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
- Footwear AI customer service results on Freshdesk at a glance
- What does the brand do?
- Which helpdesk does the brand use?
- How did they train their AI customer service agent?
- When did they decide to turn on 'direct replies' to customers?
- What was the biggest thing they did to improve their AI agent's resolution?
- How do they customize their AI agent setup to work for their business?
- What impact is their AI customer service agent having now?
- Where does the brand go from here?
Created time
May 29, 2026 01:26 PM
Title length (<60)
Author
Last optimised
Sep 15, 2026
Ecomm?
Tags
Image
Publish date
May 29, 2026
Video
Slug
footwear-ai-customer-support-case-study
Featured
Featured
Type
Case study
Ready to Publish
Ready to Publish
A footwear brand resolves 95% of ~5,300 monthly Freshdesk tickets with AI, holding 90% AI CSAT and saving ~422 hours every month.
Sell footwear to customers around the world and you inherit a support inbox that never really sleeps. Shoppers want to know where their order is, whether a shoe runs true to size, and how to start a return (the same handful of questions, thousands of times a month).
That whole load now runs through My AskAI, across both Freshdesk and the chat widget on their site. The AI resolves 95% of roughly 5,300 tickets a month at 90% AI CSAT, and saves the team about 422 hours every month.
Here's how it came together.
Footwear AI customer service results on Freshdesk at a glance
Fact | A footwear brand |
Industry | eCommerce, footwear and apparel (global DTC) |
Helpdesk | Freshdesk (email Copilot) plus the My AskAI chat widget on their site |
Support volume | ~5,300 tickets/month |
AI resolution rate | 95% across Freshdesk email and the website widget |
AI CSAT | 90% |
Time saved | ~422 hours/month |
Key features used | Self-Learning, Shopify integration, User Data API, Guidance across all three types, Custom Answers |
Knowledge sources | Freshdesk help center via the Knowledge Base connector, Self-Learning auto-drafted articles, Shopify product and order data, live User Data API lookups |
Go-live mode | Split by channel: direct replies in the website widget, Copilot drafts on Freshdesk email |
What does the brand do?
The company is a footwear and apparel brand with a signature shoe line at the center of the range, sold direct to consumers.
They ship internationally, and that global direct-to-consumer reach means high volume, repetitive questions, and customers in every timezone (which is a lot of "where's my order?" at 3am).
Which helpdesk does the brand use?
They run support across two channels. Email tickets land in Freshdesk, and live questions come in through the My AskAI chat widget on the website.
We work in both, but in two different modes. In Freshdesk, the AI runs in Internal Notes (Copilot) mode: it drafts the reply as an internal note, and a human agent gives it a quick read and sends it. In the widget, the AI replies to shoppers directly.
That split lets the team keep an agent's eye on email while the widget answers website questions on its own. Both channels share one trained AI, so the knowledge and the brand voice stay consistent wherever a customer turns up.
How did they train their AI customer service agent?
The training stack got built up in layers, and each layer closed off more of the question set.
First, they connected their Freshdesk help center through our Freshdesk Knowledge Base connector. That handed the AI every existing help article to answer from on day one.
Then they switched on Self-Learning. It drafts brand-new help articles from the questions the AI sees and the answers agents give, so the help center grows itself (no one sitting there writing docs all weekend). In the last 30 days, those auto-drafted articles answered more than 300 tickets on their own.
Next came the commerce data. They connected Shopify for products, orders and customer data, then added our User Data API for live order and customer lookups. It means the AI can answer "where is my order?" with the real tracking status, the way a human agent would (most bots just fall back to a polite policy line).
Last, they added a handful of Custom Answers for the questions that need a fixed, scripted reply every single time.
When did they decide to turn on 'direct replies' to customers?
This wasn't an all-or-nothing switch. They turned it on channel by channel.
On the website widget, the AI replies to shoppers directly (no human in the loop, by design). Chat questions tend to be lower-risk, and they're the ones where an instant answer counts for the most, so the widget went live with direct replies.
In Freshdesk, email stays in Copilot mode for now. The AI drafts the reply as a note and an agent sends it, which keeps a human in the loop on the channel where customers expect a more considered response. It's the boring-but-effective way to move fast where it's safe and stay careful where it counts.
What was the biggest thing they did to improve their AI agent's resolution?
In ecommerce, a small set of question types (order status, sizing, returns) makes up most of the inbox, so deflection rates usually land somewhere between 60% and 80%. This brand clears that comfortably at 95%, and two things did most of the heavy lifting.
The first is Self-Learning. Instead of letting the help center go stale, it keeps drafting new articles from the questions customers actually ask. Over the last 30 days, those articles answered more than 300 tickets on their own (questions that would otherwise have gone to a human).
The second is live order data. With Shopify and our User Data API connected, the AI doesn't punt "where is my order?" to an agent: it looks the order up and answers with the real status. Order-status and returns questions are the bulk of any DTC inbox, and answering them with live data instead of a canned line is what turns a good deflection rate into a great one.
Put together, the two form a loop. Self-Learning widens what the AI knows, the live data lets it act on what each customer is actually asking, and the resolution rate climbs as both get better.
How do they customize their AI agent setup to work for their business?
The setup is tuned heavily with Guidance, and they use all three types we offer.
Communication guidance keeps the AI on brand: the right tone, the right sign-off, the voice a footwear brand's customers expect. Context and Clarification guidance tells it when to ask for more before answering, so it requests an order number up front. Handover and Escalation guidance sets the rules for when a question should go to a human instead (knowing when to tap out is half the job).
On top of Guidance, they lean on a handful of Custom Answers for the replies that need to be scripted word-for-word. And they're starting to build Tasks (our agentic, multi-step workflows) so the AI can start taking actions for customers, on top of answering their questions.
What impact is their AI customer service agent having now?
Here are the numbers from the last 30 days:
- 95% AI resolution rate across Freshdesk email and the website widget
- ~5,300 tickets handled per month
- ~422 hours saved per month
- 90% AI CSAT
- 300+ tickets answered by Self-Learning's auto-drafted articles alone
The CSAT number matters every bit as much as the resolution rate. Resolving 95% of tickets only counts if customers are happy with the answers they get, and at 90% AI CSAT, they clearly are.
Where does the brand go from here?
The next step is action. The team is building out Tasks so the AI can handle multi-step workflows: the order edits, address changes and refunds that today still need a person to finish.
With Shopify and our User Data API already connected, the data those Tasks need is sitting right there, so it's mostly a matter of wiring up the workflows. Over time, more of the Freshdesk email channel can shift from Copilot drafts to direct replies as the team's confidence in the AI's email answers grows.
None of this is chasing a support team of zero. The goal is to let the AI handle the high-volume, repetitive 95% so the team can spend its time on the cases that need a person: the VIP, the odd edge case, the upset customer who really does need a human.
If you'd like to see more stories like this one, you can browse all our customer case studies, where YouGarden runs a similar Freshdesk setup. And when you want to size it up for your own store, our pricing sits on one page.
