AI sales and support agents for your Cmsmart Ecommerce or WooCommerce store

An assistant that does the intake and the lookups, while a named person keeps every price, promise and exception.

Book a free assessment How we run projects

Quoted per project, no obligation. The assessment is free and carries no price list.Updated 10 October 2026

In short

What can an AI agent safely do for an ecommerce or print shop?

An AI agent can turn a free-text enquiry into a complete specification, answer order-status questions from your order data, and draft a quote from your own pricing rules. A named person approves every quote and owns credits, dates and exceptions. Cmsmart builds it as custom development with a person in the loop, after a free assessment.

Key facts: AI sales and support agents
Scope	A chat or email assistant for sales enquiries and order-status questions on your store
What you get	Complete specifications from free text, answers grounded in your catalogue and orders, and a hand-over to a person when it is unsure
Typical inputs	Your catalogue and product rules, the source of your prices, sample enquiries, the order-status data source and escalation contacts
Deliverables	The agent on staging, a written list of what it never decides, tests on your own enquiries and a run book
How pricing works	Free assessment, then a quote per project. The model provider is named in the project agreement.

How it works

From an enquiry to an approved quote

From an enquiry to an approved quote. The language model does not work out or make up prices. It asks the buyer to supply any missing details, gets the price from your rules or engine, and only ever shows a quote that a named person has approved.

The problem

Where do AI assistants go wrong in a store?

An invented price or delivery date
A discount promised that nobody approved
An answer about an order that the system does not hold
Customer data sent to a model provider without a written agreement
A bot that never hands over to a person

Fit

Is this the right service?

A good fit	Another route fits better
Many enquiries arrive as free text and need a structured specification before anyone can quote	A small shop with few enquiries: a good form is enough
You have pricing rules, or an engine, that the agent can ask for a price	There is no price source yet: the agent can still gather specifications for your estimators, and automatic prices come later
A person is available to approve each quote	You want an agent that sets prices or approves credits on its own: we do not build that

What we do

What is in scope?

Part	Input	Deliverable
Specification extraction	A free-text enquiry	Structured fields, and a question for each field that is missing
Price lookup	Your pricing rules, price matrix or engine	A price from your rules, never computed by the model
Quote drafting	The approved price and the specification	A draft quote for a person to approve
Order status	Order, shipment or production data	Answers read from the record, and 'not known' when the record is silent
Escalation	Complaints, credits, custom items and legal questions	A hand-over with the conversation and the order reference
Data handling	Model provider, processing region and retention	Written into the project agreement before the build
Evaluation	Your own past enquiries	A test set, and adversarial tests of the list of what the agent never decides

The Cmsmart angle

What the plugin's AI tools are, and what an agent is

The AI tools in Cmsmart Product Designer act on files: they remove a background, upscale a low-resolution image or check a print file. An agent acts on conversations and orders. They can run in the same store.

Core AI tools that ship in the Cmsmart plugin

AI Print Preflight. Checks the buyer's image resolution in the designer and the print PDF in the admin order. It uses AI tokens from your Cmsmart wallet.
AI tools for print shops. Background Remover, Image Upscale and Print Preflight as one intake line.

Custom Built as custom development

The sales and support agent itself
The link from the agent to your pricing rules or engine
The order-status lookup against your order or production system
Escalation to your helpdesk or an estimator

Boundaries

What the agent never decides

We write this list with you before the build and test every line of it with adversarial prompts. The pattern is always the same: the agent hands the decision to the person who owns it.

Decision	Who owns it	What the agent does instead
Prices, discounts and price matching	Your pricing rules, then an estimator	Asks your rules for the price and shows nothing but an approved quote
Delivery or dispatch dates	Whoever runs the production schedule, with a named person	Repeats the date held in the order record, or says that none is set
Reprints, credits and payment reversals	A named person	Logs the request and passes it on together with the order reference
Whether artwork is acceptable	Prepress	Directs the buyer back to the file upload and its preflight check
Made-to-order or off-catalogue products	An estimator	Passes on the conversation and whatever specification exists so far
Complaints and legal questions	A named person	Acknowledges receipt and escalates at once

Testing

How the agent is tested

We measure extraction on a sample of your own past enquiries that was kept out of the build, and the attacks we try follow the published list of risks for applications built on large language models, such as prompt injection and improper output handling.

Source: OWASP Top 10 for LLM and generative AI applications (read 10 October 2026).

Scope

Included, not included, and extra

Clear lines before you start. The final scope is written into your proposal.

