Grid

Grid automates 90% of member support with Duckie, and shifted the job of support manager to an AI manager

90%ticket automation

Key metrics

90%

ticket automation

70K+

tickets per month

87%

lower cost per resolution

<5 min

response on 75% of tickets

Matthew Kim

Matthew Kim

Product Manager, Grid

Industry Fintech
ChatSelf-Training AIBackend Integrations
Prior to Duckie, managing the bizops team or the customer support team was about managing humans. Now, it's becoming more about someone who can manage an AI system.
Matthew Kim

Matthew Kim

Product Manager @ Grid

Matthew on the percentage of complex tickets Duckie handles

Grid is a fintech app helping millions of Americans take control of their money — cash advances up to $200, early access to tax refunds through PayBoost, credit-building, and the Grid Card. Grid's members live paycheck to paycheck and need money to move when they do, which means AI customer support is never a soft touchpoint. Every ticket has rent, a card decline, or a missed deposit behind it.

The challenge: hire or automate?

Grid's support load was hard on two fronts at once: volume was growing fast as the user base scaled, and the tickets themselves were complex — payment-allocation questions that require walking a member's full ledger, or cancellations that branch on product state, usage, and balance. The kind of edge cases a human has to hold in memory and execute perfectly every time.

Running a roughly seven-person team against a 24-hour SLA with thin overnight and weekend coverage, Grid hit a decision point that volume alone couldn't solve: keep hiring to keep up, or find a way to scale support without hiring.

Why Grid chose Duckie

Grid evaluated multiple vendors, including the largest names in the category. Two things set Duckie apart.

  • The support team builds and owns the AI support agent. Competitors put their own implementation engineers between Grid and its runbooks. Duckie lets Grid's PM build them directly — no vendor dependency — and new use cases ship as fast as the product changes.
  • Fast go-live. Duckie ingested Grid's ticket history, wrote its own runbooks, and went live in weeks where competitors quoted quarters. The launch target was 72% of tickets; a year in, Duckie handles 90%.
  • Complex ticket requests. Duckie doesn't just handle the easy stuff — it navigates Grid's most complex support automation, from payment-allocation questions that require parsing a member's entire transaction history to multi-step cancellations that branch based on product state, account balance, and usage patterns.

What changed for the team

The biggest shift was structural, and it's the part Grid's leadership cares about most. Of the original seven-person team, only two to three operators stayed on support. Roughly 60% of the team was rerouted to other functions — a capacity gain that let Grid scale support without hiring as volume grew.

"The remainder of the team has been freed up to work on other projects which has been great for the business operations team and the company overall."

— Matthew Kim, Product Manager, Grid

The one operator who stayed closest to support now runs the AI support agent itself — evaluating responses, tuning runbooks, improving the system rather than answering tickets. That's the role transformation: the support manager job went from managing humans to managing an AI.

The numbers behind it:

  • 90% of inquiries resolved by Duckie, 70,000+ tickets a month, with zero new agents hired to get there
  • Response time under 2 minutes on 75% of tickets
  • Cost per resolution dropped 87%. Because pricing is per-ticket against Duckie's throughput, the savings compound as Grid scales

"We saved 87% of our per ticket cost per resolution."

— Matthew Kim, Product Manager, Grid

Runbooks that moved the needle

Grid started where volume was highest and stakes were sharpest:

  • Cash advance status, eligibility, and repayment — instant answers at the moment they matter
  • PayBoost and paycheck tracking — Duckie reads Grid's backend live and tells members exactly where their money is
  • Ledger and payment-allocation questions — the hardest category, now resolved without an operator opening the ticket. This is the kind of automated ticket resolution that goes beyond answering questions; the AI support agent actually processes the inquiry end-to-end
  • Membership cancellations — Duckie navigates the branching product- and usage-state logic that used to live only in operator memory
  • Account, card, and transaction issues — resets, activations, disputes, settings changes, all closed inside Duckie

Looking ahead

"If Duckie disappeared right now, I think we'd be in a lot of trouble. We've put a lot of faith and trust in Duckie to stay performant and continuing to answer and support our customers. I think customers would be left in a black hole for quite a while."

— Matthew Kim, Product Manager, Grid

Grid is planning to grow roughly 3x this year, and the support automation model is built to scale with it — by adding runbooks to Duckie, not headcount to the org chart. As Grid ships new products, Duckie expands to cover them, and the support queue keeps feeding the roadmap with live signal on what members need next.

The framing other vendors sell is "our AI does the work, you manage the agents." Grid built the opposite on Duckie: Grid builds the system, Duckie is the tool that makes it possible — AI customer support that compounds as the company grows.