AI only works for credit unions when the digital infrastructure and cybersecurity are ready. Here’s how boards and leaders can build a member-centric foundation.
Technology strategy used to be a back-office project. Now it’s a board-level survival issue.
That’s the real message behind Chris Sachse’s comment on The CUInsight Network:
“Technology, in general, is really a board and CEO conversation.”
If your credit union wants to use AI for member-centric banking—fraud detection, smarter lending, financial wellness, better digital service—you don’t start with the AI tools. You start with digital infrastructure and cybersecurity.
Here’s the thing about AI for credit unions: it amplifies whatever foundation you already have. Strong, secure, flexible infrastructure turns AI into a strategic asset. Outdated, brittle systems turn it into risk.
This post builds on themes from Chris Sachse’s work at Think|Stack and connects them directly to AI adoption in credit unions. The focus: how to build the kind of digital infrastructure that actually supports member-centric AI, instead of getting in its way.
Digital Infrastructure Is Now a Board-Level Risk Issue
If your core systems, networks, and security posture can’t support AI, you’ll feel it in three places: member trust, operating costs, and regulatory pain.
Why the board has to own technology decisions
For mid-size and community credit unions, technology used to be something you “outsourced to IT.” That model broke the minute:
- Members shifted to mobile as their primary channel
- Fraud became real-time, automated, and AI-driven
- Regulators started asking tougher questions about third-party risk, cloud, and data use
AI for credit unions raises the stakes even more. You’re talking about models that:
- Touch member data at scale
- Influence credit decisions and pricing
- Run on cloud infrastructure you don’t fully control
Those are board-level issues, not product tweaks.
Boards that treat AI as a “cool feature” instead of a risk-and-strategy conversation end up with:
- Fragmented tools from different vendors that don’t integrate
- Security gaps between on-prem systems and cloud AI services
- Confusion over who owns failures when something goes wrong
Boards that get this right usually do three things:
- Set clear tech and AI risk appetite (what’s acceptable, what’s not)
- Tie AI and digital projects to business outcomes (ROA, member growth, fraud losses)
- Hold leadership accountable for infrastructure, not just features
From “projects” to infrastructure thinking
Most credit unions still manage technology as a chain of projects: new app here, chatbot there, analytics over there.
The problem? AI solutions depend on shared foundations:
- Clean, well-governed data
- Reliable identity and access management
- Standard APIs to connect core, CRM, LOS, and digital channels
If those foundations are weak, every AI initiative becomes a custom integration headache.
The reality? You’re not buying “AI projects.” You’re building a digital infrastructure that everything—AI included—will live on for the next decade.
Cloud, Cybersecurity, and AI: You Can’t Separate Them
You can’t have serious AI for credit unions without the cloud. And you can’t have cloud at scale without serious cybersecurity.
Why AI pushes you to cloud-native infrastructure
Most AI tooling—fraud models, machine learning platforms, large language models—runs on:
- Public cloud services (think major web service providers)
- Elastic compute and storage
- Containerized apps and APIs
Credit unions that cling to legacy on-prem infrastructure face three predictable problems when they try to adopt AI:
- Performance and scale: Real-time fraud and personalization models need fast, scalable compute.
- Integration friction: AI services expect API-first systems, not flat files and batch jobs.
- Cost inefficiency: Over-provisioned hardware sits idle while AI workloads spike in short bursts.
A secure, well-architected cloud environment turns AI projects from “special events” into just another workload.
Cybersecurity as the control layer for AI
AI doesn’t just increase value. It increases attack surface.
You’re now:
- Moving more sensitive data into cloud environments
- Integrating more vendors and third-party models
- Automating decisions that directly impact members and money
Strong AI adoption depends on a security architecture, not just products. At minimum, that architecture should include:
- Zero trust principles: Don’t assume anything inside your network is safe by default
- Multi-factor authentication for staff and admins across all AI tools
- Robust identity and access management with least-privilege access
- Network segmentation between AI workloads and core systems
- Continuous monitoring and logging of data flows, model access, and admin activity
Here’s the hard truth: AI will expose every lazy security shortcut you’ve taken in the past decade. Fixing those first is faster than cleaning up after a breach.
Agility: The Missing Ingredient in Member-Centric AI
The pandemic proved one thing to every credit union: agility isn’t optional.
Over a few weeks, institutions had to:
- Stand up remote work securely
- Handle massive digital traffic spikes
- Roll out new loan programs with constantly changing rules
The credit unions that adapted quickly already had flexible digital infrastructure. The ones that struggled were wrestling with brittle systems and manual workarounds.
AI amplifies this gap.
What agile infrastructure looks like in practice
A credit union that’s ready for AI usually has:
- API-first architecture connecting core, LOS, CRM, and digital channels
- Cloud-based data platforms instead of hundreds of siloed reports
- Modular services that can swap in new AI vendors without re-architecting everything
Here’s a simple example:
- You want to deploy an AI fraud detection model that needs real-time transaction data and device fingerprints.
