AI lending helps credit unions approve more good loans, faster, without losing the human touch. Here’s how to use it for inclusive, member‑centric lending.
Why AI Lending Matters For Credit Unions Right Now
Credit unions left billions of dollars in safe, profitable loans on the table last year because traditional credit models said "no" to people who would've paid them back.
That's the core problem this episode of The CUInsight Network with Jeff Keltner from Upstart is really about. Not technology for its own sake, but members who are more creditworthy than their files suggest, and credit unions that are struggling to serve them through outdated lending processes.
As part of our AI for Credit Unions: Member-Centric Banking series, this post looks at what AI lending actually changes, how it supports inclusive lending, and how to use it without turning your credit union into a cold, robotic experience. Because as Jeff says:
“Don’t underestimate the value of the human touch.”
Here’s the thing about AI in lending: it’s not about replacing loan officers. It’s about finally aligning your mission—serving members—with your decisioning, your digital experience, and your risk management.
What AI Lending Really Does (And What It Doesn’t)
AI underwriting helps a credit union see risk more clearly by analyzing far more data than a traditional credit score or rules-based model.
Jeff describes Upstart’s model as using 800+ variables, including both traditional and non‑traditional data points. That’s a huge jump from the usual handful of inputs in a scorecard model.
From “FICO plus a few rules” to rich member profiles
Traditional lending for many credit unions still looks like this:
- Pull credit score
- Check DTI and income
- Apply a rule matrix
- Approve, counteroffer, or decline
That process is easy to manage, but it’s blunt. AI lending models, by contrast, can incorporate variables like:
- Occupation and field of work
- Employment stability and income consistency
- Education and career trajectory
- Historical repayment behavior patterns
The outcome isn’t just a fancier score. It’s a more accurate prediction of default risk, especially in the so‑called subprime pool, where many “thin file” or young borrowers live.
What AI doesn’t do
There are a few myths worth killing right away:
- AI does not guarantee higher approval at higher risk. Good AI models aim for more approvals at the same or lower loss rate.
- AI does not remove compliance or fair lending responsibility. You still own model governance, testing, and oversight.
- AI does not replace every human decision. The best programs route edge cases to humans and use people where judgment and empathy matter.
The reality? AI is just a better tool for the job of predicting credit risk. What you build around that tool—policies, workflows, member experience—is where leadership matters.
Inclusive Lending: Finding the “Hidden Prime” Members
The most powerful reason for AI lending in a credit union isn’t cost savings. It’s mission alignment.
Jeff points out that more people are creditworthy than we think. Traditional scores under-call a big chunk of borrowers, especially younger members, gig workers, and those who’ve had one rough patch in an otherwise solid financial life.
How AI helps you say “yes” more often
AI models trained on large, diverse datasets can:
- Approve more members at prime or near‑prime pricing who would otherwise be declined or pushed into high‑rate products.
- Identify borrowers in the "subprime" FICO band who behave like prime based on broader variables.
- Reduce friction for strong applicants by removing unnecessary documentation steps.
A simple way to think about it:
Traditional models see a risky pool. AI helps you see who’s actually risky inside that pool—and who isn’t.
For member-centric credit unions, that means:
- More approvals without blowing up charge‑offs
- Better pricing for good members who look marginal on paper
- A fairer shot for those outside the conventional profile (younger, new to credit, immigrant, gig‑economy workers)
Why this matters for fair and inclusive lending
AI loans can support inclusive lending when they’re designed and governed well. You can:
- Test approval and pricing outcomes across demographics
- Remove variables that introduce bias
- Show regulators you’re using a more accurate, not less fair, way to assess risk
The payoff isn’t abstract. Inclusive, AI‑driven lending means:
- A first auto loan for a member who was declined elsewhere
- A debt consolidation loan that actually lowers someone’s cost of credit
- A path to build credit history with responsible terms
That’s real member impact—and it’s exactly what differentiates credit unions from large banks and fintechs.
Digital Lending That Works End‑to‑End (Not Just a Pretty Front Door)
Most credit unions don’t have a technology problem at the front of the application. They have a workflow problem behind it.
Jeff makes a sharp point: it’s not enough to put a digital form on top of a legacy lending process. If the back-end is still manual, slow, and paper-heavy, members feel it.
What “end‑to‑end” actually means
A genuinely modern, AI‑enabled lending journey looks like this:
- Smart application
- Pre‑fills known member data
- Asks only what’s needed for the product and risk profile
- Real‑time decisioning
- AI model evaluates 800+ variables
- Approvals, declines, and refer‑to‑manual outcomes happen in seconds
- Streamlined verification
- Income, identity, and employment verified digitally when possible
- Only a minority of applications need human follow‑up
- Frictionless closing and funding
- E‑sign, digital disclosures, fast funding
- Clear guidance for members at each step
Contrast that with the “digital veneer” version:
- Web form submits to a queue
- Staff re‑key data into LOS
- Follow‑up by phone or email
- Manual decision after hours or days
Same website, totally different member experience.
