Optimization of price quote preparation for cardboard packaging manufacturers

The client faced a long and inefficient process of preparing price quotes, requiring 3–7 employees across 3 departments. Preparing a quote took up to 2 days, while the long-term success rate was only 8% — meaning teams spent 92% of their efforts on quotes that never turned into orders. In a competitive environment where speed and accuracy are crucial, the complicated process resulted in lost profits and higher operational costs.

We developed an AI solution that automates price quote preparation based on historical data, order parameters, and internal pricing models. The system recommends the optimal price, generates materials for sales staff, and enables faster decision-making. This minimizes manual work, shortens response times, and improves calculation accuracy. With a clear and simple web interface, the entire process can now be handled by a single salesperson entering just a few basic order parameters, receiving results in seconds. For added trust, the application displays the most similar completed orders so salespeople can validate results if needed.

Sales department, management.

6

Python, ML frameworks (scikit-learn, TensorFlow/PyTorch), database systems, integration with ERP

Head of IT / Data / Technology

Full implementation

Operation is provided by Gauss Algorithmic via its proprietary MLOps cloud platform, with long-term support and regular model retraining.

IT/ERP specialists, sales department as end users and validators of results.

Standard database and server infrastructure, cloud environment for model training, ERP system integration. Operating costs remain at a fraction of the original manual process, with retraining performed 2–3 times per year.

The price quote preparation process became significantly faster and more accurate. Sales staff receive data-backed price recommendations based on completed orders, increasing trust in the quoting process and helping uncover insights previously overlooked. Frustrating manual tasks were eliminated, freeing up valuable capacity for R&D and quality control specialists to focus on their core roles.

Team reduction from 3–7 people across 3 departments to a single salesperson. Preparation time reduced from 1–2 days to seconds. Model accuracy 96%+.

“In today’s dynamic world, our customers expect to receive quotes without unnecessary delays. Implementing the AI model for quote creation has brought us a major boost in efficiency and speed. It also allowed our developers and technologists to focus on innovation and creativity instead of routine tasks. This shift opens up new opportunities and contributes to the company’s growth.” – Martin Hejl, Managing Director, THIMM.

High-quality data preparation, integration with the existing ERP system, close collaboration with IT and sales teams, and iterative development enabling fast feedback. Strong project sponsorship from senior management.

Heterogeneous and incomplete historical order data required extensive cleaning and structuring — the client only successfully exported the required data on the 6th attempt. Another challenge was the learning curve for sales staff, who initially distrusted the AI recommendations because they differed from their manual calculations, incorporating additional costs upfront. These challenges were resolved through continuous communication and joint workshops with the client.

Start with a pilot project focusing on a clear use case where AI has a high business impact. Dedicate time to data preparation and cleaning, and involve end users from the beginning — their trust and adoption are critical to success.