AI Lending That Truly Serves the Underserved

AI Lending That Truly Serves the Underserved

Most credit unions say they’re member-centric. Very few can honestly say what Lorrie Wohlfeil said:

“We are truly helping the underserved.”

That line from her CUInsight Network conversation isn’t marketing fluff. It’s the standard her team at Lending Solutions Consulting, Inc. (LSCI) uses when they step into the shoes of front-line lenders and live the realities of credit union lending.

Here’s the thing about AI for credit unions: if it doesn’t help you serve the underserved better, it’s just another expensive toy. The point isn’t fancier models; it’s fairer approvals, smarter risk, and members who walk away with both funds and financial confidence.

This post connects Lorrie’s member-first lending philosophy with modern AI lending tools. If you’re trying to modernize loan decisioning, improve financial wellness, and keep your mission intact, there’s a practical path that blends empathy, education, and data science.

---

What “Member-Centric Lending” Really Means in an AI Era

True member-centric lending combines three things: empathy, expertise, and execution. AI should reinforce all three, not replace them.

Lorrie’s work at LSCI is grounded in a simple belief: credit decisions are moments of trust. Members come in with a need, often stress, and sometimes shame. The best credit unions treat that moment as a chance to educate, not just approve or deny.

AI lending tools can either help or hurt that moment:

• If AI is a black box that spits out yes/no, it erodes trust.
• If AI is a decision support tool that helps staff explain why and what’s next, it builds trust.

Member-centric AI lending is built on a few non-negotiables:

• Explainability: Staff must understand why a model recommends approve/decline/conditional so they can explain it in plain language.
• Fairness: Models need to be monitored for disparate impact, especially on underserved groups your credit union exists to support.
• Coaching, not just credit: Every decision should include guidance: “Here’s how this affects your score, here’s how to improve it, here’s a product that fits where you are today.”

The reality? AI doesn’t replace member-centric lending. It exposes whether your lending culture was member-centric to begin with.

---

From Classroom to Call Center: Turning Lending Theory into Practice

Lorrie describes a model that more credit unions should copy:

Blending classroom education with hands-on experience, where consultants step into the shoes of credit union staff.

That matters because lending breaks down in the handoff between policy and practice. You can invest in fancy AI scoring, centralized risk models, and model governance docs, but if the front-line team doesn’t know how to use them in real conversations, members feel like they’re talking to a script, not a person.

How this applies to AI lending tools

If you’re rolling out AI-enabled lending (automated decisioning, risk-based pricing, pre-approvals), you need three layers of training:

1. Concept training

   • What data the AI uses (e.g., credit bureau, internal behavior data, deposit patterns)
   • How the model segments risk tiers
   • What “confidence score” or model thresholds mean in practice

2. Scenario training

   • Walk through real member stories: thin-file, gig worker, past charge-off, recovering from medical collections
   • Show how AI might score them and what options are available
   • Practice explaining approvals, counteroffers, and denials with empathy

3. Shadowing and embedded coaching

   • Just like LSCI consultants, have someone sit with your lenders and MSRs
   • Listen to calls, watch how staff use the lending system, and adjust scripts and workflows

AI loan decisioning only becomes member-centric when staff feel confident enough to override, question, or augment the AI in specific situations. That confidence comes from hands-on support, not an LMS video.

---

Using AI to Serve the Underserved Without Raising Risk

Most credit unions want to help underserved members but get stuck on one thing: risk.

AI actually helps here—if you design and govern it well.

1. Go beyond the traditional credit score

Underserved members are often:

• Young adults with thin credit files
• Immigrants or cash-economy workers
• Members recovering from life events (medical, divorce, job loss)

A single bureau score punishes them. An AI lending engine can incorporate more signals:

• Deposit stability and income consistency
• Savings patterns and direct deposit history
• Payment histories on internal products (e.g., credit builder loans, share-secured cards)
• Engagement signals (login frequency, use of digital banking, contact center history)

When I’ve seen this implemented well, credit unions were able to safely approve 10–20% more loans in targeted segments without a spike in losses, because they finally recognized good risk that traditional scores ignored.

