Mike HeapMike is an experienced Product Manager who focuses on all the “non-development” areas of My AskAI, from finance and customer success to product design, copywriting, testing and more.
Table of Contents
- Group-payments AI customer service results on Zendesk at a glance
- What does the platform do?
- Which helpdesk does the platform use?
- How did they train their AI customer service agent?
- Why the team runs its AI in copilot mode
- The biggest lever in their AI setup
- How does the team customize their AI agent setup?
- Setting the tone with Communication Guidance
- Knowing when to ask with Context & Clarification Guidance
- What impact is the AI copilot having now?
- Where does the team go from here?
Created time
Jun 3, 2026 02:00 PM
Title length (<60)
Author
Last optimised
Sep 15, 2026
Ecomm?
Image
Publish date
Jun 3, 2026
Video
Slug
group-payments-ai-customer-support-case-study
Featured
Featured
Type
Case study
Ready to Publish
Ready to Publish
A group-payments platform handles around 1,600 Zendesk support tickets a month with My AskAI running as an AI copilot. Self-Learning drafts the answers, and a growing library of those drafts now feeds about 1,200 replies a month, so the knowledge base keeps building itself while a human stays on every reply.
This company is a payments platform built for groups, and the support behind it is high-stakes by nature. When a ticket is about a payment that didn't land, a refund, or access to an account holding a group's money, the answer has to be right. A small team fields a steady stream of those questions through Zendesk, somewhere around 1,600 a month.
So rather than put a public-facing bot in front of customers, they use us as a copilot. The AI drafts the reply, a human checks it, and the human sends it. This is the kind of rollout I wish more payments teams copied, because it gets the speed without handing the customer over to a robot.
Here's how it came together.
Group-payments AI customer service results on Zendesk at a glance
Fact | A group-payments platform |
Industry | Fintech, payment collection for groups and organizations |
Helpdesk | Zendesk (tickets and live chat, mostly email) |
Support volume | ~1,600 tickets/month |
Key features used | Self-Learning, Custom Answers, Communication Guidance, Context & Clarification Guidance |
Knowledge sources | Zendesk help center (155 articles) via the Knowledge Base connector, website sync (~71 pages), a large Custom Answers library |
Go-live mode | Internal-note Copilot mode from the start, no direct replies |
What does the platform do?
It is an online platform that lets groups and organizations collect payments, forms and sign-ups in one place. Schools, PTAs, sports teams, clubs, churches, nonprofits and HOAs use it to gather dues, event fees, fundraiser money and registrations without chasing cash and checks.
Which helpdesk does the platform use?
They run support on Zendesk, handling tickets and live chat in a workflow that's mostly email. We plug straight into that Zendesk setup.
The rollout actually started on the technical side. Their CTO did the first setup inside Zendesk (after an early walkthrough with our team), then handed day-to-day ownership to the support team to run and tune. There was no drawn-out bake-off against another AI vendor first, which meant the team could tune the copilot around how they answer customers from day one.
How did they train their AI customer service agent?
The agent learns from the same material the support team already keeps up to date. Three sources do most of the work.
First, the Zendesk help center, connected through the Knowledge Base connector. That pulled in 155 help-center articles. During setup, the team split the public articles from the internal ones (so the AI only ever drafts from content that's fine to show a customer).
Second, their public website, pulled in through website sync, which added roughly 71 pages of product and policy content on top of the help center.
Third, a large library of Custom Answers, hand-written for the questions where the team wants the wording exactly right rather than paraphrased from an article.
Curating that knowledge base matters as much as adding to it. Early on, a duplicate source had crept in and was dragging answer quality down, so removing it was one of the first cleanup jobs. The lesson is a familiar one: the AI is only ever as good as the material you point it at.
Why the team runs its AI in copilot mode
They deliberately keep My AskAI in internal-notes, copilot mode. The AI drafts a reply inside the Zendesk ticket, and a human reviews it before anything reaches the customer (drafts only, never sent on their own).
For a payments platform, that's a deliberate choice and a smart one. Tickets about money movement, refunds and account access carry consequences a generic chatbot answer can't, so they keep a person on every reply.
The copilot still does the heavy lifting (yes, even the tricky payment ones): it reads the ticket, finds the right knowledge and writes the draft. The agent's job shifts from writing every reply from scratch to checking and sending.
The biggest lever in their AI setup
The single biggest thing they turned on is Self-Learning. It watches how tickets get resolved and auto-drafts help articles from those resolutions, so the knowledge base grows out of the support the team is already doing.
That compounding shows up in the numbers. Self-Learning's drafts have been used across roughly 1,200 of their monthly responses. Paired with the hand-curated Custom Answers library, it splits the work neatly: Self-Learning proposes new coverage from real tickets, humans approve what's right, and the agent gets sharper on the next batch without anyone sitting down to write articles by hand (it's one of our favorite features for exactly this reason).
How does the team customize their AI agent setup?
Out of the box, any AI agent sounds generic. They shaped theirs with two layers of Guidance.
Setting the tone with Communication Guidance
Communication Guidance controls how answers are phrased: the brand's tone, the structure, and what the AI should and shouldn't say. A concrete example came up early on.
The AI was suggesting phone calls to customers, which doesn't fit their largely older, less technical user base (no screen sharing, and a written answer simply works better for them). Communication Guidance is where you correct that kind of behavior, so the drafts sound like the brand and point customers at help that actually suits them.
Knowing when to ask with Context & Clarification Guidance
Context & Clarification Guidance tells the AI when to ask a follow-up before it drafts. Payment and account questions often hinge on one missing detail, and a confident answer to the wrong version of the question is worse than no answer at all. This guidance keeps the copilot asking for the detail it needs.
What impact is the AI copilot having now?
Today the copilot is woven into the support team's daily work:
- Around 1,600 Zendesk tickets a month are handled with My AskAI drafting in copilot mode.
- Roughly 1,200 of those monthly replies drew on a Self-Learning draft.
The AI runs as a copilot, so the goal is agent speed and consistency, with a person still on every customer reply.
The payoff is a solid, well-sourced draft waiting in every ticket, so agents answer faster and stay consistent. For a payments team, we'd argue that's the right trade to make.
Where does the team go from here?
They are still early in what the copilot can do, and a few things are on the roadmap.
The nearest-term piece is auto-tagging, so the AI only drafts on the tickets it should and stays off the automated, do-not-reply emails. After that, the plan is to let the AI start replying directly to customers on the safest, most repetitive ticket types, once the team is happy with the drafts it's seeing.
Further out, they're eyeing the User Data API, which would let the AI answer account and transaction-specific questions like payment status. Zendesk messaging is on the list to add when they're ready (no extra cost on our side), and they plan to roll out the AI Support Copilot Chrome extension so agents get the same drafting help wherever they're working.
It's a measured path, which fits a company whose customers are trusting it with their money. It is also not the only team taking it, since a creator-monetization platform runs the same internal-notes copilot setup on Zendesk, with direct replies switched off by choice. If you'd like to see more customer stories like this one, browse how Edel Optics and Honeygain run their Zendesk support on My AskAI, or all our case studies. And if you're weighing it up for your own team, here's how the pricing works.
