Management Analytics AI Agent for PrintMarket

Client

A Ukrainian online retailer specializing in print and promotional products, with a broad catalog, a complex product model, and a large volume of operational and marketing data.

Problem

The data needed for management decisions is stored across different systems and has different structures.

Sales, leads, advertising traffic, manager performance, and other metrics had to be collected, compared, and analyzed separately before getting an answer even to a relatively simple business question.

For example:

• how sales changed compared with the previous period;
• which managers show better or worse performance trends;
• which traffic sources generate more qualified leads;
• how conversion is changing;
• which categories or business areas are growing;
• where unusual deviations have appeared.

For such questions, management had to work with several reports or involve an employee to prepare a separate analysis.

As a result, there was an unnecessary delay between asking a business question and getting an answer, while the depth of analysis depended on the time and availability of the people working with the data.

Solution

For PrintMarket, we developed an AI-powered management analytics system that brings company data into a single analytical environment and allows users to work with it through natural language.

Management does not need to build a report in advance or know the structure of the database.

It is enough to ask a question, for example:

“Compare sales for the last two quarters, show the trend by manager, and explain where the largest decline occurred.”

The system determines which data is needed to answer the question, performs the relevant analysis, and presents the result in a clear format.

Depending on the request, the response can include:

• a text-based analytical conclusion;
• tables;
• charts;
• period comparisons;
• breakdowns by managers, sources, or other parameters;
• identified changes and anomalies.

As a result, the AI agent works not as a regular chat with predefined answers, but as an interface to the company’s real business data — allowing users to ask different analytical questions in natural language and receive answers instantly, 24/7.

A separate part of the project is data preparation. Information from different sources may have different structures, formats, field names, and levels of quality, so before analysis we clean, normalize, and structure the data.

Data from CRM, ERP, the website, advertising platforms, spreadsheets, and other sources is brought into a unified logic and automatically loaded into structured BigQuery tables. This creates a stable foundation for calculations, comparisons, and the AI agent’s work with actual business data.