Practical steps to modernize your lending journey
You don’t have to replace everything at once. I’ve seen credit unions get real traction by:
- Starting with one product (often unsecured personal loans or auto refi) as the AI‑pilot.
- Mapping the current state process end‑to‑end, with time stamps and member pain points.
- Defining clear metrics up front: approval rate, average APR, time‑to‑decision, funded rate, NPS.
- Partnering with an AI lending platform while keeping internal control over rules, tiers, and member communication.
The goal isn’t just “AI in underwriting.” The goal is a faster, cleaner, less stressful borrowing experience for members, powered by better decisioning.
Keeping the Human Touch Where It Matters Most
Jeff’s quote—“Don’t underestimate the value of the human touch”—isn’t nostalgia. It’s strategy.
The winning credit unions won’t be the ones that automate everything. They’ll be the ones that automate the right things and humanize the rest.
Where AI should lead
AI is perfect for:
- Straight‑through approvals on clear, low‑risk applications
- Real‑time pricing and offers, so members don’t wait days
- 24/7 digital access, especially after work hours and on weekends
If a member with a strong profile wants a simple personal loan at 10:30 PM, they shouldn’t have to wait for the branch to open. Let AI underwriting and digital fulfillment handle that.
Where humans should stay in the loop
Humans shine when:
- A member is on the edge of approval and needs context and coaching
- There’s been a life event—job loss, medical issue, divorce—and they’re asking, “What are my options?”
- A decline needs explanation and an alternative path (e.g., secured loan, smaller amount, co‑signer)
AI can say yes or no more accurately. But only humans can say, “Here’s why this happened, and here’s how we help you get to yes in the future.”
A practical approach I like:
- Use AI to handle 80–90% of simple decisions.
- Route the remaining 10–20%—plus any member who asks for it—to experienced lending staff.
- Train those staff as financial guides, not just decision-makers.
That’s how you scale lending while still feeling like a credit union, not a faceless app.
How Credit Union Leaders Can Get Started With AI Lending
If you’re a CEO, CIO, or lending executive, the hardest part isn’t believing the tech works. It’s deciding how to start without overwhelming your team or your risk committee.
Here’s a straightforward blueprint.
1. Clarify the problem you’re solving
Don’t start with “We want AI.” Start with:
- “We want higher approval rates at the same loss rate.”
- “We want decisions in under 60 seconds for at least 70% of apps.”
- “We want to expand inclusive lending to younger and thin‑file members.”
That clarity will guide partner selection, model design, and internal change management.
2. Pick one product and one segment
Common starting points:
- Unsecured personal loans for debt consolidation
- Auto refinance
- Small personal lines of credit
Limit scope. Get wins. Then expand.
3. Build a joint risk framework
With your internal risk team and any AI partner:
- Define target loss rates and return thresholds
- Set routing rules: which apps are instant, which require manual review
- Agree on monitoring: monthly performance, fairness, and exceptions
You’re not giving up control; you’re upgrading the tools you use to execute your risk appetite.
4. Design the member journey, not just the model
Make sure your digital lending experience reflects credit union values:
- Plain‑language explanations of decisions
- Clear paths to talk to a real person
- Proactive education: “Here’s what would improve your odds next time.”
AI for credit unions should never feel like a black box. It should feel like faster, fairer, more transparent lending.
5. Communicate the “why” to staff and members
Staff need to hear:
- “AI is here to remove busywork, not your job.”
- “You’ll spend more time solving real member problems, less time keying data.”
Members need to hear:
- “We’re using more information to understand you as a whole person.”
- “We can approve more good loans, faster, without raising risk.”
When you connect AI lending back to your member‑first mission, adoption gets a lot easier.
Where AI Lending Fits in Member‑Centric Banking
AI in lending is one pillar of member‑centric banking for credit unions, alongside fraud detection, smarter member service automation, financial wellness tools, and better analytics.
Lending is where members feel your values in the most direct way: you either support their goals, or you don’t. That’s why getting this right matters so much.
Use AI underwriting to:
- Say yes more often to the right people
- Offer fairer pricing to members who deserve it
- Deliver digital experiences that feel modern but still human
Then keep the human touch exactly where it makes your credit union special—coaching, empathy, and real conversations about money.
The credit unions that win the next decade won’t be the ones with the most technology. They’ll be the ones that use AI to see their members more clearly and serve them more personally.
If your lending still feels like a bottleneck—slow, rigid, and frustrating for staff and members—that’s your signal. There’s a better way to approach this, and now’s the right time to start.