Best AI agent builder for ecommerce (Shopify and everything next to it)

There is no single best builder. Inbox is the free native start. Gorgias is the ecommerce desk. Tidio and Chatbase are quota apps. eesel sits on the desk you already pay for. A general builder is the buy when refusal, brand, and handoff have to be designed — not picked from a plan page.

Monday, 9

. The same three tickets as last Monday.

Where is order 1842. Can I return the navy knit. I want a refund, I changed my mind, it shipped yesterday.

A chatbot vs AI agent split shows up here, not in a demo. The bot recites the policy. The agent looks up the order, starts the return when the policy allows it, and refuses the refund when it does not — then writes the note so nobody is reconstructing the thread at 11.

That is the job. The App Store listing is not.

What should an ecommerce agent actually do?

Baymard’s accounts research found that 50% of respondents named order tracking the most important account feature — and 67% of test sites still failed to give shoppers the full tracking picture on the site itself (Baymard, updated 2026). Shopify’s own WISMO note is blunter: if shoppers cannot see where the parcel is, they open a ticket (Shopify, 2025).

The builder is only as good as the seven jobs it can finish without a person.

We score every tool on the same seven columns. The real estate companion uses the same card with different rows.

A chatbot can talk. An agent has to do those seven.

How we scored the builders

Official pages, August 2026: Gorgias pricing and AI Agent, Shopify Inbox plus Inbox help, Tidio, eesel, Chatbase, YourGPT pricing and Shopify. If a page says contact sales, we say that.

Strong / Partial / Weak is a capability read, not a league table.

The scorecard

*Refund refusal is a designed rule, not a toggle. The platform can refuse. The install will not, until someone writes the rule and tests it against real tickets.

Is Shopify Inbox enough?

Start here if the store is new and the team still answers chat themselves.

Inbox is free on Basic, Grow, Advanced, and Plus (App Store). The agent can answer product questions, recommend from the catalog, look up an order, and add items to cart (Shopify Help). Assign it to all chats, only when staff are offline, or never.

The honest limit is the surface, not the model. Shopify says the agent is early access and only for certain merchants. It needs new customer accounts. Order lookups prompt a sign-in. The “you are chatting with AI” notice cannot be removed. It only runs in the storefront widget — not email, WhatsApp, or the desk you already use.

Shopify’s own docs already require a sign-in for order lookup. A WISMO answer that stops at a login is a ticket that arrives by email ten minutes later. Buy Inbox for free coverage while a person still owns the desk. Do not buy it as the last tool.

Is Gorgias the default Shopify desk?

For a lot of mid-market Shopify brands, yes — and Gorgias says so. They claim they power conversations for 40% of Shopify brands (pricing). The AI Agent is built for WISMO, returns, subscription edits, and discounts against live inventory (support skills). Courtney Bajek at Orthofeet, on their site: they aimed at 30% automation in 30 days and passed 50%.

Published bundles: Starter $40/month (50 tickets, 30 automated); Basic $90/month monthly / $77 annual (300 tickets, still 30 automated); Pro $550/month / $471 annual (2,000 tickets, 190 automated). Automated overage is $1.50. Tickets overage at $0.40 or $0.36. Starter has no brand voice, guardrails, or Actions — those start on Basic.

The cost-to-run problem is the double meter. An AI resolution still sits in the helpdesk. A sale week can hit ticket overage and AI overage on the same thread. The $1.50 automated overage is on Gorgias’s pricing page. Right buy if you need that inbox. Wrong buy if you wanted a quiet widget and a predictable bill.

What about Tidio and Lyro?

Tidio is the app merchants install when they want a chat bubble this afternoon. Lyro is the AI agent on top.

Tidio pricing (annual, August 2026): Starter $24.17/month (100 human conversations); Growth from $49.17; Plus from $300 plus usage; Premium is contact sales. Lyro standalone starts at $32.50/month for 50 AI conversations (annual). Every account gets 50 Lyro conversations once, free. The agent page also quotes about $0.50 per conversation. Native Shopify actions start on Growth. Hiding Tidio branding is $16.67/month on Growth, included on Plus.

Lyro can recommend from the live store and run order-update actions. Handoff is a first-class control. Tidio says it answers only from the content you give it.

