GiveCard started as a nonprofit in Boston that gave out cash assistance. Getting that money to people was hard. Banks wanted high minimums, and the alternatives were checks and gift cards, each paid out by hand. So GiveCard built its own tools, then saw that other nonprofits and government programs needed them too.

Today GiveCard is infrastructure for mass payouts. Its customers run guaranteed income pilots, financial assistance programs and emergency disaster relief. Their recipients get a prepaid debit card or a bank transfer, a portal to track every transaction, and fraud protection on the card.

"We have a cheesy term that we say here, money movement made magical, and that's both on the customer end and on the recipient end," says Sofia Pedro, Head of Product at GiveCard. Sofia has been at GiveCard since it was a nonprofit. "It should be completely intuitive to receive these funds and spend it."

GiveCard runs support for every recipient itself, and Sofia treats it as a different kind of support. "Since we're in the business of giving people money, it is very sensitive topic," she says. "A lot of the time, this may be the first financial tool this population has ever used, and we often have no idea what kind of stressful situation someone might be in."

She describes the moment GiveCard designs for. "Having your card decline at checkout with no explanation is a stressful experience, and getting those moments right has always been a priority for us."

Language matters as much as speed. GiveCard works with refugee resettlement agencies, and many recipients do not speak English as their first language. "Multilingual is a huge, huge factor," Sofia says.

GiveCard's phone line ran on a traditional IVR for basic requests, accompanied by live support. Sofia was benchmarking AI vendors, with voice and multilingual support as her two requirements.

Then the federal government shut down on 1 October 2025, and SNAP recipients stopped receiving their benefits. Local governments and nonprofits across the country needed to send emergency payouts fast, and their recipients would call with questions. The shutdown lasted 43 days.

GiveCard had to choose. "We needed to pick a vendor that was going to be extremely fast and that we felt was going to be supporting us throughout the process," Sofia says.

GiveCard signed with Lorikeet on 31 October 2025 and went live on 3 November. "We actually built workflows specific for this event within two to three days, which was crazy," Sofia says.

Two things made that speed possible. GiveCard knew what recipients would ask, from its experience with emergency relief programs. And Sofia could build the workflows herself. "There was no hierarchy, bureaucracy across teams," she says. "I was really able to take on the work and just build it out myself with the help of the Lorikeet team and just a couple of people from my side."

Across the emergency program, Lorikeet answered about 60,000 calls from GiveCard cardholders. At launch the agent worked out what each caller needed and passed them to the right support team.

The hardest part is identity, and it is what the two teams are building now. Each GiveCard customer shares different recipient data under its own agreement, and some programs keep recipients anonymous.

"We have a really unique customer base that have really tight data sharing agreements across their recipients, so we don't actually have consistent data on our recipient. It's very much based on the program that they're in," Sofia says. "That means we need to customize the verification flow based on the payout method that they received."

In the flow they are building, the agent asks what the recipient needs and works out the intent, and each intent triggers its own workflow. The workflow already knows whether the question needs verification. If it does not, the recipient gets an answer straight away. If it does, the agent identifies the recipient and the program that paid them, then runs that program's verification steps through GiveCard's API. If a program allows verification only by email and phone number, the workflow verifies with those two. Once verified, the agent can check balance, transaction history, set a pin or understand why a transaction got declined.

"If it doesn't, it could be an easy five second phone call where the response is given back, and it's troubleshooted right away," Sofia says. "If not, it goes through a series of verification steps in order to provide that answer."

At any step the recipient can ask for a person. The call moves to a human agent with its context. "What we're striving for here is being able to carry a recipient throughout the entire support process without them having to continuously repeat themselves all the time," she says.

With the agent answering common questions, GiveCard has been able to upskill its human agents. "One thing that we're now able to spend a lot more time on that we weren't able to do previously is really, really upskilling our human agents," Sofia says. The result is agents trained to handle more specialized, complex flows with less context switching.

GiveCard went live on voice first because the shutdown left no time to sequence channels. With a normal timeline, Sofia would build in a different order. "I would definitely recommend going the approach of email and chat first over voice, having the foundations in place and then tweaking the voice nuances later on," she says. Her reason is that expectations are rising. "They are also experiencing CX with other products that are also receiving answers immediately, so the expectations are increasing a lot faster."

GiveCard is now looking at how to use Lorikeet on its general support lines, and plans to add more channels later. Sofia is also noticing new opportunities. "There's so many ways to really make customer support a core strength that we just didn't even realize, like, oh, we should now start doing outbound calls to notify recipients of when their card is gonna get loaded," she says.

Sofia expects the next emergency. "There's always going to be a next event that comes, there's gonna be a big new customer that comes and wants to do mass payout distribution," Sofia says, "and I feel like I now have the confidence to be able to do customized workflows very, very quickly because we've had that successful launch in the past."