AI sales agent for ecommerce
that answers, advises and sells 24/7.
An AI agent for your online store. It answers product questions from your catalog, recommends the right product and size, clears up doubts at checkout and hands over to your team when a person is needed, on your website, WhatsApp, Instagram and Telegram.
- Answers only from your data
- Human handover built in
- AI disclosure as the EU AI Act requires
Does the Trail jacket work for cycling in the rain? I'm 182 cm and usually wear M.
Yes. It's fully waterproof and cut longer at the back for riding. At 182 cm, M fits most customers. Choose L if you wear a thick layer underneath.
checked · catalog · size chart · reviews
What's delivery to Rotterdam, and can I return it?
Free delivery over €50, arriving Thursday. Returns are free within 30 days. Shall I add the Trail jacket in M to your cart?
added to cart · Trail jacket · M
In short
What is an AI sales agent for ecommerce?
An AI sales agent for ecommerce is software that talks to shoppers in chat and messaging apps, answers their questions from the store's own catalog, stock, prices and policies, recommends products and helps them complete the purchase. Unlike a rule-based chatbot, it understands questions written in the shopper's own words. Unlike a general AI assistant, it is connected to the store's systems and only answers from verified data. NewProWeb builds these agents and integrates them with online stores, with a handover to human staff for anything the agent should not handle.
The problem in numbers
Online stores lose sales in conversations that never happen.
53%
of pre-purchase questions at the median store arrive outside weekday business hours.
56%
of e-commerce sites fail to adequately support what shoppers search for.
+21.7%
conversion when shoppers got a generative-AI pre-sale chatbot instead of no service, in a randomised experiment with 44,614 shoppers.
Free AI audit
How much of this happens in your store?
Tell us about your store, your channels and the questions your team answers every day. We'll email you a free audit that shows whether an AI sales agent pays off for you, and where to start.
- About 3 minutes
- Report by email
- No call, no obligation
Problems and solutions
Where online stores lose sales, and what the data says.
Every problem below is backed by published research, linked next to each figure. Where a source is a vendor study or covers a single market, we say so.
01/ Problem
Shoppers ask questions when nobody is there to answer.
Online shopping peaks in the evening and at weekends, but most service teams work office hours. A question asked at 21:00 often gets its answer the next morning, after the shopper has moved on.
At the median store, 53% of pre-purchase questions arrive outside Monday to Friday, 9:00 to 18:00. The median first reply takes 2.2 hours during business hours and about 15 hours after hours.
72% of the 100 largest online stores in Belgium and the Netherlands offered no evening customer service in test purchases during Black Friday 2025.
At Galaxus in Germany and Switzerland, more than one in five orders from January to May 2021 was placed between 20:00 and 23:00.
How the agent solves it
Answers in seconds, at any hour.
The agent replies immediately, day and night, from your catalog, stock and policies. Conversations it can't finish go to your team with the full history, so the morning shift starts with context instead of a backlog.
- 24/7 answers in your store's languages
- Instant help on product pages, in the cart and at checkout
- Overnight handovers queued for your team, with context
02/ Problem
Slow or missing answers cost sales.
Before buying, shoppers want to be sure: does it fit, does it work with what I have, when will it arrive? Without a quick, accurate answer, many simply leave.
86% of consumers say responsiveness and accurate resolution strongly influence their purchase decisions, and 74% now expect 24/7 service. Survey of 6,182 consumers in 22 countries, including Germany, France, Spain, Italy, the Netherlands and the UK.
In a randomised field experiment with 44,614 shoppers on a large cross-border retail platform, a generative-AI pre-sale chatbot raised sales by 16.3% and conversion by 21.7% compared with shoppers who got no service. Returns and product ratings did not get worse.
How the agent solves it
Accurate answers before doubt turns into an exit.
The agent handles the questions shoppers actually ask before buying: fit, compatibility, materials, delivery dates, payment and returns. Every answer comes from your data, so it is specific to the product and to the shopper's situation.
- Product, compatibility and availability questions
- Delivery dates and costs for the shopper's location
- Payment, warranty and returns explained in plain words
03/ Problem
Search fails when shoppers describe what they need.
Shoppers don't search like a catalog. They type "a gift for a runner under €100" or "a charger for my laptop", and most on-site search engines don't understand requests like that.
56% of e-commerce sites fail to adequately support users' search needs. Problems are most common with non-product queries (66% of sites), compatibility queries (44%) and use-case queries (43%).
71% of consumers expect personalised interactions and 76% are frustrated when they don't get them. Personalisation most often lifts revenue by 10 to 15%.
How the agent solves it
A shop assistant that understands the request.
Shoppers describe what they need in their own words. The agent asks one or two clarifying questions, then recommends a short list from your catalog and explains why each item fits.
- Natural-language product finder and comparisons
- Compatibility checks against your product data
- Recommendations limited to items in stock
04/ Problem
Seven in ten carts are abandoned.
Some abandoned carts are just browsing. Many others end at a doubt in the last step: an unexpected cost, a delivery date, the returns policy, trust in the store.
The average documented online cart abandonment rate is 70.22%, based on 50 studies.
