Implementing AI in an E-commerce Company

Implementing AI in an e-commerce business starts with a process audit and ends with a ChatGPT or Claude assistant that responds based on data from your ERP, BaseLinker, and online store. Changelog carries out such implementations for online stores, wholesalers, and logistics companies: it integrates AI with the systems you already use and automates customer support, product descriptions, and translations.

The changelog implements AI only in processes where the return on investment can be quantified: hours of work saved or lower order fulfillment costs.

Let's talk about AI implementation

Thank you! Your message has been sent

Oops! Something went wrong while submitting the form.



What challenges do e-commerce companies
face before implementing AI?

First

I know that AI can help, but I don't know where to start when it comes to implementing AI in my company

The team tested ChatGPT on their own, but no one figured out which process should be automated first or how much benefit it would bring.

Second

My team is wasting time on repetitive tasks

My team manually answers the same customer questions, writes product descriptions, lists products on marketplaces, and copies data between systems. These repetitive tasks take away from the time needed to serve customers and grow sales.

Third

I'm afraid a custom solution will be too expensive

I'm concerned about the high implementation costs and the lengthy development process. I'm looking for a solution that fits our company's budget, and I want to know if the investment will pay off.

Fourth

I have sales data, but I'm having trouble drawing conclusions from it

Sales and returns data are scattered across different systems. I can't quickly use this data to make decisions about inventory and pricing, so I rely on intuition too often.

Fifth

I don't know how to use ChatGPT safely at work

Employees paste customer data and price lists into ChatGPT or Claude, and the company doesn't know whether the provider might use them to train models.

Sixth

I have the data in the ERP system, and I ask questions in ChatGPT

The team uses ChatGPT or Claude, but to answer questions about inventory levels or accounts receivable, they manually extract data from Subiekt, Comarch ERP Optima, or BaseLinker and paste it into the chat.

What does the implementation of AI in Changelog entail?

The implementation of AI in Changelog consists of four components: an AI audit, AI agents for repetitive tasks, MCP servers that connect ChatGPT and Claude to the ERP system, and model selection based on data sensitivity. Each component can be ordered separately, but the pilot phase always precedes the AI audit.

First

AI Audit: Which Processes Will Pay for Themselves the Fastest?

An AI audit is an analysis of a company’s processes and data that identifies where AI will reduce labor costs and how long it will take to pay for itself. In an AI audit, Changelog measures labor hours in customer service, content, sales, and logistics; verifies whether data in ERP, PIM, and OMS systems is suitable for use by AI; and provides a list of 2–3 pilot projects along with their costs and projected payback periods.

Second

AI Agents and AI Automation for Businesses: Customer Service, PIM, and Content

An AI agent is a program that, based on data from a company’s systems, independently performs repetitive tasks and refers cases it is unsure about to a human. Changelog builds AI agents for three tasks: providing customer service responses regarding order status and returns, generating product descriptions for PIM, and bulk translating product pages for new markets.

Customer Service (CS): Preparing responses regarding order status and returns based on the data provided. We establish the rules for approving responses and escalating issues to a consultant during the onboarding process.

PIM, descriptions, and translations: Preparing descriptions and translations of product cards based on product data, with content verification prior to use. The scope of integration with PIM and the store depends on the available integrations.

Third

MCP Server: How to Connect ChatGPT and Claude to Your ERP?

MCP (Model Context Protocol) is an open standard that allows AI assistants—including Claude and ChatGPT—to read data from a company’s systems. Changelog’s MCP server enables an employee to ask a question in a chat, and ChatGPT or Claude responds based on current data from the ERP system, rather than from the model’s memory. Examples of questions answered by the MCP server:

Subiekt GT and Subiekt nexo: “How much of product X do we have in stock, and when was the last delivery?”, “Show me the unsettled documents for vendor Y.”

Comarch ERP Optima and XL: “Which customers have accounts receivable that are more than 30 days past due?”, “Compare this quarter’s sales with the previous quarter’s by product group.”

enova365: sales reports, product records, and documents without the ability to export to Excel.

BaseLinker and SellRocket:“Which Allegro orders have been waiting to be shipped for more than 48 hours?”, “How many returns were there this week, and for what reasons?”