- In a modern environment, that’s a standard stream of data feeding into a model via APIs.
- In a legacy environment, you’re trying to retrofit real-time intelligence onto batch files and mainframes.
The first path takes weeks. The second can drag into years—if it’s even possible.
Member experience depends on agility
Member-centric AI isn’t just about the algorithm. It’s about how quickly you can respond to what the algorithm tells you.
Examples:
- A risk model flags unusual behavior. Can you adjust limits or reach out to the member in minutes, or does it pass through three manual queues?
- A financial wellness model identifies members likely to benefit from coaching. Do you have workflows and digital channels ready, or does the insight die in a spreadsheet?
A dependable digital platform has three traits Chris Sachse highlights often: flexible, nimble, and secure. Skip any one of those, and your AI ambitions stall.
Budgeting for Cybersecurity and AI Infrastructure (Without Guesswork)
Most credit unions underfund security and overfund shiny tools. AI makes that imbalance dangerous.
Start with a realistic cybersecurity baseline
Before you add AI, you need a minimum security baseline that covers:
- Foundational controls
- MFA everywhere possible
- Endpoint protection on all devices
- Email security and phishing defenses
- Identity, access, and network controls
- Role-based access
- VPN or secure remote access
- Segmented network zones
- Data protection and resilience
- Regular, tested backups
- Encryption in transit and at rest
- Incident response plans and tabletop exercises
From there, you can layer AI-specific needs:
- Secure data pipelines from core/LOS/CRM to AI platforms
- Model access controls and activity logs
- Vendor risk reviews focused on data usage and retention
How much should a credit union budget for security?
There’s no universal number, but I’ve found one practical approach that works for many institutions:
- Treat cybersecurity and infrastructure as a percentage of operating expenses, not a leftover line item.
- Benchmark yourself: if peers your size are investing 2–3x more in security and cloud, you’re not “efficient”—you’re exposed.
Then, tie AI investments directly to measurable outcomes, such as:
- Reduced fraud losses (e.g., 20–40% lower card fraud)
- Faster loan decisioning (e.g., from days to minutes)
- Higher digital engagement (e.g., 30% more members using app features)
If the AI project can’t be connected to a metric, it’s not ready—and your infrastructure budget shouldn’t bend to support it.
Build a 24–36 month roadmap, not a shopping list
Instead of asking, “Which AI tools should we buy in 2026?” ask:
- What infrastructure gaps do we need to close this year?
- Which security controls are non-negotiable before we deploy AI more broadly?
- What member problems are we solving first: fraud, lending, or service?
Map AI and infrastructure together across 3 years:
- Year 1 – Get the basics right
- Cloud strategy, security baseline, identity management, core integrations
- Year 2 – Targeted AI use cases
- Fraud detection, collections optimization, simple member service automation
- Year 3 – Integrated member-centric AI
- Personalized offers, proactive financial wellness, omnichannel experiences
This way, infrastructure supports AI in a planned sequence, instead of reacting to one-off vendor pitches.
Turning AI and Infrastructure Into a Member-Centric Strategy
All the technology talk is pointless if it doesn’t improve members’ lives.
The most effective credit unions treat AI, cloud, and cybersecurity as tools for one goal: member-centric banking. That means:
- Safer accounts and faster fraud response
- Fairer, more consistent lending decisions
- Proactive guidance that keeps members financially healthy
- Digital experiences that feel personal, not generic
To get there, you need alignment across people, not just systems.
Make technology a shared language across the organization
Chris Sachse often emphasizes communication and empowerment. I agree: you can’t build member-centric AI in a silo.
Practical steps:
- Bring IT, risk, operations, and marketing into the same planning room for AI initiatives
- Translate technical decisions into member impact for the board
- Invest in training, so staff understand what AI is doing and can explain it to members
AI should never feel like a black box. Members should understand, at a basic level, why a decision was made and how their data is protected.
Where to start if you’re behind
If your credit union is still early in this journey, here’s a simple first step sequence:
- Assess your current digital infrastructure and security posture
- Use an external partner or internal audit to identify gaps
- Prioritize one or two AI use cases tied directly to member value
- Fraud detection or smarter loan decisioning are usually best starting points
- Align the board and leadership
- Set expectations for investment, risk, and timeframes
- Build a cloud and data roadmap that supports those use cases
- Don’t buy AI tools that your infrastructure can’t support securely
This series—AI for Credit Unions: Member-Centric Banking—is all about these practical steps. Tools matter, but the real differentiator is how you architect, secure, and govern the environment they run in.
The credit unions that win over the next five years won’t be the ones with the flashiest chatbot. They’ll be the ones whose boards treated digital infrastructure, security, and AI as one integrated, member-focused strategy.
Now’s the time to decide which side of that line you’re on.