The bill is three sliders — human conversations, Lyro, and Flows — and they do not move together. A fair mid-market start. Not a designed brand surface, and not a Gorgias-style desk until Plus.

When is eesel the smarter install?

eesel does not want to be your storefront. It wants to sit in Zendesk, Freshdesk, Intercom, Shopify, Slack — the tools already open — and take the tickets a person should not touch.

Published price: $0.40 per regular task — one ticket or one chat, however many messages (eesel pricing). Light lookups are free. No platform fee, no seats, no minimum. $50 free to start. Enterprise is $1,000/month plus usage.

A teammate can brief it in Slack: “Anything over $500 in refunds, loop me in first.” That is refund refusal as an instruction, not a hope. WISMO and returns work if Shopify and the desk are connected. They also say you still pay when the answer is wrong.

eesel is weak on brand surface because it is not trying to be one. Right for the morning pile in Zendesk. Wrong if the storefront has to sound like the brand.

Why include Chatbase?

Because merchants keep putting it on the same shortlist, and it publishes a price.

Chatbase, yearly: Free (50 credits; idle agents deleted after 14 days); Hobby $32/month (700 credits); Standard $120 (4,000, helpdesk, voice, Shopify); Pro $400 (15,000, 5 seats). Extra credits $40 per 1,000. Removing the badge is $1,188/year. Enterprise is contact us.

A capable general agent with a credit meter and a branding tax. Not trained on ecommerce the way Gorgias claims to be. WISMO and returns depend on the actions you wire. Fine if the catalog is simple. Thin if you needed a desk.

The Shopify conversation around these apps is not theoretical. Merchants are already arguing the shortlist in public:

Read that thread for the tone, not the winner. People want WISMO and a human they can reach. They do not want another bubble that apologises.

When should you buy an app, and when should you hire a studio?

Buy the app when all of this is true:

1. One Shopify store, one language, one desk.
2. The top intents are WISMO, sizing, and a standard return window.
3. The vendor widget is acceptable on the product page.
4. Someone on the team will own the knowledge every week.

Hire a studio when any of this is true:

• The agent has to refuse a refund the way a trained lead would, not the way a model guesses.
• The widget has to look like the site, not like an app.
• The ticket still has to land in the desk you already run, tagged, with a person only on the exceptions.
• Voice, WhatsApp, or after-hours coverage is part of the same agent, not a second vendor.

Aurelia’s support desk was the second case. Earlier chat tools had been switched off — they could talk, but they could not do the work. We wrote the brief from live tickets, put the help centre in as memory, and used n8n to open the case and tag the product. Elena Marchetti, who runs CX there: “They treated the agent like a piece of the brand, not a side project.” That is the difference. The document was wrong is the longer note. The builder is the substrate.

If the work between tools is the actual mess — Shopify to the desk to the board — read n8n vs Zapier vs Make before you add another AI subscription.

Where does a general builder fit? (YourGPT, late on purpose)

We ship customer agents on YourGPT. That is a stack choice, not a crown. They will ask which builder a studio uses. The honest answer: the one that can hold your knowledge, your tools, and your handoff — and still be reviewed a month later.

The Shopify connector is on every plan, no extra connector fee (YourGPT Shopify). Published annual plans (pricing): Essential $39/month ($59 monthly); Professional $79 ($129) with functions and human escalation; Advanced $349 ($499) with custom branding and no “Powered by”; Enterprise is contact. Credits burn by model, not by ticket. The Shopify page says the agent can read catalog, inventory, and orders; look up status; start a return; recommend; transfer with the thread. Concurrent chats on that integration: 500.

The cons are real. Reindex is a dashboard action or a schedule — their FAQ says so. There is no Loop/Recharge skill list. Refund refusal is a function plus a written rule, which is why the document has to be true. The badge stays until Advanced. Heavy models on every greeting will surprise the credit meter.

That is also why the Chatbase × Shopify conversation keeps circulating. General builders are getting closer to the catalog. The question is still who writes the refusal.

What does cost-to-run actually look like?

Ignore the trial. Price a normal month and a sale month.