The MCP server from Changelog is read-only by default. Write operations—such as changing an order’s status—are enabled by Changelog on a selective basis and require user confirmation. Every query to the MCP server is logged, and permissions are assigned to specific employees.

Fourth

ChatGPT for Businesses Without Data Leaks: How Do We Choose the Right Model Based on Data Sensitivity?

The changelog selects one of three model variants for each process individually, because not all data requires its own server, and not all data can be sent to the cloud. In all three variants, access is limited to a minimum, and every use is logged.

Option	Where is the data processed?	When to Choose It
A Business Plan for ChatGPT or Claude	At the provider's site, without training the models on the company's data	Team Collaboration in Chat, MCP Servers for ERP and OMS
Model via API	At the provider, without training models, within the integration	AI Agents in Customer Service, Descriptions and Translations for PIM
A model in a private cloud or on a company server	Exclusively on the customer's infrastructure	The most sensitive data, such as personal information and contracts

In addition,

Data analysis and synthesis as an additional application of AI

AI can support the analysis of sales, returns, and inventory levels, as well as the preparation of recommendations for inventory planning. The scope of the analysis and the frequency of updates depend on the data and available connections. The results must be evaluated by the person responsible for business decisions.

Get in touch



or call +48 797 161 330

What You'll Gain by Investing in AI Solutions from Changelog

Implementing artificial intelligence isn’t just about buying another trendy gadget—it’s a strategic investment that instantly frees up tied-up resources in your company. Partnering with Changelog means the following for your business:

First

Significant time savings

Your team saves dozens of hours each week. We eliminate the need to manually create thousands of descriptions in the PIM, translate products in bulk, and tediously answer the same customer inquiries over and over. Your team can finally focus on strategy and sales.

Second

Data Under Control

We tailor the processing method to the sensitivity of the data and the company’s requirements. We determine the scope of access, the location of processing, and the data retention policies prior to implementation. When using ChatGPT, Claude, or an API, selected data may be shared with the model provider. The terms of use, including for training purposes, depend on the selected service, the agreement, and the configuration.

Third

Answers based on data from your systems

Integrating the assistant with the company’s systems allows it to access shared data on products and orders. This does not eliminate errors in the model. In the project, we define how to verify responses and escalate issues to a human.

Fourth

Increased Conversion Rates (Hyper-Personalization)

By implementing smart semantic search engines and recommendations, you enable customers to find what they need instantly—even if they misspell a name or use a colloquial term. This directly reduces the number of abandoned carts and increases AOV (average order value).

Fifth

Instant Scaling (Cross-Border)

Entering foreign markets is no longer a technological or cost barrier. With the help of AI, the process of adapting thousands of product pages to local language requirements and SEO standards is reduced from several months to just a few days.

Sixth

Measurable return on investment (ROI)

You get a system tailored specifically to your pain points. We focus on real-world business processes, so the cost of a dedicated implementation quickly pays for itself through back-office optimization and the automation of customer service operations.

What does the step-by-step process of implementing AI in a company look like?

Implementing AI at an e-commerce company like Changelog involves 5 steps and begins with an audit, not with choosing a tool:

Stage 1

AI Audit

Changelog maps processes and measures how long they take.

Stage 2

Select 2–3 processes

The pilot program includes processes with the shortest payback time and data that is ready for use by AI.

Stage 3

Pilot Program

The changelog implements a single process, such as the MCP server for Subiekt GT or the customer service agent.

Stage 4

Measurement

The changelog compares the time and cost of the process before and after the pilot, based on the company's data.

Step 5

Extension

Subsequent processes are implemented only after the pilot has demonstrated results.

Examples of AI Implementations in E-commerce

Problem	AI Solution	Systems
A sales representative calls the warehouse to check inventory levels	ChatGPT or Claude Assistant via the MCP server	Subiekt GT, Comarch ERP Optima
Customer Service answers the same questions about the status of a shipment	The AI agent checks the status and responds; it forwards difficult cases to a consultant	BaseLinker, ERP, Allegro
Entering a new market requires the translation of thousands of product sheets	Bulk translations and SEO descriptions uploaded to PIM	PIM, online store, marketplaces
Customers can't find products in the store	Semantic Search Engine and Recommendations	Online Store
Intuition-Based Stocking	Analysis of Sales and Returns with a Demand Forecast	ERP, BaseLinker, Excel

We have been cooperating with Changelog in technical support for integration between sales systems. From the beginning, the cooperation can be described as professional. The gentlemen make every effort to ensure that the quality of support is at the highest level. We also cooperate in the implementation of further system integrations. We can always count on quick response in case of need or occurring problems. In addition, the implementation of patches and solving technical problems is also done at a fast pace, the response time is very short, which facilitates work and helps with further development. We look forward to long-term cooperation at a sustained high level.