Among US shoppers who abandoned a checkout, the top reasons were extra costs that were too high (40%), slow delivery (20%), not trusting the site with card details (19%), an unsatisfactory returns policy (13%) and not seeing the total cost up front (12%).
How the agent solves it
Doubts answered the moment they appear.
The agent is available in the cart and at checkout to answer questions on total cost, delivery and returns with exact figures. With the shopper's consent, it can follow up later in their channel. Discounts appear only if your rules allow them.
- Total cost, delivery dates and returns on request
- Consent-based follow-ups in WhatsApp, Telegram or email
- Your discount rules, never invented offers
05/ Problem
Wrong size, wrong product, expensive returns.
In fashion and many other categories, a large share of returns comes from buying the wrong size or a product that doesn't match expectations. Every return costs shipping, handling and margin.
Zalando reports that about 50% of ordered items are returned across its markets, and that its size and fit tools prevented 9% of size-related returns in 2026.
Size is the number one reason for returns at Zalando.
Surveyed apparel retailers attribute 70% of returns to poor fit or style.
Nearly one in four parcels sent to online shoppers in Germany was returned in 2021.
How the agent solves it
Size and fit advice before the order.
The agent uses your size charts, product measurements and fit notes to recommend a size, and tells shoppers honestly when a product isn't right for them. Fewer surprises at delivery mean fewer returns.
- Size recommendations from your size charts and fit notes
- Honest answers when a product doesn't fit the need
- Clear expectations on materials, colours and use
06/ Problem
Shoppers want to message you, not wait in a queue.
Customers increasingly expect to reach a store the way they reach their friends, in WhatsApp, Instagram or Telegram. Most stores answer these channels by hand, if at all.
72.4% of consumers say they are more likely to buy from a brand that offers messaging, and 66.8% feel frustrated when it isn't offered. Survey of 11,056 adults in 22 markets, commissioned by Meta.
WhatsApp reached 90% of UK online adults in May 2025, and 74% used it every day.
44% of the 100 largest online stores in Belgium and the Netherlands use WhatsApp for customer service.
How the agent solves it
One agent in every channel.
The same agent works in your website chat, WhatsApp, Instagram, Telegram and Messenger, with one knowledge base and one set of rules. This is conversational commerce without extra staff: shoppers can start on Instagram and finish on your website without repeating themselves.
- Website chat, WhatsApp, Instagram, Telegram and Messenger
- One knowledge base and one rulebook for all channels
- Conversation history kept per customer
07/ Problem
Every live conversation costs money, and peaks overload teams.
Routine questions take up most of a service team's day, and hiring for Black Friday or seasonal peaks is slow and expensive.
Live channels such as phone, chat and email cost an average of $8.01 per contact, compared with about $0.10 for self-service.
The average hourly labour cost in the EU was €34.9 in 2025.
Klarna's AI assistant handled two-thirds of its customer service chats in its first month, 2.3 million conversations, and cut the time to resolve an errand from 11 minutes to under 2.
In 2025, Klarna said cost had become too dominant a factor and quality had suffered, and it began hiring people again so customers can always reach a human. The lesson: let AI handle routine work and keep people for the rest.
How the agent solves it
AI for routine questions, people for what matters.
The agent takes the repetitive questions, so your team spends its time on complex orders, complaints and key customers. It scales with traffic during peaks without extra hiring, and hands over instead of guessing.
- Routine questions answered without a queue
- Scales through Black Friday and seasonal peaks
- Your team focuses on complex and high-value cases
Free AI audit
See which of these problems an AI agent would solve in your store.
Our engineers rank the use cases by value and effort for your store, and tell you honestly which ones aren't worth it yet.
- Use cases ranked for your store
- Pilot proposal included
- Free, no obligation
What it does
What the AI sales agent does in your store.
Product answers from your catalog
Specifications, materials, compatibility and availability, read from your product data rather than guessed.
Guided selling
Asks clarifying questions, recommends a short list and explains the differences, like a good shop assistant.
Size and fit advice
Recommends a size from your size charts, measurements and fit notes, and says when a product runs small or large.
Delivery and returns
Exact costs, delivery dates and return conditions for the shopper's location and order.
Cart and checkout help
Adds items to the cart or shares a checkout link where your platform allows it, and explains payment options.
Order status
Looks up orders and delivery status for verified customers, which takes routine "where is my order" questions off your team.
Consent-based follow-ups
Reminders about a saved cart or a product back in stock, sent in the shopper's channel only with their permission.
Human handover
Passes the conversation to your team in your help desk or CRM, with the history and a short summary.
Insights from every conversation
Shows the most common questions, missing product information and the reasons shoppers hesitate.
How it works
From your catalog to a working agent.
Connect your data. We connect the agent to your catalog, stock, prices, policies and order system, and agree what it may and may not do.
A knowledge base that stays in sync with your store
Pilot on real conversations. We test the agent on real questions from your past chats and emails, then launch it on one channel with success criteria agreed up front.
Measured results and a go / no-go decision
Roll out to your channels. Website chat, WhatsApp, Instagram or Telegram, with handover to your help desk or CRM.
