GoBrise
AI chatbot for e-commerce

AI chatbot for e-commerce stores

Shoppers ask before they buy, and they ask at the worst possible times — late at night, mid-checkout, during a promotion when your queue is already full. A chatbot that answers from your own shipping, returns and product policies keeps those conversations moving, and passes anything about a specific order straight to your team.

  • Enquiry volume multiplies during promotions and seasonal peaks
  • Pre-purchase questions decide the sale within minutes, not hours
  • The same delivery, returns and sizing questions repeat all day
  • Order-specific problems still need a person, not a canned reply

Last reviewed 17 September 2026

Quick summary

$0

Human live chat entry point

The Free plan includes 25 AI replies / month, unlimited human live chat and a non-removable “Powered by GoBrise” badge.

1 thread

AI and agent context

When a person takes over, the existing conversation stays intact and the AI stops replying.

24/7

Policy-answer availability

Approved shipping, returns and product guidance can be answered outside store hours.

500

Starter AI conversations

Starter includes 500 AI conversations per month; one conversation can contain several messages.

What shoppers actually ask

These are the conversation types an e-commerce assistant handles every day. Each answer is drawn from content you have written and approved, not from general knowledge.

Order status and delivery times

Dispatch windows, carrier options, cut-off times and delivery estimates answered from your published shipping policy, at any hour. Anything tied to one specific order is handed to an agent with the whole conversation attached.

Returns, exchanges and refunds

Return windows, condition requirements, who pays return postage and how long a refund takes — answered consistently every time. Refund decisions and complaints escalate to a human by design.

Product and sizing questions

Materials, dimensions, compatibility, care instructions and sizing guidance pulled from your own product content, so the answer matches what is on the product page.

Stock, pricing and promotions

Restock policy, discount code rules, bundle terms and price-match policy answered directly in the widget, instead of becoming an abandoned cart.

How it works on a store

Load your shipping, returns and product policies into the knowledge base. The AI agent may only answer from that content, so it repeats what you have already published rather than inventing a policy. When retrieval is weak, when the assistant is unsure, or when a shopper asks for a person, the conversation escalates and your team continues in the same thread.

  • Grounded answers from your own store policies and product content
  • Handles peaks without seasonal headcount for repeat questions
  • Escalation to a human for order, refund and complaint decisions
  • White-label branding so the widget matches your storefront
  • Unlimited human live chat on every plan, including Free
  1. Step 1

    Choose the questions the assistant should own

    Start with repeatable questions whose answers are already controlled by the store: delivery zones, dispatch cut-offs, return windows, exchange conditions, product dimensions, materials, care instructions and promotion rules. Keep judgement calls out of the first release. A damaged parcel, a disputed refund or a request to amend an order belongs with a person because the correct answer depends on the order record and your team's authority.

  2. Step 2

    Prepare one clear source for each policy

    Add the current policy text to the knowledge base and remove contradictions before launch. If domestic and international returns differ, state both and name the regions. If a product has compatibility restrictions, write them explicitly. The assistant retrieves passages from this material, so a concise, complete source is more useful than several pages that disagree. Growth and higher plans can crawl website content and update it automatically; on lower plans, your team edits sources manually.

  3. Step 3

    Install and style the storefront widget

    Place the GoBrise JavaScript snippet on the store theme or through the site's script-management area. Configure the logo, colour and welcome message so the chat entry point feels native to the storefront. Growth and higher plans remove GoBrise branding; Free and Starter retain the published branding treatment. Test product pages, cart, checkout-adjacent pages and mobile layouts before exposing the widget to all shoppers.

  4. Step 4

    Test real questions, including awkward ones

    Use wording customers actually type: misspellings, short follow-ups and questions that combine two policies. Check whether an answer cites the right rule and whether an order-specific or unsupported request reaches a person. Include boundary tests such as an expired return window, a delivery location you do not serve and a discount combination your policy forbids. Fix weak answers by improving the source content rather than adding a vague instruction to sound confident.

  5. Step 5

    Run the daily queue as one conversation stream

    Routine policy questions can be answered immediately. When retrieval is weak and the model is uncertain, it confidently cannot answer, or the shopper explicitly asks for help, the thread escalates. An agent sees what the shopper asked and what the AI already said, then continues in that same thread. Once a human replies or takes over, the AI remains silent in the conversation, avoiding competing answers.

  6. Step 6

    Review gaps after campaigns and policy changes

    Look for repeated escalations after a sale, new product launch or shipping change. A cluster of unanswered questions is usually a content task: add the missing promotion term, compatibility note or regional policy and test it again. Update time-sensitive copy when a campaign ends. The aim is not to prevent every handoff; it is to automate questions with stable answers and preserve human judgement where the outcome depends on an order or exception.

Who an AI chatbot for e-commerce is — and isn't — for

The useful dividing line is not store size. It is whether a meaningful share of your questions have stable, written answers and whether a person is available for the exceptions.

A strong fit when

  • Your team repeatedly answers shipping, returns, sizing, product-care or promotion-policy questions.
  • You have accurate policies and product information that can become the assistant's approved source material.
  • You want immediate answers outside working hours but still want agents to own refunds, complaints and order changes.
  • You need a branded website chat experience and one thread for automated replies and human follow-up.