Agnieszka Pacuk

TRefl sa

We wanted to streamline the work of our store - to better manage e-commerce processes, to maximize the use of available tools and technologies. The Changelog team (formerly DreamTeam) provided a solution created exactly for us, tailored to the specific queries, needs and real capabilities of the company. The implementation of the entire strategy went smoothly and did not reflect on the business at hand. They quickly detected and fixed the main errors that were stopping our operations, while maintaining efficient communication with third parties. I also appreciate that Changelog (formerly DreamTeam) is able to explain quite intricate processes in an understandable way, and its specialists were available to us practically all day.

Daniel Walendziak

Gym Glamour

It is impossible to overestimate the professionalism and commitment that Changelog.co.uk put into creating our website. Their team not only provided us with the perfect tool to present our products to the group, but also understood what our brand is all about. Taking care of the smallest details, they created a site that perfectly reflects our philosophy and aesthetics. Their approach to the project, focus on our needs and perfectionism made the whole process smooth and efficient. Thanks to Changelog.pl, we not only have a functional site, but also a visual gem that sets us apart from the competition. We highly recommend Changelog.pl to any company looking for a professional and reliable partner.

Luke Lener

Golden MArk

Cooperation with Changelog was very successful. The implementation of the changes in the Sellintegro plug-in was successful and saved us a month of one man's work at a cost of 1/3 of an employee's salary, not to mention the savings and avoided mistakes in the future.No other company wanted to take on this task. For the Changelog team, it was an undemanding challenge. Contact with the Project Manager transparent, factual and precise. Minor childhood diseases of the implemented change removed promptly. Great flexibility in the context of cooperation. If I need some additional functionality that Sellintegro does not offer, I will definitely hit the guys from Changelog first.

Kamil Jablonski

Sport Cross

Thanks to our cooperation with Changelog, we implemented BaseLinker and integrated SellIntegro to fully meet the needs of our Vilo brand. The implementation was carried out professionally, and the solutions introduced allowed us to automate key processes and eliminate manual tasks. As a result, we saved significant time resources, and the risk of operational errors was completely eliminated. We would also like to emphasize the exceptional commitment of the Changelog team, especially the Support Department (Adam), which stood out for its responsiveness and readiness to act in any situation. The implementation was completed as planned and on time, which further confirms the professionalism of this cooperation. We heartily recommend it!

Adrian Krzyzanowski

Vilo Home (VOX S.A.)

Read what customers are saying about us

You don't have to take our word for it, see what our business partners are saying.





A team reviewing documents during a meeting at the office

Implementing AI in an e-commerce company: a brief overview

The implementation of AI at an e-commerce company using Changelog begins with an AI audit, which identifies 2–3 processes with the fastest return on investment. The most commonly implemented solutions include MCP servers that integrate ChatGPT and Claude with Subiekt, Comarch ERP, and BaseLinker, as well as AI agents in customer service and PIM. Company data remains under control thanks to business plans that do not require model training or a model in a private cloud.

How We Use AI

We use AI where it makes sense. First, we analyze the process; then we choose the right tool. If AI doesn't reduce the workload or lower costs, we don't implement it.

People are responsible for the results. With AI, we write code, tests, and documentation faster, but every change is reviewed and approved by our specialist.

Data under control. We only use services that do not train models on your data. We grant only minimal access, and every use is logged.

Q&AMP;A

Frequently Asked Questions About Implementing AI in a Company

Cost, MCP integrations, security, and response control—find out what we determine before deploying AI.

01 How much does it cost to implement AI in an e-commerce company?



Changelog provides a detailed quote after the audit, once it is clear how much time the implementation will save.