2. Turn denials into financial wellness moments

This is where Lorrie’s emphasis on education really shows up. AI can pre-generate:

• Reasons for denial in member-friendly language
• A “path to approval” checklist (e.g., pay down $800 on Card X, set up direct deposit, maintain on-time payments for 90 days)
• A recommended financial wellness plan: small-dollar savings goals, credit-builder products, or debt consolidation scenarios

Your staff’s job then shifts from “I’m sorry, you don’t qualify” to “Here’s exactly what we can work on together so this is a yes next time.”

That aligns perfectly with the credit union mission and actually reduces future delinquency, because members who understand their situation make better decisions.

3. Monitor AI models for fairness

Serving underserved members means being ruthless about fairness. That includes:

• Regular model audits for disparate impact across age, geography, and protected classes (using compliant proxy methods)
• Reviewing override patterns: Are staff consistently overriding the model in certain communities or branches?
• Tuning strategy rules so that members with similar risk patterns receive similar treatment, regardless of zip code or background

If you combine this technical governance with Lorrie’s culture of empathy and accountability, you get something powerful: AI that scales fair decisions instead of scaling old biases.

---

Building a Member-Centric AI Lending Playbook

If you’re a credit union leader trying to modernize lending, here’s a practical playbook that reflects what LSCI teaches plus what AI makes possible.

Step 1: Define what “helping the underserved” means for you

Vague goals lead to vague outcomes. Be specific:

• Target segments (e.g., thin-file under 30, members with FICO 580–640, gig workers with variable income)
• Desired outcomes (e.g., 15% increase in approvals for thin-file members, 20% reduction in high-cost external borrowing)
• Guardrails (loss targets, max DTI, product limits for high-risk tiers)

Write this down as a member-centric lending mission and make sure every AI project points back to it.

Step 2: Map current lending journeys

Sit with your lending and contact center teams the way Lorrie’s consultants do.

• Listen to loan calls and branch conversations
• Watch how staff use your LOS and CRM screens
• Note where members get confused, anxious, or frustrated

You’ll quickly see where AI can help:

• Pre-filling applications with existing data
• Pre-qualifying offers in digital banking
• Suggesting alternative products when an application is weak
• Surfacing coaching tips to staff in real time

Step 3: Introduce AI decision support, not auto-pilot

Start by augmenting human decisions:

• Use AI models to recommend approvals/declines with confidence levels
• Give staff a clear view: “Model recommends approve at X rate, here’s why”
• Set policies for when overrides are allowed and what documentation is required

Gradually automate only the low-risk, well-understood segments (e.g., prime auto refis, payroll-backed personal loans), and keep complex or underserved segments in augmented mode longer.

Step 4: Train like LSCI: classroom + in-the-chair

Build a training plan inspired by Lorrie’s approach:

• Classroom: Explain how the AI works, what it’s allowed to do, and how it supports the credit union’s mission.
• Role-play: Practice tough conversations—denials, counteroffers, higher-than-expected rates—using AI outputs.
• In-the-chair coaching: Lending leaders or external consultants sit with staff during live use and give immediate feedback.

You’ll know the training is working when staff stop saying, “The system says no,” and start saying, “Here’s what your profile looks like and how we can improve it together.”

Step 5: Track both financial and human outcomes

If you only track approvals and yields, you’ll miss whether your AI is actually member-centric. Add:

• Member satisfaction after lending interactions
• Conversion rates from denial → future approval
• Engagement with financial wellness tools after a loan decision
• Performance of loans made to your targeted underserved segments

This is where leadership matters. If you reward only speed and volume, empathy disappears. If you reward long-term relationships and outcomes, AI becomes a tool for member advocacy, not just profit.

---

Keeping Humanity at the Center of AI Lending

What struck me most from Lorrie Wohlfeil’s story is how personal this work is for her—built on her father Rex Johnson’s legacy and reinforced by years in the credit union trenches. That mindset is exactly what keeps AI lending honest.

AI for credit unions should never be about replacing people. It’s about:

• Giving your lenders better visibility into risk
• Giving your members clearer paths to financial progress
• Giving your underserved communities fairer access to credit

If your AI roadmap doesn’t support those goals, it’s time to rethink it.

For this "AI for Credit Unions: Member-Centric Banking" series, one theme keeps repeating: technology only works when it amplifies your mission. Lorrie’s team shows how powerful it is when training, empathy, and real-world practice come first.

The next step is yours: choose one underserved segment, define what “helping” them really means, and design your AI lending strategy around that promise. The tools exist. The question is whether your culture is ready to use them in the way your members actually deserve.