Some links here may earn us a small commission. It has no bearing on what we recommend, and we have turned down partnerships that would have.

A 2026 buyer’s guide to Gorgias, Richpanel and Remark. Pricing verified September 10, 2026.

We get some version of this question from nearly every client who has read a headline about AI shopping and wants to know where to start, and the short version of what we tell them is that the question is premature, because there is no single tool called an AI shopping assistant to evaluate. There is a support desk that has learned to close a sale while it resolves a complaint, a merchandising layer that happens to talk back to the shopper in plain language, and a genuinely new category where the software completes the transaction on the customer’s behalf rather than merely encouraging one. Give us roughly ten minutes and you will leave knowing which of those three you are actually shopping for, and what Gorgias, Richpanel and Remark cost once the AI is priced in.

What this guide covers

Why “AI shopping assistant” describes three separate products, and how to tell which one a given founder actually needs.

The variables that decide the true monthly cost, since ticket and interaction volume are billed separately from the base plan.

Gorgias, Richpanel and Remark in full, with every published tier and what triggers the next one.

A side-by-side table, then a recommendation sorted by the problem you are actually trying to solve.

The questions founders ask us most, including whether an AI agent can be trusted with a refund and what happens when it gets a recommendation wrong.

The short version. If the problem is a support queue quietly costing you sales because tickets sit unanswered while a customer decides whether to buy or bounce, Gorgias is built for that, and its AI Agent add-on scales with resolution volume rather than seats. If customers ask pre-purchase questions your team never sees because they abandon rather than email, Richpanel’s self-service layer is the cheaper, faster fix, gated by order volume rather than tickets. If you sell a considered, higher-price product where the conversation itself is part of the value, the way a good salesperson on a boutique floor is part of the value, Remark is a different kind of purchase, closer to hiring a tireless clienteling advisor than to installing a helpdesk. Everything below explains how we sort a brand into one of those three, and where the line between them blurs enough to need a second opinion.

The question that actually sorts this, before any vendor’s name comes up

Ask where the conversation currently breaks down for your customers, not where you would like a robot to sit, because the honest answer does most of the sorting for you. A brand whose inbox already strains a two-person team has a resolution problem, and the fix is an agent that closes routine tickets correctly and hands the rest to a human without the customer noticing the seam. A brand whose product page gets plenty of traffic but a thin conversion rate usually has a pre-purchase question problem, where shoppers leave silently rather than ask, and the fix there is closer to a concierge than a helpdesk. A brand selling something expensive enough that customers want a second opinion before buying, apparel at a real price point or furniture that has to fit a specific room, needs a digital version of the person who used to stand on the sales floor. None of this is strictly about order volume, and a brand doing a modest number of monthly orders at a high average order value can justify the pricier, more conversational tier well before a pure order-count threshold would suggest it is ready, so treat the roughly one thousand orders a month mark we use to define the mid-market band, $2 million to $50 million in revenue, as directional rather than a rule. Two of the three vendors here also meter their AI add-ons by ticket or interaction volume rather than by seat, a detail that belongs in the decision as much as the base price does.

What the record shows about answers, trust and the agent doing the talking

The most directly relevant independent source here is not a market research report but a specification. The Agentic Commerce Protocol, published jointly as an open standard under an Apache 2.0 license and documented at agenticcommerce.dev, defines how a business stays the merchant of record while an AI agent initiates and completes a checkout on a customer’s behalf, retaining control over which products can be sold, how they are presented, and how the order gets fulfilled. That distinction, between an assistant that recommends and an agent that transacts, is precisely the seam this article is trying to locate, and the fact that a standards body spelled out merchant-of-record protections tells you the transacting category is further along, and further scrutinized, than most founders assume. Separately, the Baymard Institute’s running tally of fifty cart abandonment studies puts the average documented abandonment rate at 70.22 percent, a figure that has stayed stubbornly high for years despite a wave of chat widgets promising to fix it, a useful check against any claim that a chat bubble alone will move it without the underlying answer quality being right.

Metering is the whole ballgame. Gorgias bills against ticket volume with the AI Agent priced as a per-account add-on carrying its own interaction ceiling, so a busy inbox pays twice, once for the base plan and again for automation past the included interactions. Richpanel splits its product line in a way that is easy to miss. The self-service layer that most resembles an AI shopping assistant is priced by monthly order volume, while its separate agent-seat helpdesk is priced per user, so confirm which product a quote actually covers. Remark does not publish pricing at all, because it scopes every deployment on storefront sessions, conversion rate, average order value and support ticket volume, so the honest number to budget is whatever that scoping conversation produces. None of the three lets you buy the AI layer without the platform beneath it, so the real comparison is never assistant versus assistant, it is the whole stack each vendor asks you to run your support or merchandising through.

