People approve
The assistant proposes; your team checks matches before an offer or order goes on.
AI assistants that work on your product data.
Assistants that match products to enquiries and orders, and answer customers from your catalogue.
When it makes sense
Your team searches for products instead of selling them.
Most enquiries ask for products you already have. The time goes into finding them.
- Sales searches several systems to answer one enquiry.
- An enquiry with a list of items is matched to products by hand, line by line.
- Customers ask by an old code, a competitor name or a short description.
- Shoppers cannot find a parameter in the store and call instead.
Product data in numbers
10M+ SKUs in our clients' systems.
The catalogues an assistant searches. Figures from our own projects.
- 10M+SKUs in our clients' systems
- 70+PIM and e-commerce implementations
- 300+projects delivered
- 30+experts and consultants
- 2010operating since
- 35+PIM implementations
- openProd MCP serverassistants search your real product data
- LemonHub.AI, our productanswers shoppers from your PIM data
Sources: LemonMind, November 2026
How it works
From an enquiry to a checked offer.
The assistant reads the enquiry and searches your products through the openProd MCP server. People decide what goes out.
Input
ERP
prices and stock
Customer enquiry
e-mail, PDF or a list of items
Product catalogue
your PIM data
openProd MCP serverto the assistant
Assistant
AI assistant
reads each line and searches your catalogue
- codes
- attributes
- alternatives
- quantities
matches
Review
Your team
checks the matches and approves
prices and stock from ERP, approved orders to ERP
sent as
Output
- Offer
- Order
- Answer in the store
approved orders to ERP
prices and stock from ERP
Which enquiries could an assistant handle?
30 minutes, no obligation. Show us real enquiries and we pick one type for a pilot.
Use cases
Three assistants we build on your data.
Assistants search your real catalogue, so they propose products you actually sell.
Rules
Rules we agree before the first enquiry.
An assistant is useful only when everyone knows what it may do.
Agreed data scope
We agree which data the assistant may read and which systems it may write to.
Real products only
It searches your catalogue through the openProd MCP server, so it proposes what you sell.
Measured on your enquiries
The pilot is judged on real enquiries against a measure agreed at the start.
Pilot
Start with a pilot on one type of enquiry.
- 01
Enquiry review
You show us real enquiries or orders. We pick one type, for example orders sent as a PDF list.
ResultOne enquiry type and a measure - 02
Pilot
We connect the assistant to your catalogue and run it on real enquiries. Your team checks every match.
ResultMatches on your enquiries, checked by people - 03
Rollout
If the pilot works, we add enquiry types, channels and the ERP, with rules for review.
ResultAn assistant in daily sales work
Case studies
Product selection and AI on product data, in practice.
A product selection program and LemonHub.AI in the PIM: the catalogue data an assistant works on.
Product selection tool
Heatpex
An online selection tool for the Heatpex Aria ventilation system: engineers design a system, save projects and download a PDF, and Heatpex gets B2B leads.
Read case study
AI product content
Ceramika Paradyż
Generative AI in the PIM and online shop of Ceramika Paradyż: product descriptions and translations prepared with LemonHub.AI.
Read case studyFAQ
Before the first pilot.
What is the openProd MCP server?
openProd, our own product at openprod.io, has an MCP server for AI tools. Through it an assistant searches your product data, so it chooses from your real products. More on openProd.
Does the assistant send offers or orders by itself?
No. It proposes matches, and your team checks them before an offer or order goes on.
Is this the same as LemonHub.AI?
Partly. A shopping assistant in your store can use AI.Assistant from LemonHub.AI. Matching products to enquiries and orders is built for your process and uses the openProd MCP server.
Can it work with our ERP?
Yes, where the use case needs it, for example to check prices and stock or to pass an approved order on. We agree this in the pilot scope.
Which AI models do you use?
We choose the model for each task and for your data rules, and agree which data the assistant may read.
How do we start?
Show us one real enquiry. We pick one type of enquiry for a pilot and agree how to measure it. Background: RAG on your own product data.
Olgierd MrozikCEO, LemonMind
Start with one enquiry
Show us one real enquiry.
In a free 30-minute call we look at your enquiries and orders and agree a pilot on one type.
30 minutes. Google Calendar opens in a new tab.
- 2010 founded
- 300+ projects
- 70+ PIM and e-commerce implementations