation is to start building three things now: AI-legible product and rate data, fraud detection that can tell a member’s AI agent from a malicious bot, and clear permission structures for what an agent can do on a member’s behalf. Wide Open Ventures ran its own survey through AI BASECAMP, gathering responses from 123 credit union executives across eight departments. The results show why Filene’s recommendations are harder to act on than they sound: leadership-wide AI knowledge averages 3.02 out of 5, IT readiness sits at 3.13, and 73 percent of technology leaders already anticipate integration challenges with existing systems, before agentic commerce enters the picture. The infrastructure Filene describes touches marketing, IT, and risk at the same time. Our data suggests most credit unions aren’t yet aligned enough across those departments to build it.
The Scenario Filene Is Describing Isn’t Hypothetical
Filene Research Institute published a piece earlier this year making a fairly direct argument: AI agents are already comparing and selecting financial products on behalf of consumers, and most credit unions don’t currently show up in that comparison at all. The infrastructure for this isn’t a future state. Visa has launched its Intelligent Commerce program with more than 100 partners, Mastercard has enabled agent-based payments across every U.S. issuer, and OpenAI and Stripe have built a shared protocol for AI-driven commerce, with Google backing a competing standard alongside Walmart, Target, and other major retailers. Filene also points to consumer behavior catching up quickly, noting that more than half of U.S. consumers used AI tools to shop during the most recent holiday season.
Filene frames the risk through two scenarios worth sitting with. In one, a member asks an AI agent to find the best personal loan rate. The agent pulls structured data from every lender whose rates and terms are machine-readable, compares them in seconds, and the member never sees a list that includes a credit union whose product data isn’t structured for that kind of lookup. In the other, a credit union’s fraud system flags a member’s own AI agent as suspicious activity, blocks it, and the agent effectively learns to route around that institution going forward.
Filene’s recommendation is to start building three things now: making product and rate data legible to AI agents (what they describe as AI Engine Optimization, distinct from traditional SEO), building fraud detection that can tell a legitimate member-authorized agent from a malicious bot, and establishing clear permission structures for what an AI agent is allowed to do on a member’s behalf, including spending limits and approval thresholds.
What Our Own Data Says About Where Credit Unions Actually Stand
We ran our own survey earlier this year through AI BASECAMP, our executive alignment program, and heard from 123 credit union executives across eight departments: CEO, Operations, Technology and IT, Risk and Compliance, Marketing and Member Experience, Lending and Credit, Finance and Accounting, and HR.
A few numbers from that survey are worth holding up against Filene’s recommendations. Executives are reasonably comfortable discussing AI in general terms, rating their personal comfort at 3.97 out of 5. But when asked to rate the actual level of AI knowledge across their leadership team, that number drops to 3.02. Clarity around how leadership has communicated AI’s purpose and expectations came in at 3.31, and executive team alignment on AI’s role came in at 3.19. None of these numbers are disastrous on their own. But they describe an industry that is comfortable with the topic of AI without necessarily being coordinated about what to do with it, which matters quite a bit when the work in question requires multiple departments to move together.
Mapping Filene’s Three Priorities Onto Our Department-Level Data
Filene’s three recommendations, legibility, fraud detection, and permissions infrastructure, each land on a specific combination of departments, and our survey gives a reasonably granular picture of how those departments are positioned.
AI legibility is fundamentally a marketing and IT problem. Product and rate data has to be structured in a way that’s both accurate and machine-readable, which means marketing and IT need to be working from the same source of truth. On the marketing side, our data is encouraging: Marketing and Member Experience had the highest departmental readiness score in the entire survey at 3.41, and 82.35 percent of marketing respondents identified behavioral segmentation and targeting as a top AI opportunity, with 70.59 percent pointing to predictive personalization. Marketing also overwhelmingly recognizes that this isn’t a solo effort: 100 percent of marketing respondents said Technology and IT should be a close collaborator on AI deployment. The catch is that IT’s own readiness score came in lower, at 3.13, and 73.33 percent of IT respondents said they anticipate integration challenges with existing systems. Marketing may be ready to think about legibility. Whether the underlying systems are ready to support it is a separate question.
Agent-aware fraud detection lands squarely on Risk and Compliance, and this is where our data produced one of the more striking numbers in the whole survey. 100 percent of Risk and Compliance respondents identified fraud detection and prevention as a top AI opportunity for their department, the only unanimous response to any opportunity question across all eight departments. At the same time, Risk and Compliance readiness came in at 3.29, and 71.43 percent of respondents said they foresee regulatory or compliance risks associated with AI adoption. The department most focused on fraud is also the department most aware of how much regulatory uncertainty surrounds this work.