Probably not the right fit when

  • Most enquiries require live access to order, warehouse or carrier systems that are not connected to your support workflow.
  • Your policies change informally and are not written down, because the assistant cannot reliably infer rules your team has not documented.
  • You want the bot to approve refunds, modify orders or make case-by-case commercial decisions without human review.
  • You need a complete commerce suite with merchandising, email marketing, fulfilment and marketplace operations in one product.

Frequently asked questions

Does GoBrise integrate directly with Shopify?

GoBrise is installed on a storefront with a JavaScript widget, so it can appear on a Shopify storefront when the theme or store setup allows custom scripts. Do not assume that this means native access to Shopify orders, customers, refunds or inventory. The current public product is strongest at answering from approved store content and handing account-specific work to a person. Confirm any required Shopify data integration with the GoBrise team before choosing a plan.

Can it tell a shopper where a specific order is?

It can explain published dispatch windows, delivery methods and tracking instructions from your knowledge base. A live status for order 1234 requires access to that order's data and suitable identity checks. Without a confirmed order-system integration, GoBrise should hand that request to an agent rather than expose or invent order information. This boundary is intentional: general policy is safe to automate, while a customer's record needs verified context.

Can it handle refund policy differences by country or region?

Yes, if the regional differences are clearly written in the source content. Name each covered country or region, its return window, exclusions, postage rule and refund timing. Avoid uploading two generic policies that conflict without explaining where each applies. The assistant can then retrieve the relevant rule from the shopper's question; ambiguous cases or exception requests should still be escalated for a human decision.

Will it recommend products or help with sizing?

It can answer factual product questions from the material you provide, including dimensions, materials, compatibility, care instructions and published size guidance. It should not invent a fit guarantee or claim stock it cannot verify. For useful recommendations, document the selection criteria in plain language and test common comparisons. Where suitability depends on personal judgement, safety or information not supplied by the shopper, route the conversation to your team.

What happens when the knowledge base does not contain the answer?

A missing answer should become a handoff, not a polished guess. GoBrise evaluates both the strength of the retrieved content and the model's own certainty. Weak evidence alone does not force a handoff if the assistant can still give an appropriate grounded response, but uncertainty, an explicit request for a person or a confident statement that the material cannot answer the question triggers escalation. The waiting agent receives the existing transcript.

Can a human take over without the AI continuing to reply?

Yes. AI and human messages share one conversation. If an agent replies or uses takeover, the AI stops speaking in that thread permanently. A conversation that was automatically escalated before any agent touched it may return to AI handling on the shopper's next message, but human involvement is the hard boundary. This prevents a shopper receiving two competing answers while an agent is resolving an order or complaint.

How should we prepare product and policy content?

Prioritise completeness and consistency over volume. Give each important rule a direct heading, include the conditions and exceptions, and state where regional rules apply. Product entries should carry the facts customers use to decide: dimensions, materials, compatibility, care and limitations. Remove expired promotion text and duplicate policies. Then test the exact phrases found in your support inbox; those tests reveal missing context faster than reviewing the content in isolation.

Can we use it only for human live chat at first?

Yes. The Free plan includes 25 AI replies / month, unlimited human live chat and a non-removable “Powered by GoBrise” badge. That gives a team a way to install the widget, organise incoming conversations and learn what shoppers ask before paying for AI volume. Starter includes 500 AI conversations per month, while Growth and Business increase the allowance. Human-handled conversations do not count as AI conversations, and one AI conversation may contain several back-and-forth messages.

Does the widget support white-label branding?

The widget can use your logo, colours and welcome message. Growth and Business include white-label presentation without GoBrise branding; Free and Starter use the published GoBrise branding treatment. If a fully branded support experience is important, compare the current plan details before rollout rather than assuming every visual option is included at every tier.

How do we evaluate it before showing it to every shopper?

Build a test set from real pre-sale and post-purchase conversations. Include easy policy questions, short follow-ups, misspellings, conflicting requests and cases that should reach a person. Check factual correctness, not just tone. Verify the widget on desktop and mobile, and confirm agent alerts and takeover. Launch on a limited set of pages or during staffed hours first if your theme setup supports that, then broaden coverage after the content gaps are fixed.

Explore by use case

AI chatbot for dropshipping businesses

Set honest shipping-time expectations automatically from your supplier data, while stuck-order questions reach a person. Start free.

AI chatbot for order tracking and returns support

Handle the return-policy and shipping-process questions that dominate your inbox, with real order-status lookups reaching a person unless connected. Start free.

AI chatbot for Shopify store owners

Answer Shopify shoppers' order, shipping and return questions instantly from your own policies, with account-specific lookups escalated to a person. Start free.

AI chatbot for subscription box businesses

Handle recurring billing, pause and cancellation questions instantly from your published terms, with account-specific disputes escalated to a person. Start free.

E-commerce & Retail solution

How the platform is used across online retail teams.

Customer support

Take the repetitive questions, keep people for the rest.

Human handoff

How escalation to a live agent works in practice.

Pricing

Free plan with unlimited human live chat, paid AI tiers from $29.

Answer the questions that decide the sale

Start free with unlimited human live chat, load your policies, and let the AI cover the hours your team cannot.