How to Add an AI Agent to Gorgias, Zendesk or Kustomer Without Replacing Your Helpdesk

The short answer. Keep Gorgias, Zendesk or Kustomer as the system of record and put an AI agent such as REP AI in front of it. The agent answers shoppers in your store chat from your catalog, policies and order data, then escalates to your helpdesk with the conversation attached, so your team continues the thread instead of starting again.
The work is six steps: which jobs the agent takes, what it may see and do, how conversations reach the helpdesk, who owns that queue, how its rules fit your helpdesk's own automations, and how you will tell whether it worked.
Disclosure: REP AI publishes this guide and sells an AI agent that works alongside the helpdesks named here. Helpdesk pricing facts below come from each company's own pages and were captured on the dates shown. The customer results below come from REP AI's own published case studies, not from the stores or from an independent source. If anything is out of date, tell us and we will correct it.
This guide describes a method. It does not promise a result, because results depend on your store, your helpdesk setup and your questions. The two customer examples are one store's account each, as REP AI reported it in its case study, not a forecast for yours.
Why add a layer instead of replacing the helpdesk
Your helpdesk holds your customer history, your macros, your reporting and the habits of your team. Replacing it means migrating data, retraining people and changing how you report, all at once, while customers keep writing in. Adding an agent in front of it changes one thing: who answers first.
That makes a layer easy to test and easy to remove. Your helpdesk keeps working the way it does today, and you judge the agent on its own results. Replacing a helpdesk is sometimes the right call, for example when the helpdesk itself is the problem. That is a different decision and this guide does not cover it.
How the three helpdesks price their plans
Pricing units below are from each company's pricing page. The Gorgias row was checked on gorgias.com on 17 November 2026; the Zendesk and Kustomer rows were captured 2026-10-06. Check the source before you rely on a detail.
| Helpdesk | Pricing basis | Source |
|---|---|---|
| Gorgias | Helpdesk tiers by ticket volume, from Starter at $10 a month for 50 tickets to Advanced at $750 a month billed annually for 5,000 tickets, never priced per agent. AI Agent is on every plan and charged per automated interaction: $0.90 on annual plans, $1.00 on monthly plans, and $1.50 past the plan's allowance. Helpdesk and AI Agent bundles start at $40 a month (Starter, monthly billing) and $77 a month (Basic, billed annually). Free trial: 7 days. (As listed on Gorgias's site on 17 November 2026.) | gorgias.com/pricing |
| Zendesk | Per-agent plans; AI resolutions are priced separately | zendesk.com/pricing |
| Kustomer | Not published on the page captured | kustomer.com/pricing |
Pricing units differ, so two bills can grow for different reasons: tickets, agents, automated interactions or resolutions on the helpdesk side, and traffic or conversations on the agent side. Work out your own volumes before you compare.
If your helpdesk already includes an AI agent, read what it covers before you add another. The helpdesk's own AI may already handle your email tickets well. A layer earns its place where it covers something the helpdesk's AI does not: for example the conversation in your store chat before a shopper has a ticket at all.
How the pieces fit
Shopper on your store
|
v
AI agent in store chat ---> answers from your catalog,
| policies and order data
|
v
Escalation to your helpdesk, with the conversation attached
(Gorgias, Zendesk, Kustomer, Freshdesk or Richpanel)The seam between the two is the escalation, which is why much of this guide is about getting it right.
Step 1: Decide who does which job
Write the split down before you connect anything. A simple starting point:
| Job | Agent in store chat | Helpdesk and your team |
|---|---|---|
| Pre-purchase questions (fit, shipping, compatibility) | Answers from your catalog and policies | Receives escalations |
| Order status and tracking | Answers from order and carrier data | Takes exceptions: lost package, address error |
| Returns and exchanges | Explains policy, collects the details | Approves anything outside policy |
| Cancel or change requests | Handles within the limits you set | Handles the rest |
| Complaints, damaged goods, refunds outside policy | Collects details, then escalates | Owns the conversation |
| Email tickets already in the helpdesk | Not in scope unless you choose to add it | Stays as it is |
Email, social and WhatsApp AI from REP AI are a later decision, not part of the first rollout. The table above is a starting point. Change it to match your own tickets. If you have not yet sorted your tickets by topic, the guide AI Customer Service for Shopify Stores: The Complete Guide to What to Automate and What to Hand Off shows how.
