Grid automates 91% of member support with Duckie, handling 70,000+ tickets a month without growing the team

Grid's members don't write in with idle questions. They write in when a cash advance is pending and rent is due, when a PayBoost deposit hasn't landed, when a card isn't working at the checkout. Every ticket has weight, and every hour on the clock matters.

As the app's user base grew, so did the volume of those tickets, and Grid's team was left with a question they couldn't solve by hiring: "How do we keep hitting our 24-hour SLA, and get members their money faster, without doubling headcount and doubling costs?"

So Grid ran a tight evaluation of AI support platforms. The bar was high on four fronts:

• Response accuracy members could actually trust
• Real resolutions, not just polite deflection
• A short path to production, measured in weeks not quarters
• A security posture that held up to fintech-grade scrutiny

The point wasn't to shave off ticket volume. It was to give members a better experience and hand the team back the bandwidth to focus on the cases that actually required a human.

Most of the vendors Grid looked at wanted a long runway. Months of implementation, a professional services engagement, sometimes an internal AI team to maintain runbooks, and even after all that, deflection tended to flatten out around 30 to 40%.

Duckie worked differently. The AI ingested Grid's existing ticket history, wrote its own runbooks, and let Grid's agents test and correct it before anything went live. What took competitors a quarter took Duckie a couple of weeks.

How fast it got live

• Automated workflows shipped within weeks of kickoff
• First rollout focused on FAQ-type tickets to pressure-test Duckie on real member traffic
• Zero implementation services, zero runbook writing, zero AI hires on Grid's side

How deep the resolutions went

• Duckie plugged into Grid's backend to read and act on live product state
• 91% of daily inquiries resolved by the AI
• 75% of those tickets closed in under five minutes

Throughout the rollout, Duckie's team stayed embedded with Grid's product and ops leads, treating the deployment as a living system rather than a finished handoff. And that's the part that kept compounding: Duckie's AI flags its own blind spots, drafts fixes, and ships them once Grid approves. The system sharpens itself every week without Grid's engineers lifting a finger.

The net effect is a support org that gets better the more it runs, and a team that doesn't have to grow for it to happen.

Running on a platform that trains itself changed what Grid's operators actually do day to day. They stopped being runbook authors and started being reviewers of an AI that proposes its own improvements. Every ticket becomes training data, every correction tightens the next response, and every new edge case gets promoted into a suggested workflow instead of sitting in a backlog.

That shift shows up in the numbers:

• New use cases go live in days, not sprints
• Responses stay on-brand and consistent across thousands of conversations
• One platform handles the whole loop, no separate tools for routing, knowledge, or QA
• Support capacity scales with the product, not with the org chart

Grid and Duckie started where the volume was highest and the member stakes were sharpest, the ticket types any fintech ops leader can recite from memory.

• Cash advance status and eligibility. Members asking about their limit, their approval status, or when repayment is coming out now get an answer in seconds, at exactly the moment that answer matters most.
• PayBoost and paycheck questions. Duckie reads live from Grid's backend, tells the member exactly where their PayBoost or paycheck is in the flow, and resolves the ticket without an agent ever opening it.
• Account, card, and transaction issues. Password resets, card activations, dispute inquiries, and settings changes now close out inside Duckie, leaving agents free for the harder, messier cases that actually need a human in the loop.

Every workflow was tuned to close the loop with the member, in Grid's voice, without the cold-bot feel.

The headline: 91% automation, 70,000+ tickets a month, and zero new agents hired to get there.

The support queue has quietly turned into one of Grid's best product signals. What members ask about, what frustrates them, and what they try to do next all flow back into the roadmap. Support stopped being a line item that scales with the user base and started being a function that scales itself, and a lens into what Grid should build next.