• Inbox: $0. The cost is staff time and the tickets that bounce to email because of the login wall.
• Gorgias: helpdesk tier + AI bundle + $1.50 automated overage + $0.36–$0.40 ticket overage. A resolved AI chat can still count as a ticket. Confirm in-account; Gorgias revises the page.
• Tidio: human conversations + Lyro conversations + Flows visitors. Lyro standalone from $32.50/month (50 chats, annual) or about $0.50 a conversation on the agent page.
• eesel: $0.40 × tickets the AI touches. 500 tickets is $200. 2,500 is $1,000. Failed answers still bill.
• Chatbase: plan + $40/1,000 extra credits + $1,188/year to hide the badge.
• YourGPT: flat plan + credit burn. Branding needs Advanced.

If you cannot sit with the sale-week invoice and still want the agent on, you bought the wrong meter.

Which builder should you pick?

• Free, native, early store: Shopify Inbox. Keep a person on the desk.
• Already drowning in a Shopify helpdesk: Gorgias. Read the $1.50 line twice.
• Want a bubble this week, Shopify actions, a human nearby: Tidio on Growth, with a Lyro quota you can defend.
• Already on Zendesk or Freshdesk, hate another inbox: eesel.
• Want a general agent plus Shopify and will maintain sources: Chatbase, or YourGPT if the widget and functions need to be the product.
• Brand surface, designed refusal, n8n into the existing desk: a studio build on a general builder. That is the Aurelia shape. It is also the work we do on agents.

The real estate version of this card — after-hours, budget, listings, calendar, CRM — is here. Same columns. Different job.

Send a week of tickets and the current help centre. We will write back with what an agent should answer, what it should refuse, and which of these six is even in the conversation. The note goes here.

FAQ

What is the best AI agent builder for a Shopify store?

There isn’t one. Shopify Inbox is the free native start. Gorgias is the dedicated ecommerce helpdesk. Tidio and Chatbase are storefront apps with conversation or credit quotas. eesel is the right install if the desk already exists. A general builder (we use YourGPT) is the right substrate when the brand surface and the refusal rules have to be designed. Match the builder to the seven jobs, not to a homepage claim.

How much does a Shopify AI agent cost?

Inbox is free. Third-party apps start in the tens of dollars and climb with volume. Gorgias Starter with AI is $40/month for 50 tickets and 30 automated interactions, then $1.50 per extra automated interaction. Tidio Starter is $24.17/month annual; Lyro starts at $32.50/month for 50 AI conversations. eesel is $0.40 per ticket the AI handles. Chatbase Hobby is $32/month annual. YourGPT Essential is $39/month annual. Confirm every figure on the vendor page before you budget a sale week.

Can an AI agent look up a Shopify order (WISMO)?

Yes, if it is connected to live order data. Inbox can look up orders after the shopper signs in. Gorgias, Tidio (Growth+ Shopify actions), eesel (with Shopify connected), and YourGPT’s Shopify integration all publish order-status skills. The failure mode is not the model. It is a login wall, a stale sync, or a tracking page that still sends the shopper to the carrier.

Will the agent issue refunds it should refuse?

It will, unless refusal is a written rule with a tool permission behind it. Gorgias gates Actions and brand guidance by plan. eesel can be briefed in plain language (“over $500, loop me in”). Tidio and Inbox follow the content you give them, which is not the same as a hard no. On a general builder, refund refusal is a function plus a test set from real tickets. Do not ship the agent until it fails a refund on purpose.

Shopify Inbox vs Gorgias — which should I install?

Inbox if you want free storefront chat and a person still owns the desk. Gorgias if you need a real helpdesk across email, social, and chat, and you will pay per ticket and per automated resolution. Inbox’s agent is early access, storefront-only, and cannot hide the AI notice. Gorgias is the opposite problem: it can do the work, and the meter will show it.

Do I need a studio or can I just install the app?

Install the app when WISMO and a standard return window are the whole job and someone on the team will maintain the knowledge. Hire a studio when the agent has to sound like the brand, refuse like a trained lead, and file into the desk you already run. That is service design plus automation, not an App Store afternoon. We start from the tickets, not from the widget.

Is YourGPT a Shopify chatbot?

No. A chatbot answers. An agent takes action — order lookup, return start, catalog recommendation, human transfer. YourGPT is a general agent platform with a Shopify connector on every plan. We use it when the knowledge, the tools, and the handoff need to be designed. It is not a Gorgias replacement out of the box, and it is not free of a knowledge project.