Agent permissions infrastructure, which by definition cuts across IT, Risk and Compliance, and Operations, is where the cross-departmental piece becomes unavoidable. 90 percent of IT respondents and 71.43 percent of Risk and Compliance respondents both rated Operations as an essential collaborator for AI deployment. The departments know they need each other for this kind of work. What’s less clear from our data is whether the leadership-level alignment exists to actually coordinate it, particularly given that 61.9 percent of executives across the whole survey cited leadership alignment and understanding as one of the top three barriers to AI adoption generally, ahead of technology infrastructure, employee training, and regulatory constraints.
The Gap Isn’t Awareness. It’s Sequencing.
95 percent of executives in our survey rated cross-departmental collaboration as essential or important to successful AI adoption, about as close to consensus as a survey question gets. But department-level readiness scores cluster fairly tightly in the 3.0 to 3.4 range across the board, with Finance and Accounting as the one notable outlier at 2.40. Everyone agrees that AI adoption has to be coordinated. Almost nobody is currently positioned to coordinate it.
This is the gap that sits underneath Filene’s recommendations. Building AI-legible product data, agent-aware fraud detection, and permission infrastructure are all reasonable things for a credit union to start working on. But each of them requires departments that don’t typically coordinate closely, marketing and IT, risk and IT, IT and operations, to work from a shared understanding of priorities, definitions, and ownership. Skipping that step tends to produce exactly the kind of disconnected pilot projects that look like progress on a roadmap but don’t add up to anything coherent six months later.
Where AI BASECAMP Fits
AI BASECAMP exists for this stage of the process, before any of the infrastructure work Filene describes gets underway. It’s an executive alignment engagement built around structured surveys, department-level conversations, and working sessions, with the goal of getting a leadership team onto common ground about what AI means for their institution and how departments that need to work together actually will. It doesn’t produce a vendor shortlist or an implementation roadmap. What it produces is the shared understanding that makes those downstream decisions easier to get right.
If Filene’s research describes where the puck is going, AI BASECAMP is designed to help a credit union figure out whether its departments are currently positioned to get there together, and what needs to happen first if they aren’t.
Frequently Asked Questions
What does Filene mean by “AI legibility” for credit unions? Filene’s research argues that AI agents acting on behalf of consumers, comparing loan rates, choosing payment methods, and so on, rely on structured, machine-readable data rather than traditional websites. If a credit union’s product and rate information isn’t structured for that kind of lookup, it may not appear in an AI agent’s comparison at all, regardless of how competitive the actual product is. Filene describes this as “AI Engine Optimization,” distinct from traditional search engine optimization.
How does Wide Open Ventures’ AI BASECAMP survey data relate to this? Our AI BASECAMP survey of 123 credit union executives across eight departments shows that the departments most relevant to Filene’s recommendations, Marketing, IT, and Risk and Compliance, each have specific readiness gaps. Marketing has the highest departmental readiness score in our survey (3.41) and clear interest in personalization and segmentation, but IT readiness is lower (3.13) and 73 percent of IT respondents anticipate integration challenges. Risk and Compliance unanimously identified fraud detection as a top AI priority, but also reported significant regulatory uncertainty.
Which departments need to be involved in preparing for agentic commerce? Based on Filene’s three recommendations, AI legibility, agent-aware fraud detection, and agent permissions infrastructure, and our survey data, the core departments are Marketing and Member Experience, Technology and IT, Risk and Compliance, and Operations. Our data shows these departments already recognize each other as necessary collaborators, but leadership-level alignment on AI priorities, rated 3.19 out of 5 in our survey, lags behind that recognition.
Is our credit union behind if we haven’t started any of this? Most credit unions haven’t. Filene’s framing is meant to create urgency about a shift that’s already underway, not to suggest any individual institution has missed its window. Our survey data suggests that even institutions comfortable discussing AI in general terms, executives rated their personal comfort at 3.97 out of 5, are still working through the cross-departmental alignment that would need to happen before infrastructure like AI-legible product data or agent-aware fraud detection could be built effectively.
What is AI BASECAMP? AI BASECAMP is Wide Open Ventures’ executive alignment program for credit union leadership teams. It uses department-level surveys, structured conversations, and working sessions to help leadership teams reach a shared understanding of AI priorities, ownership, and governance, before moving into vendor selection, implementation, or formal roadmaps.
What’s a reasonable first step for a credit union thinking about this? Start with an honest, department-by-department read on where things actually stand, rather than assuming readiness based on how comfortable leadership feels discussing AI in general. The biggest barrier our survey identified wasn’t technology infrastructure. It was leadership alignment and understanding, cited by 61.9 percent of executives, which is exactly the kind of gap that structured cross-departmental conversations can close before any technical work begins. For more on how we’re thinking about this, check out our insights page.