Gorgias: the support agent built to close a sale while it resolves a ticket

Metered by monthly ticket volume, with the AI Agent priced as a separate add-on billed against automated interactions. The vendor states it now powers customer conversations for roughly forty percent of Shopify brands, a claim worth treating as a vendor claim rather than an independently audited figure.

Verified September 10, 2026 from gorgias.com/pricing and re-verified October 29, 2026, with no changes. Overage on tickets runs from $0.40 down to $0.36 per ticket depending on tier, billed separately from AI Agent overage. Custom pricing is available above 5,000 monthly conversations.

Gorgias earns its reputation because the base helpdesk and the AI layer share the same ticket data, so an automated reply and a human reply read as one continuous conversation, and the Pro tier is marketed explicitly around AI that resolves and sells, closing a return conversation with a store-credit upsell rather than stopping at resolution. The revenue analytics on higher tiers let a support team prove, in dollars, that the automated layer generates orders rather than just cutting cost. The real cost of the layered metering is that a brand scaling past both its ticket ceiling and its interaction ceiling in the same month gets billed on both counters at once, and the jump from Starter to Basic changes the ticket allowance sixfold while barely moving the interaction ceiling. It suits a brand whose primary need is a support problem wearing a sales opportunity.

In their words

By Gorgias’s own account, it is built for ecommerce businesses where customer conversations are tied closely to the order experience, to conversion and to repeat purchase, bringing support across every channel, ecommerce workflows, and AI-assisted support and shopping assistance into one platform so a team can clear routine interactions efficiently while keeping service quality and human judgment in view. Gorgias is candid about where it fits less well: it is strongest when support volume and customer experience are meaningful levers for the business, and a very small store with few tickets, or an organization outside commerce looking mainly for general case management or an internal service desk, may be better served by something simpler or more general-purpose.

“At Gorgias, we see customer conversations as part of the commerce experience, not a separate afterthought. We bring support, automation, and shopping assistance together so ecommerce teams can handle service and sales in the context of how they actually sell.” Lauren Knehnetsky, Senior Commercial Partner Manager, Gorgias

On how it views our work, Gorgias’s team wrote: “Atelier’s emphasis on removing excess before adding complexity, and aligning design, technology, and commercial performance, resonates with our view of durable ecommerce growth. We respect an approach that treats the customer experience and the systems behind it as part of one business problem.” Gorgias

Lauren Knehnetsky

Senior Commercial Partner Manager, Gorgias

Richpanel: order-aware self-service that keeps a question from becoming a ticket

The self-service, AI-shopping-assistant-facing product is metered by monthly order volume rather than tickets or seats, and the vendor positions this line explicitly as a lower-friction substitute for a traditional ticket queue.

Verified September 10, 2026 from richpanel.com/pricing. Richpanel‘s separate agent-seat Help Desk product is priced per user per month and is a distinct purchase from the order-metered self-service line above; confirm which line item a quote refers to before comparing it to a competitor.

Richpanel’s self-service line is the closest thing here to a genuinely free entry point, since a store under a thousand monthly orders can run order tracking and a basic FAQ flow at no cost and prove whether deflecting pre-purchase questions actually moves conversion before committing to a paid tier. Because it is metered by order volume rather than ticket volume, the bill is indifferent to how chatty customers happen to be, rewarding a brand that fields a lot of low-value questions rather than punishing it. It sits in a middle position, doing more than a pure helpdesk by answering questions before they become tickets, while still falling short of the true discovery and merchandising layer covered in the companion guide, since its flows are built around order status rather than helping a shopper compare products they have never bought. The real limitation is that the free and Pro tiers cap out fast, and the step to Pro Max at $360 a month is large relative to what it adds.

Remark: the agent built to talk a shopper through a decision, then complete it

Remark publishes no pricing and no range; every deployment is scoped to the brand on storefront sessions, conversion rate, average order value and support ticket volume, which is standard for a bespoke implementation rather than a self-serve widget install.

Verified September 10, 2026 from remark.ai, which has no pricing page; pricing model confirmed directly by Remark, October 27, 2026.

Remark’s own site describes it as a single connected system that learns from every conversation and is trained through real expert dialogue, positioning it for health and wellness, luxury fashion and home furnishings brands where a shopper benefits from something closer to advice than to answers, and its live demo shows the assistant fielding a question about a dress’s fit and then guiding the shopper toward adding it to the bag and checking out with a stored payment method, the transacting behavior that separates this category from the first two. That places Remark closer to the agentic commerce end of the spectrum, since the conversation functions as the sales conversation itself, with checkout completed inside the same chat rather than the shopper routed back to a product page. The honest caveat is that this is a narrower fit, built for considered purchases where hesitation is about fit rather than order status, and it carries none of the transparent pricing the other two publish, so evaluating it requires an actual sales conversation.