Step 2: Give the agent your policies and its limits
An agent only knows what it reads. Give it your shipping, returns, cancellation and product information, and remove conflicts between pages first. Then set limits: what it may do alone, what it must check first, and what it must pass on. Decide your escalation triggers before launch, and write them down where your team can read them.
Step 3: Connect the helpdesk and test one handoff end to end
REP AI escalates to your team with the conversation attached, and works with Gorgias, Zendesk, Kustomer, Freshdesk and Richpanel. How a ticket looks in your helpdesk depends on your account settings, so check each of these in your own account before you go live:
- It arrives where your team looks. The ticket lands in the view or queue someone actually watches.
- The conversation is readable. The full exchange is attached and makes sense without the chat window.
- You know who the customer is. Name, email and any order number are visible.
- The subject line helps. Your team can tell what the ticket is about before opening it.
- Replies reach the customer. Your team's reply goes out through the channel the customer expects.
- One conversation, one ticket. A single conversation does not create duplicate tickets.
- The customer does not repeat themselves. Your team can answer from what is attached.
Run these tests with a handful of internal conversations that cover different reasons for escalation. Fix what you find in your helpdesk's own settings, then test again.
Step 4: Decide who owns the queue
An escalation that nobody owns is a dead end. Name the person or role who picks up escalated tickets, the hours they cover, and what the agent tells the shopper outside those hours. Promise a reply window only if your team can meet it. Review the escalated tickets weekly and sort them by reason. Each repeated reason is either a policy answer the agent is missing or a topic that should stay with people.
Step 5: Stop two automations from colliding
Your helpdesk probably has its own rules: auto-replies, auto-close timers, satisfaction surveys, assignment rules. Check how each behaves on an escalated ticket. A satisfaction survey that fires before a person has replied, or an auto-close timer that shuts a ticket nobody has read, will undo the good work upstream. If your helpdesk's AI agent and the added agent could answer the same message, decide which owns which channel, so a shopper does not get two different answers.
Step 6: Roll out in stages and measure
- Test internally with real past questions, including awkward ones.
- Go live for part of your traffic and keep the rest as it was.
- Compare against a baseline you recorded before you started: share of tickets by topic, time to close, and repeat contacts for the same order.
- Define "resolved" in advance, for example closed with no human touch and not reopened within seven days. Use the same definition before and after.
- Read transcripts. Numbers show whether it worked. Transcripts show why.
Treat a short test as an indication, not proof, because ticket volume changes week to week.
Notes by helpdesk
If you use Gorgias. Gorgias lists REP AI as a partner app in its app directory (checked 17 November 2026). REP AI escalates to Gorgias with the conversation attached. Gorgias also sells its own AI Agent, so decide up front which one answers which channel. Gorgias counts a conversation as a billable ticket once a human agent, a Rule or AI Agent replies, so check how escalated chats count against your plan. For the setup, see the REP AI and Gorgias integration; for plan costs, Gorgias pricing explained; and for a head-to-head view, REP AI vs Gorgias. Gorgias's AI Agent page, read on 17 November 2026, cites a 60% average automation rate across 5 brands, so ask which conversations that figure counts before you split channels. If you are weighing a different helpdesk instead, see Gorgias alternatives for Shopify.
If you use Zendesk. If your support operation spans several channels and teams, keeping Zendesk as the system of record is a sensible default. REP AI escalates to Zendesk with the conversation attached, so the layer sits in front of the platform you already run.
If you use Kustomer. REP AI escalates to Kustomer with the conversation attached. Check in your own account how the escalated conversation appears on the customer record, since that depends on your configuration.
