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
- What does Honeygain do?
- Which helpdesk does Honeygain use?
- How did Honeygain train their AI customer service agent?
- When did Honeygain decide to turn on 'direct replies' to customers?
- What was the biggest thing Honeygain did to improve their AI agent's resolution?
- How do Honeygain customize their AI agent setup to work for their business?
- Setting the tone for high-emotion tickets
- Asking the right clarifying question first
- Knowing when to hand a player to a human
- What impact is Honeygain's AI customer service agent having now?
- Where do Honeygain go from here?
Created time
Jun 3, 2026 09:38 AM
Title length (<60)
Author
Ecomm?
Tags
Image
Publish date
Jun 3, 2026
Slug
my-askai-honeygain-case-study
Featured
Featured
Type
Case study
Ready to Publish
Ready to Publish
Honeygain resolves around 3,060 of its ~3,400 monthly Zendesk tickets with AI: a 90% AI resolution rate, while holding 78% AI CSAT and saving roughly 507 hours every month.
Let's be honest: a passive-income app lives and dies on one question from its users, and that question is "where's my money?". Honeygain gets it around the clock, in volume, and almost always with a charge of suspicion behind it. "Why is my withdrawal under review?", "why was my account flagged?", "where are my credits?": that's a hard inbox to staff, and an expensive one to run at the revenue-per-user this kind of app works with.
Today the AI handles most of it. Honeygain resolves 90% of roughly 3,400 monthly tickets inside Zendesk with My AskAI, gets about 507 hours back every month, and does it while holding a 78% AI CSAT in a category where plenty of tickets are unhappy before anyone replies. Here's how it came together.
What does Honeygain do?
Honeygain is a free app that pays people for the internet bandwidth they aren't using. You install it, it shares your spare connection with Honeygain's network, and you earn money you can take out through PayPal or in crypto once you pass a minimum threshold.
It's a big, consumer-scale base. Honeygain reports more than 12 million users worldwide, over a million payouts completed, and seven years in business (fun fact: the average payout works out at around $27), with apps across Android, Windows, macOS and Linux so one person can earn on several devices at once. You can read more on their site.
Which helpdesk does Honeygain use?
Honeygain runs support on Zendesk, in the Support Inbox and Tickets. That's where their agents already live, so the plan was simple: put an AI agent into the inbox they already had.
Rather than lean on Zendesk's own AI add-on, Honeygain dropped My AskAI straight into Zendesk Tickets. We work the same queue the human agents do, pick up tickets as they land, and draft or send replies in place. For the team, nothing about where they work changed.
How did Honeygain train their AI customer service agent?
The first source was the one they'd already invested in: their Zendesk help center. Honeygain connected it through the Zendesk Knowledge Base connector, so every article they'd written (payouts, devices, eligibility, account questions: the greatest hits of a rewards inbox) became answerable on day one.
On top of that they built out a large library of Custom Answers (the scripted, exact responses for the questions that come up constantly and have one right answer). Minimum payout thresholds, supported countries, how a withdrawal is processed, what a particular account status means: these are the ones where a paraphrase isn't good enough, and a Custom Answer pins the reply down word for word.
Between the two, the agent started from a base of real Honeygain knowledge rather than a generic model guessing at policy. That's the difference, in our experience, between an AI that just deflects and one that answers a payout question correctly.
When did Honeygain decide to turn on 'direct replies' to customers?
Honeygain deployed the AI in Zendesk Tickets in "reply only to the first message" mode. The AI takes the first reply on every new ticket; if the conversation carries on, a human agent picks it up from there.
It's a deliberate way to scope a high-stakes inbox. The first message on a reward or withdrawal ticket is usually the most repetitive part (a status question, a how-to, a policy check), and that's exactly what the AI is strongest at.
Anything that turns into a back-and-forth, which is where the trickier and more emotional cases show up, lands with a person. A resolution here means the ticket never needed to be escalated to a human, and we keep that escalation easy on purpose.
What was the biggest thing Honeygain did to improve their AI agent's resolution?
The biggest single lever here was Self-Learning. Honeygain turned it on and let the knowledge base grow itself.
Self-Learning watches the questions the AI couldn't confidently answer, and the replies human agents send after a handover, then drafts new help content from them. Over the last 30 days, those auto-drafted articles answered roughly 600 tickets on their own (no one had to sit down and write them).
That matters more in this category than in most. Offers change, payout rules get updated, new account-policy edge cases appear, and a help center maintained by hand drifts out of date fast. Self-Learning closes those gaps as they surface, which is what holds a 90% resolution rate steady at 3,400 tickets a month instead of letting it sag as the questions move.
How do Honeygain customize their AI agent setup to work for their business?
Beyond knowledge, most of the "make it feel like ours" work happened through Guidance, the rules for how the AI talks, what it asks, and when it steps back. Honeygain uses it across three areas.
Setting the tone for high-emotion tickets
A lot of Honeygain's inbox arrives frustrated. Communication Guidance sets how the AI replies to those tickets (the tone, the structure, the brand voice) so a "where's my payout" answer reads as calm and human rather than curt or robotic. Getting the tone right is half the battle when the customer already suspects they've been wronged.
Asking the right clarifying question first
Payout questions are unanswerable without detail: which device, which payout method, how much, what the account status is. Context & Clarification Guidance tells the AI to gather that first instead of guessing, so the reply that follows is actually about the customer's situation rather than a generic policy paragraph.
Knowing when to hand a player to a human
Not every ticket should be automated, and Honeygain is explicit about which ones shouldn't. Handover & Escalation Guidance routes the cases that need a person (suspected fraud, account-ban appeals, payment disputes) to the human team instead of letting the AI improvise on something sensitive.
That's the other half of why the resolution number stays honest. The AI isn't trying to win tickets it has no business answering.
What impact is Honeygain's AI customer service agent having now?
The numbers, over the last 30 days:
- 90% AI resolution rate across ~3,400 monthly Zendesk tickets, with around 3,060 resolved by the AI without ever being escalated to a human.
- ~3,400 tickets handled per month.
- ~507 hours saved per month (roughly nine minutes of agent time per ticket).
- 78% AI CSAT. That's a strong score for this category specifically. When a chunk of your inbox is ban appeals and withheld payouts, some tickets arrive unhappy no matter how good the answer is. The AI clears the high-volume, answerable questions, and the human team keeps the cases that need judgment and a softer touch.
- ~600 tickets a month answered by Self-Learning's auto-drafted knowledge, content the system wrote itself.
In our experience that result is the whole point: automate the repetitive volume that makes up most of a passive-income app's inbox, and give the team back the hours for the cases that need a person.
Where do Honeygain go from here?
The clearest next lever is live account data. Right now the AI answers payout questions from policy, but with the User Data API it could answer them from the customer's actual transaction state: the real status of this user's withdrawal. For a "where's my money" inbox, that's the difference between a good answer and the exact answer.
From there, Tasks and Tools open up the agentic side: checking a payout's status, prompting a stalled KYC step, or surfacing account flags for the trust-and-safety team to act on. And with a 12-million-user base spread across the world, multilingual coverage is the natural way to keep one team serving every market.
None of this is about removing the humans. I've spent enough time around support inboxes to know that "100% automation" was never the goal for a high-emotion category: the goal is for the AI to take everything it can answer well, so our people are free for the ones only a human can handle.
If you'd like to see more customer stories like Honeygain's, browse all our case studies. If you're running support on Zendesk yourself, that's where My AskAI slots in, and the pricing is usage-based, so your bill stays flat as the AI gets better.