Included

A written scope and a never-decides list
Specification extraction with questions for missing fields
Price lookup from your pricing rules or engine
Order-status answers read from your data
Hand-over to a person, with the conversation attached
Tests on your own enquiries and a run book

Not included

Setting prices, discounts, delivery dates or credits. A person owns each of them.
A pricing engine, unless it is scoped separately
Model fees charged by the provider
Sourcing or buying of any data
Licences, hosting and third-party fees (plugins, fonts, images, services, provider accounts)
Anything that is not written in the approved scope

Extra, quoted separately

A pricing engine or a price matrix import
AI tools for files in the plugin, such as print preflight
Reporting on what customers ask

What we need from you

Your catalogue and the rules for each product
Where prices come from, and who approves a quote
A set of real past enquiries, with the answers that were sent
Where order and production status live, and who handles escalations
A staging copy of your store, or approval to create one
Someone on your side who can make decisions when a question comes up

Process

How the work runs

Each phase is a milestone, approved and invoiced one at a time. The timeline is set in the assessment because it depends on your store, your products and your systems.

Free assessment. We read your brief, your store and your constraints, and tell you whether the service fits and what it would involve.
Written scope and quote. Included, not included, extra and what we need from you, set out before any work starts.
Build on a staging copy, in milestones you approve one at a time in your dashboard.
Acceptance with your own products, files and orders, including the awkward ones.
Handover with a run book, then support through tickets in your dashboard.

This service in three phases

What happens, and what you receive

Phase	What happens	What you receive
1. Scope	We agree the enquiries the agent handles, the price source, the escalation contacts and the never-decides list.	A signed scope and boundary list
2. Build	We build the intake, the price lookup, the status lookup and the hand-overs on staging.	The agent working on your sample enquiries
3. Accept and hand over	We measure extraction on a held-out sample, run adversarial prompts against the boundary list and train your approvers.	An acceptance record and a run book

Free assessment

Start with a free assessment

Describe your store and what you need. We reply with whether the service fits, what it would involve and a quote, with no obligation.

Book a free assessment How we run projects

FAQ

Frequently asked questions

Does a person always check a quote before it is sent?

Yes. The agent only prepares a draft from the price your rules return, and a named person must approve it before the buyer receives it.

We do not have a pricing engine. Is that a blocker?

Not for the first stage. The agent can still gather full specifications for your estimators, and automatic prices follow once there is a pricing engine or a price matrix.

Which AI model does it use?

The provider is chosen per project. The project agreement states who it is, where the data is processed and how long it is kept, before the build starts.

How do you know the agent behaves?

We test extraction on a held-out sample of your own enquiries, and we run adversarial prompts against every line of the never-decides list. The check is repeated when the model, the prompt or the data changes.

How does an agent differ from the AI tools that come with the plugin?

Those tools act on files, such as removing a background or checking a print PDF. An agent talks to buyers and reads orders. A store can use both.

Can you work with a client in my country?

Yes. Every Cmsmart service, from custom development, migration and integration to full store builds, is delivered remotely. Clients in any country and any industry can work with the team; Netbase has worked with clients in 80+ countries. Projects run online, through a written proposal, email or WhatsApp, and tickets in the Cmsmart dashboard.

Do you work in my industry?

Yes. Clients in any country and any industry can work with the team; Netbase has worked with clients in 80+ countries. Published Cmsmart projects include stores for web-to-print, merchandise and apparel, signage, packaging and labels, and gifts and promotional products. See the client work at cmsmart.net/projects.

How do I pay securely?

Contracts and payments go to the company, Netbase JSC: plans are paid through PayPal; services are invoiced by Netbase JSC under a written proposal and paid by bank transfer to Netbase JSC's company account or by PayPal Business. Cmsmart (Cmsmart Ecommerce) is the ecommerce brand of Netbase JSC, a company based in Hanoi, Vietnam.

Who do I talk to?

Talk to the CEO, Huy Nguyen (David): message him on WhatsApp at +84 937 869 689 or email project@cmsmart.net. The contact page at cmsmart.net/contact lists every channel, and Request a quote (cmsmart.net/project-proposal) sends your project brief to the team. Tell us the scope and we reply with a plan and a quote.

Related services

Learn more

The team behind Cmsmart and its published work

2012Netbase since 2012Cmsmart Ecommerce is a Netbase JSC company
6,300+client projectssince 2012
80+countriesclients worldwide
17,133proposals writtenin our project system, live count

Client projects and countries: Netbase portfolio library and WorkSuite project records, as of Sep 2026. Proposals: records in our Cloodo WorkSuite project system, all-time, refreshed every 6 hours.

Talk to an ecommerce expert

Huy Nguyen (David)

Founder & CEO, Cmsmart · Netbase JSC

Every request is read by the team before you get a written scope and estimate.

Request a proposal Chat with David on WhatsApp

Published Cmsmart projects

Project records from our portfolio. They show the wider work of the team, not this exact service.

Stores on Cmsmart Ecommerce

AI sales and support agentsFree assessment

Cmsmart Ecommerce provides custom ecommerce and WooCommerce development services for web-to-print, product personalization, plugin support and API integration.

Cmsmart Ecommerce expert team

Web-to-print, WooCommerce, B2B and AI specialists who plan, build and support stores together.

Meet the team →

Top