American Heritage Credit Union answers 50,000 member conversations a month without losing the human touch

At a Glance:

• Mission: Deliver instant, always-on service to 325,000 members, and give Contact Center representatives room for the conversations that shape a member's financial life.
• Solution: "Gigi," a voice and chat AI agent built with SoundHound AI, running more than 30 workflows and roughly 100 FAQs.
• Results:
  
  • 598,000+ member interactions handled in 11 months
  • 100% completion on dispute inquiries, loan applications, and contact requests
  • 88% of fraud and dispute chats resolved without a representative
• What’s Next: A hybrid agentic model and proactive account alerts that reach members first.

The mission: answer every member instantly, at any hour

It's the first of the month and a member's Social Security deposit hasn't arrived. The money exists. It's sitting in a settlement queue, waiting on a settlement date set by the payroll company or the sending institution. Until someone tells them that date, they don't know whether to pay the electric bill today or wait until Thursday.

Across town, another member drops a debit card down a sewer grate climbing out of the car. There's no time to get to a branch before the weekend.

American Heritage Credit Union fields both of those calls, and thousands more like them, every single day.

Both matter enormously to the person making them. Account balance questions, transaction lookups, and card replacements are the highest-volume, lowest-complexity calls in banking, and they arrive in numbers no contact center can staff against.

The real cost of those calls isn't the time they take. It's what a representative could be doing instead. A member weighing a mortgage against a home equity line, opening an estate account after a death in the family, or working out why a loan payoff quote moved needs someone who can ask a second question and listen to the answer. Those conversations decide whether a family consolidates debt at a good rate or a bad one. At a not-for-profit cooperative, they are the entire point — and they're the first thing squeezed when the queue fills with routine requests.

American Heritage Credit Union saw that trade-off early and refused to accept it.

Founded in 1948 as the Budd Workers Federal Credit Union, American Heritage today holds more than $5 billion in assets and serves over 325,000 members and 800 Workplace Partners across 35+ locations in greater Philadelphia. It ranks among the top 70 largest credit unions in the United States and has been named Pennsylvania's 2025 Best-in-State Credit Union by Forbes for the third consecutive year.

It was also among the first credit unions in the country to put a conversational AI agent in front of its members — early enough that there was no peer playbook to copy. So American Heritage wrote one.

The solution: an AI agent the credit union built as its own

Meet Gigi. She answers American Heritage's phones and its chat window, and for most members she's the first point of contact with the credit union.

The name comes from a retired VP of Human Resources, known for being unfailingly patient with whoever walked through her door. It's what her grandchildren call her.

Gigi works the same way — patiently, and without needing a second attempt. Gigi authenticates members, reads balances and recent transactions, moves money between accounts, resets online teller passwords, activates and replaces debit cards, and quotes loan payoff amounts to members budgeting without the paperwork in front of them. 24/7. She finds the branch that happens to be on the way to your sister's place in South Philadelphia, with the hours and available services attached. She walks a member through opening an estate account, asking whether they already have a death certificate. And every November she invites members to Grand Illumination, the credit union's holiday tree lighting with the Philadelphia Boys Choir & Chorale.

Two people know Gigi better than anyone. Rebecca Londo designs every conversation and workflow. David Lulko, a software developer, builds the backend that connects her to the credit union's Symitar core. Between them they've grown Gigi into more than 30 workflows and roughly 100 FAQs across three environments — and they've been at it since before she went live.

They haven't done it alone. "The delivery team we have today, 10 out of 10 would recommend," says Londo, Digital Analyst at American Heritage Credit Union.

That partnership shows up in how little Gigi asks of the team now. "It used to be every day, all the time. But now we've gotten her to a steady spot where there isn't really much maintenance required," says Lulko, Software Developer at American Heritage Credit Union.

The results: 598,000 conversations, answered on the first ring

Gigi handled more than 598,000 member interactions in eleven months.

[[block:stats]]
598,000+ | member interactions handled, July 2025 – May 2026
1,897 | voice conversations answered instantly, every day
88% | of fraud and dispute chats resolved without a representative
100% | completion on dispute inquiries, loan applications, and contact requests
[[/block]]

The depth is as striking as the volume. Gigi resolves 78% of account maintenance chats and 70% of rate and CD inquiries on her own. She absorbs the 5% of inbound voice traffic that arrives as ghost and dead-air calls — noise no representative ever hears. And she keeps getting better: self-service has climbed every single quarter since American Heritage started measuring, without a single rebuild.

What that frees up is harder to put on a chart, so Lulko puts it plainly: "I'd imagine our Contact Center would be completely swamped and overwhelmed with calls — a lot of dropped calls."

Volume is only half of it. The work that reaches a human changed shape.

"The cases that they handle are a lot more complicated and actually useful. They used to get spam a lot. Now it seems like it's decreased a lot."
‍
David Lulko
Software Developer, American Heritage Credit Union

Where AI hands off to people

The handoff is where American Heritage pulls ahead of most AI deployments.

When Gigi reaches the limits of what she can resolve, the entire conversation moves to a representative with a single click. Every question the member asked. Every answer Gigi gave. Authentication status flagged at the top of the transcript, in plain language, so the rep sees instantly that this member is already verified.

The member doesn't recite their account number a second time. The representative doesn't re-run a verification that's already done. They pick up mid-thought — "I see you've been trying to reset your password, let me take it from here" — and they often close the loop by showing the member how to get further with Gigi next time.

Most organizations treat the AI agent and the Contact Center as two separate service channels. At American Heritage, they're one continuous conversation, and the member never feels the seam.

"That's a big part of the credit union philosophy — people helping people, member outreach, member experience. Really making them feel part of it. You're talking to an actual person when you're talking to the bot."
‍
Rebecca Londo
Digital Analyst, American Heritage Credit Union

What's next: agentic AI, one workflow at a time

American Heritage is quickly expanding agentic workflows for Gigi, allowing her to reason with more flexibility, communicate more naturally, and tackle even more complex customer needs on her own. 

Gigi is also learning to proactively raise things the member didn't call about. The moment someone authenticates, she checks for anything sitting on the account that needs attention. If the credit union has flagged a member's mail as undeliverable, Gigi says so, confirms the address on the spot, and either clears the warning or walks them through updating it. Debit cards sitting unactivated are next, along with delinquent loans and negative balances.

Every candidate use case clears the same bar it always has. "Is it useful to the members? That's the number one question. Are we going to spend all this time building something that they don't really need or they're not going to use?" says Londo.‍

Seventy-eight years after a group of factory workers pooled their money to help each other out, the philosophy holds. The tools are just getting better.

"Gigi, from my perspective, is a lot more powerful than I initially thought. So we should be able to use her in a variety of different ways."
‍
David Lulko
Software Developer, American Heritage Credit Union

Interested in starting your own AI journey with SoundHound?

We'd love the opportunity to discuss how we can help you achieve your goals with agentic AI.

Talk to an expert