AI assistant for e-commerce
An AI assistant for your online store,built on your real catalogue.
The AI assistant for e-commerce is a shopping chat that knows your store's catalogue, filters and policies. It helps visitors choose the right product, checks stock and order status directly in the platform, and hands the conversation over to your team when a person is needed.
Visitors leave when they cannot find an answer quickly
In an online store, many questions come up right before a purchase: “Will it fit my model?”, “When will it arrive?”, “Can I return it?”. If the answer arrives by email the next day, the customer has already bought from another store.
- Product pages contain the information, but customers do not know which filter or specification matters for their needs
- Questions about order status take up a large part of the support team's time
- A generic chatbot can invent stock levels, prices or delivery times, and that damages trust
- Abandoned carts get no follow-up at all or, on the contrary, receive messages sent without a legal basis
Who it is for
- Online stores with large catalogues, where customers get lost among variants, sizes or technical specifications
- Customer service teams that get the same questions about delivery, returns and order status every day
- Brands selling products that need a recommendation (car parts, cosmetics, equipment, building materials)
- Stores that want to recover abandoned carts without sending unsolicited messages
What the AI assistant does for your online store
- 01
Guidance to the right product
Asks a few clarifying questions and suggests products from the catalogue, explaining why they fit based on their real attributes.
- 02
Answers about availability
Checks stock and price directly in the platform at the moment of the question, without guessing or memorising them.
- 03
Order and delivery status
After verifying the customer's identity, reads the order status and the AWB (courier tracking number) and explains the next step in plain language.
- 04
Return and warranty policies
Answers from the terms and conditions published by the store and sends the link to the official page.
- 05
Cart recovery messages
Prepares personalised messages only for customers who have agreed to receive communications, according to the configured rules.
- 06
Handover to the team
When a question falls outside the rules (a complaint, a special order, an unhappy customer), passes on the conversation with its full context.
How it works
How a conversation works
The steps marked in yellow are approved by a person.
- Input
- AI
- Action
- Human in the loop
- Result
- 01
Input
Customer's question
The visitor types in the widget on the site or on the product page.
- 02
AI
Understands the need
The assistant identifies the intent and asks clarifying questions if details are missing.
- 03
Action
Checks the catalogue
Queries products, stock, price or the order in the platform in real time.
- 04
AI
Writes the answer
Suggests products or explains the status using only the data received from the system.
- 05
Human in the loop
The team handles exceptions
Complaints, discount requests or unclear cases go to a colleague.
- 06
Result
Answer with a link
The customer receives the answer and the link to the product, order or policy.
What store data the assistant uses
- The product catalogue with structured attributes (dimensions, compatibility, materials, variants)
- Current stock and price, read via API at the time of each question
- Order and delivery data, accessible only after the customer is verified
- Published policies: delivery, returns, warranty, payment methods
- Consent status for marketing communications, from the email platform or CRM
- Tone rules and limits set by the team (what the assistant never promises)
Possible integrations
- WooCommerce
- Shopify
- PrestaShop
- Magento / Adobe Commerce
- ERP or inventory management system
- Couriers (Fan Courier, Sameday, Cargus) for AWB status
- Klaviyo or Mailchimp
- Helpdesk (Zendesk, Freshdesk, Gorgias)
- Google Analytics 4
- WooCommerce
- Shopify
- PrestaShop
- Magento / Adobe Commerce
- ERP or inventory management system
- Couriers (Fan Courier, Sameday, Cargus) for AWB status
- Klaviyo or Mailchimp
- Helpdesk (Zendesk, Freshdesk, Gorgias)
- Google Analytics 4
Each integration's compatibility is checked during the analysis phase, for the systems and versions you use.