02 What is an MCP server?



The MCP server acts as an intermediary through which ChatGPT or Claude retrieves data from the company’s systems—such as inventory levels, orders, and accounts receivable from the ERP. An employee asks a question in the chat, and the AI assistant responds based on current data, with read-only access by default.

03 Is it possible to connect ChatGPT or Claude to Subiekt GT or Comarch ERP Optima?



Yes. Changelog builds MCP servers for Subiekt GT and nexo, Comarch ERP Optima and XL, enova365, BaseLinker, and SellRocket. Initially, the server has read-only access.

04 Is ChatGPT for Business Safe for Company Data?



Yes, if the company uses a business plan or an API where the provider does not train models on the customer's data. For the most sensitive data, Changelog runs the model in a private cloud or on the company's server.

05 Can we use the ChatGPT or Claude licenses we already have?



Usually, yes, because MCP servers run on ChatGPT and Claude business plans, which allow you to connect your own connectors. In the AI audit, Changelog checks whether your plan allows this.

06 What if an AI assistant gives a customer the wrong answer?



Changelog's AI assistants respond solely based on data from the company's systems, and when they don't have a definitive answer, they refer the matter to a consultant. During the first few weeks, responses may be subject to human approval.

Implementing AI in an E-commerce Company

What challenges do e-commerce companies
face before implementing AI?

I know that AI can help, but I don't know where to start when it comes to implementing AI in my company

My team is wasting time on repetitive tasks

I'm afraid a custom solution will be too expensive

I have sales data, but I'm having trouble drawing conclusions from it

I don't know how to use ChatGPT safely at work

I have the data in the ERP system, and I ask questions in ChatGPT

What does the implementation of AI in Changelog entail?

AI Audit: Which Processes Will Pay for Themselves the Fastest?

AI Agents and AI Automation for Businesses: Customer Service, PIM, and Content

MCP Server: How to Connect ChatGPT and Claude to Your ERP?

ChatGPT for Businesses Without Data Leaks: How Do We Choose the Right Model Based on Data Sensitivity?

Data analysis and synthesis as an additional application of AI

What You'll Gain by Investing in AI Solutions from Changelog

Significant time savings

Data Under Control

Answers based on data from your systems

Increased Conversion Rates (Hyper-Personalization)

Instant Scaling (Cross-Border)

Measurable return on investment (ROI)

What does the step-by-step process of implementing AI in a company look like?

AI Audit

Select 2–3 processes

Pilot Program

Measurement

Extension

Examples of AI Implementations in E-commerce

Read what customers are saying about us

Implementing AI in an e-commerce company: a brief overview

How We Use AI

Frequently Asked Questions About Implementing AI in a Company

Get a Quote for AI Implementation

Contact information

Company data

Shortcuts

Implementing AI in an E-commerce Company

What challenges do e-commerce companies face before implementing AI?

I know that AI can help, but I don't know where to start when it comes to implementing AI in my company

My team is wasting time on repetitive tasks

I'm afraid a custom solution will be too expensive

I have sales data, but I'm having trouble drawing conclusions from it

I don't know how to use ChatGPT safely at work

I have the data in the ERP system, and I ask questions in ChatGPT

What does the implementation of AI in Changelog entail?

AI Audit: Which Processes Will Pay for Themselves the Fastest?

AI Agents and AI Automation for Businesses: Customer Service, PIM, and Content

MCP Server: How to Connect ChatGPT and Claude to Your ERP?

ChatGPT for Businesses Without Data Leaks: How Do We Choose the Right Model Based on Data Sensitivity?

Data analysis and synthesis as an additional application of AI

What You'll Gain by Investing in AI Solutions from Changelog

Significant time savings

Data Under Control

Answers based on data from your systems

Increased Conversion Rates (Hyper-Personalization)

Instant Scaling (Cross-Border)

Measurable return on investment (ROI)

What does the step-by-step process of implementing AI in a company look like?

AI Audit

Select 2–3 processes

Pilot Program

Measurement

Extension

Examples of AI Implementations in E-commerce

Read what customers are saying about us

Implementing AI in an e-commerce company: a brief overview

How We Use AI

Frequently Asked Questions About Implementing AI in a Company

Get a Quote for AI Implementation

Contact information

Company data

Shortcuts

What challenges do e-commerce companies
face before implementing AI?