In their words

By Remark’s own account it is built for brands where the quality of the shopping experience matters as much as the transaction itself, and it is particularly suited to premium brands whose products benefit from guidance, discovery, education, or fit and compatibility expertise, along with a more personalized path to purchase. Its stated aim is to bring the experience of a great in-store advisor online: understanding what a shopper is looking for, helping them navigate a catalog, answering nuanced product questions and making relevant recommendations throughout the journey, with virtual try-on and product visualization carrying that beyond the conversation so a shopper can see a product in context, understand fit and style, and buy with more confidence.

“There’s still a gap between how people shop online and the kind of experience they get from a great advisor in a store. The best advisors understand what a customer is looking for, help them make sense of their options, and give them confidence in the decision. Remark is built to bring more of that experience online.”

Theo Satloff

Chief Executive Officer, Remark

On how it views our work, Remark wrote: “What we respect about Atelier’s approach is their belief that a great ecommerce experience should answer the questions a shopper actually has, rather than simply try to persuade them harder. Their focus on clarity, reducing friction, and helping customers find the right information at the right moment is very aligned with how we think about the shopping experience at Remark: the best experiences help people make better, more confident decisions.” Remark

Side by side

Pricing verified September 10, 2026 from each vendor’s own pricing page.

For the brand whose support inbox is the actual bottleneck, tickets piling up faster than a small team can answer them, Gorgias with the AI Agent add-on is the more complete answer, since automation and the human queue share one system and one customer record. For the brand that suspects it is losing sales to unanswered pre-purchase questions but has never proven it, Richpanel’s self-service line is the cheaper way to find out, and the free tier means the test costs nothing but the time to install it. For the brand selling a considered purchase where a real conversation about fit or styling justifies the price, Remark is worth a demo even though it will not show up on a spreadsheet next to the other two, because it solves a fundamentally different problem than either helpdesk. A brand doing meaningful revenue at a lower order count, the kind of high-average-order-value business where a single sale funds a week of software, should not rule out the pricier or more bespoke option purely because its order volume looks small on paper.

Is an AI shopping assistant the same thing as a chatbot?

No, a chatbot generally follows a scripted decision tree, while the tools in this guide use a language model to read unscripted questions and either resolve them, deflect them, or in Remark’s case carry them through to a completed purchase.

Can these tools actually process a refund or a return on their own?

Gorgias and Richpanel can both be configured to approve routine returns and refunds within rules you set, such as a value ceiling or a return window, while anything outside those rules routes to a human, the setup we recommend for any brand just starting with automation.

What happens when the AI gives a customer the wrong answer?

All three vendors log every automated conversation for review, and the practical fix is a weekly audit of a sample of resolved conversations in the first month, tightening the rules or the knowledge base each time the agent gets something wrong.

Do I need a certain order volume before an AI shopping assistant makes sense?

No fixed number applies, and a high-average-order-value business can justify one of these tools well before a pure order-count threshold like the roughly one thousand orders a month referenced elsewhere in this series would suggest.

Is this the same thing as the discovery and merchandising layer other guides mention?

No, and clearing up that confusion is the reason this guide exists: a support-and-sales agent like Gorgias, an order-aware self-service tool like Richpanel, and a true discovery and merchandising layer, covered in the companion guide on strategic merchandising, are three different purchases.

Can an AI agent complete a purchase without a human ever touching it?

Yes, that is precisely the category Remark and the emerging agentic commerce standards are built around, where the assistant initiates and completes checkout on the customer’s behalf while the business remains the merchant of record throughout.

Will an AI shopping assistant hurt my brand’s voice?

Only if it is trained on generic scripts rather than your own past conversations and product language, which is why every vendor here emphasizes training the agent on real transcripts rather than shipping a one-size-fits-all persona.

How long does it take to see whether one of these is working?

Plan on a genuine month before judging it, since the first two weeks are mostly spent correcting the agent’s early mistakes and the resolution rate and attached revenue only stabilize into a trustworthy number in the second two.

Where to go next

If your real gap is the merchandising and discovery layer rather than a support agent, read the companion guide to that category next, which covers what that layer is and the point at which it becomes the constraint on growth. If you would rather have us sort your situation than sort it yourself, our app and retention stack work covers this kind of evaluation, and you can see how we think about vendor selection in our conversion and experience practice. Get in touch and we will walk your data against this framework with you.