One agent wherever shoppers ask
Improve every week. We review conversations, fill gaps in product data and instructions, and track conversion and running costs.
An agent that gets better with use
Integrations
Works with your store and channels.
We connect the agent through each platform's API, with access limited to what its job needs. If part of your stack isn't listed, we build the connector.
Store platforms
Channels
CRM and help desk
AI models
Trust and control
An agent your customers can trust, and you can control.
Shoppers are wary of chatbots, and with reason. Our agents are built to avoid the failures that make customers give up on them.
Answers only from your data. Prices, stock, delivery and policies come from your systems. If the answer isn't there, the agent says so and hands over instead of guessing. Companies are held responsible for what their chatbots say: in 2024, a Canadian tribunal held Air Canada liable for its chatbot's wrong answer about a refund.
A person is always one step away. 87% of customers say access to a human agent is essential when companies use generative AI in service, and only 27% would try a chatbot again after a bad experience. Handover is built in, with the full conversation.
Clear AI disclosure. Since 2 August 2026, the EU AI Act requires that people are told when they are interacting with an AI system. The agent introduces itself as an AI assistant in its first message.
Your rules for prices and offers. The agent never invents discounts or promises. Offers follow the rules you set, and anything outside them goes to your team.
Data protection by design. Access rights limited to the agent's job, minimal personal data, and hosting that matches your requirements, including keeping data in the EU.
Logs you can review. Every conversation and every action the agent takes is logged, so you can check what it said and why.
What we measure
Results you can check, not promises.
We don't promise a fixed uplift: results depend on your traffic, catalog and processes. Instead, we agree targets for the pilot and report against them.
- Conversion from chats
- The share of shoppers who buy after talking to the agent, compared with similar shoppers who didn't.
- First response time
- Seconds instead of hours, measured day and night.
- Resolved without a person
- The share of conversations the agent completes, and the reason for every handover.
- Revenue influenced
- Orders placed after a conversation with the agent.
- Size-related returns
- For fashion and footwear, returns of items bought after the agent's size advice.
- Cost per conversation
- Model and hosting costs per conversation, tracked every month.
FAQ
AI sales agent questions
It is an AI system that talks to shoppers in chat and messaging apps, answers their questions from your catalog, stock and policies, recommends products and helps them complete the purchase. It is connected to your store's systems and hands over to your team when a person is needed.
Rule-based bots follow fixed scripts and break when a question doesn't match them. A general AI chatbot understands free text but may not know your products. Our AI sales agent combines both strengths: it understands the shopper's own words and answers only from your verified store data, and it can act in your systems, for example by checking stock or adding an item to the cart.
Store platforms such as Shopify, WooCommerce, Magento / Adobe Commerce, Shopware, BigCommerce and custom stores via API; channels such as website chat, WhatsApp, Instagram, Messenger, Telegram and email; and CRMs and help desks such as HubSpot, Salesforce, Zendesk and Gorgias. So it can work as an AI sales agent for Shopify, for WhatsApp, or for both at once.
No. The agent answers from your systems and is instructed not to answer when the information isn't there. In that case it says so and hands the conversation to your team. Every conversation is logged so you can review it.
Yes. In the EU, the AI Act's transparency rules apply from 2 August 2026: people must be told when they are interacting with an AI system. Our agents introduce themselves as AI assistants. Because shoppers judge disclosed bots more critically, answer quality and an easy route to a person matter even more.
Yes. It can ask the right questions about quantities, delivery and terms, create a lead in your CRM and pass it to your sales team with a summary.
It depends on your catalog, systems and channels. We start with a pilot on one channel and agree the timeline after the free AI audit, once we know your setup.
The pilot and integration are quoted after the free AI audit, based on your scope. Running costs depend mainly on conversation volume and the model used, and we estimate them up front and track them monthly.
We set it up for GDPR: the agent only accesses the data its job needs, personal data is kept to a minimum, conversations are logged, and hosting can stay in the EU. Where we process personal data on your behalf, we sign a data processing agreement, as GDPR requires.
Sources
- 01Baymard Institute, cart abandonment rate
- 02Baymard Institute, e-commerce search benchmark
- 03Gorgias Ecom Lab, 2026 (vendor research)
- 04Ecommerce News Europe, 2026 (Returnless, Salesupply, Sendcloud)
- 05Galaxus, 2022
- 06Zendesk CX Trends 2026
- 07Fang et al., arXiv 2510.12049 (working paper)
- 08McKinsey, personalisation, 2021
- 09Zalando, size and fit
- 10Zalando, returns, 2025
- 11McKinsey, apparel returns, 2021
- 12University of Bamberg, 2022
- 13Meta / Kantar, State of Business Messaging
- 14Ofcom, Online Nation 2025
- 15Gartner, 2019
- 16Eurostat, 2026
- 17Klarna, 2024
- 18CX Dive, 2025
- 19Gartner, 2026
- 20Moffatt v. Air Canada, 2024 BCCRT 149
- 21EU AI Act, Article 50
See what an AI sales agent would do in your store.
Tell us about your store, your channels and the questions your team answers every day. We'll send you a free AI audit with the use cases worth doing first.