Freshdesk and Richpanel users follow the same six steps.
Two stores that list a helpdesk and use REP AI
REP AI's case studies on two stores list a helpdesk in their platform details. Each is one store's account, as REP AI reported it. Neither tells you how a layer will behave in your helpdesk, so run the checks in Step 3 in your own account.
Underoutfit: product questions before they become tickets
Underoutfit sells underwear and intimate garments. REP AI's case study on Underoutfit says shoppers could not judge how garments would fit from product images alone. The store uses REP AI as an AI concierge, which the page describes as answering product questions and making recommendations around the clock. On support, its co-founder says: "By helping out with basic questions, Rep was able to reduce customer support tickets." (Felix Leshno, Co-Founder of Underoutfit.) The case study gives no ticket figure, so treat it as a description, not a measurement.
Satya Jewelry: a goal of scaling support without more headcount
Satya Jewelry is a New York-based lifestyle brand known for handcrafted jewelry. REP AI's case study on Satya Jewelry lists a helpdesk and Klaviyo as its platform and names a goal of scaling customer support without increasing headcount. The store implemented REP AI as an on-site concierge.
REP AI's case study on Satya Jewelry reports that 94% of customer support inquiries were handled automatically. The page does not state the period the figure covers, so read it as one store's figure as REP AI reported it, not a rate to plan around.
The case study also describes a review habit that matches Step 4. The team used its "missing information data" to update its FAQ and site content, for example on how to redeem loyalty points. That is the loop Step 4 describes: a repeated gap is either a policy answer the agent is missing or a topic that should stay with people.
What a layer adds to your costs
Your helpdesk keeps its own pricing. The agent adds a second line. REP AI's core plans are priced by monthly website sessions, and its add-ons (helpdesk, email, social and WhatsApp AI) carry separate pricing. You do not need REP AI's helpdesk to follow this guide. Add the two lines together at your own volumes before you decide, and remember that the units differ.
When replacing the helpdesk makes more sense
Adding a layer does not fix a helpdesk that is too slow, too costly or poorly suited to your store. If that is your real problem, a layer will help with some conversations and leave the rest. REP AI also has a helpdesk.
You can try conversations in the REP AI simulator at /simulator, or start the 14-day free trial, which needs no credit card. Our integrations page is at /integrations.
Related guides: read more integrations and architecture articles.
FAQ
How do I add an AI agent to my existing Gorgias or Zendesk setup without replacing it? Put the agent in front of the helpdesk. Split the jobs, give the agent your policies and limits, connect it so escalations arrive in the helpdesk with the conversation attached, name an owner for that queue, test one handoff end to end, then roll out to part of your traffic and compare.
Which AI sales agent works with Gorgias? Gorgias lists REP AI as a partner app in its app directory (checked 17 November 2026), and REP AI escalates to Gorgias with the conversation attached. Other products also connect to Gorgias, so compare them on the same real questions, and read how each prices.
Is there an AI agent for Kustomer ecommerce? REP AI escalates to Kustomer with the conversation attached. Test the handoff in your own account, since how it appears on the customer record depends on your configuration.
Can I use REP AI with Zendesk? Yes. REP AI escalates to Zendesk with the conversation attached, and Zendesk stays your system of record.
Will two AI agents conflict? They can if both answer the same message. Decide which agent owns which channel and topic, and check how your helpdesk's rules treat an escalated ticket.
Do I need to migrate tickets or customers? Not to add an agent in front of your helpdesk. Your tickets and customer records stay where they are.
More from REP AI

Which Ecommerce Helpdesks Are Not Priced per Ticket? (2026)
A dated list of how ecommerce helpdesks and AI agents charge: per ticket, per agent, per conversation, per resolution or per session. It shows which bills rise as your AI resolves more and where REP AI's session pricing fits.
Want to see what your store would do with Rep?
Run your real Shopify catalog through the simulator before you commit to anything. No sales call required.