Control and transparency
What stays with your team
- Any discount, voucher or price exception that is not already published in the store
- Complaints, return requests outside the policy and situations with unhappy customers
- Cart recovery message templates and sending rules, approved before launch
- Order changes (address, product, cancellation), which are made by the team or only with explicit confirmation
Limits and how we manage them
- Recommendations are only as good as the catalogue attributes are complete, which is why we check them during the analysis
- It does not show stock or price if the system cannot provide them at that moment; it says it does not have the information
- It does not send cart recovery messages without valid consent and an unsubscribe option
- It cannot guarantee delivery times; it only communicates the estimate given by the courier or the store
- For products with safety or medical implications, it recommends consulting a specialist
How we measure the quality of the store assistant
We agree the indicators together before launch and track them afterwards. The values depend on the process, the data and the volume.
- Accuracy of product data
- We compare answers with stock and price in the system on a sample of conversations, checked by the team
- Relevance of recommendations
- The team assesses, on a set of real questions, whether the suggested products match the need described
- Human handover rate
- We track how many conversations reach the team and why, so we can adjust the rules
- Assisted conversions
- We measure in analytics the orders that went through a conversation with the assistant, compared with the rest
- Compliance of cart messages
- We regularly check that every message sent has recorded consent and an unsubscribe link
How we implement the assistant in your store
- 01
Analysing the catalogue and the questions
We start from customers' real conversations and emails and check how complete the product attributes are.
- 02
Connecting to the platform
We set up read-only access to the catalogue, stock and orders, plus customer identity verification.
- 03
Testing on real questions
We run the assistant on a set of questions from your history, and the team corrects the answers and the rules.
- 04
Gradual launch
We start with a few categories or pages, monitor the conversations and expand once the results have been verified.
Use cases
The scenarios below are conceptual examples that show how the assistant can be used. They do not describe delivered implementations.
- Car parts
Compatibility by model and year
The customer gives the make, model and year of the car, and the assistant filters compatible parts from the catalogue and flags when compatibility needs to be confirmed by a colleague.
Conceptual example - Cosmetics
A routine by skin type
The assistant asks three questions about skin type and preferences and suggests products in stock, with the ingredients shown from the official product sheet.
Conceptual example - DIY and construction
Approximate quantity calculation
For a customer who wants to tile a surface, the assistant estimates the number of boxes based on the dimensions in the product sheet and recommends checking with a consultant.
Conceptual example
Frequently asked questions
Can the assistant invent a price or a stock level?
We configure it to show price and stock only from the response received from the platform at that moment. If the system does not respond, the assistant says it does not have the information and offers a way to get in touch instead.
What happens when the assistant does not know the answer?
It acknowledges that it does not have the information, does not guess, and offers a handover to the team or a contact form with the customer's question already filled in.
How are customer and order data protected?
Order data is shown only after the customer is verified (for example, order number plus email). The assistant's access is limited to what it needs, and conversations are kept according to the retention policy agreed with you.
Can we send cart recovery messages to all visitors?
No. Marketing messages are sent only to people who have consented to receive such communications, with a clear unsubscribe option. The assistant checks the consent status before preparing the message.
Does it work with our store platform?
The common platforms have APIs for the catalogue and orders, but we check actual compatibility during the analysis stage, including for custom modules or your own ERP.
How is it different from the AI agent for customer support?
The e-commerce assistant works mainly before the purchase: it guides visitors to the right product in the catalogue and answers questions about stock and delivery. The support agent handles post-order requests, complaints and tickets. A store can combine the two.
Does the assistant replace the site search?
No, the two complement each other. Search remains for people who know what they want, while the assistant helps when a customer needs a recommendation. For better results in classic search, there is the semantic search solution.
Can’t find the answer? Write to us and we’ll reply directly.
Want an assistant that knows your catalogue?
Tell us which platform you use and which questions you get most often. We will suggest a realistic starting point, in the categories where it matters most.
Related solutions
- Semantic searchSearch by meaning across your website, catalogue or documents: relevant results even when users do not type the exact words.
- AI agent for customer supportAnswers customers from approved information, checks order status in your systems and hands over to colleagues the cases that need a person.
- AI agent for lead qualificationAsks the qualification questions agreed with your team, scores each lead against the agreed criteria and sends it to the right person.