Haeger and Schmidt AI delivery bills

Delivery bills // Case study

Haeger & Schmidt Logistics

Company profile:

Founded in 1887, Haeger & Schmidt Logistic has established itself as a leading player in the logistics industry, specializing in Rhine shipping and international transport. With 250 employees and 12 locations in Europe, the company handles 8.5 million tons of freight and 350,000 TEU annually.

Out

demands

Complex delivery note processing

The task was to automate the manual and error-prone process of processing delivery notes, which involved scanning, typing and entering the data into the ERP system. The variety of delivery note formats and languages posed a particular challenge.

Solution approach

Multi-level AI integration

Developing a solution required an innovative approach, as the variety and complexity of the delivery bills challenged the capabilities of generative AI models. A multi-stage processing system that includes both advanced document recognition and semantic analysis has been implemented.

1st Workshop // Use Case Definition

Inspire and motivate
Brainstorming
Definition of AI use cases (usually 3 use cases)

2. assessment

Three days on site with all relevant departments
Review of the use cases
Roadmap for implementation

3. design sprint

Work shadowing in the relevant departments
Sprint 0: Definition of all work packages/delivery packages and quality targets

4. development of the solution

Sprint 1: Creation of a working prototype
Sprint 2: Testing and tuning the prototype

5. delivery of the solution

Review of quality objectives
Delivery in the customer’s environment

The result

Increased efficiency and employee satisfaction

The automation of the process has significantly increased efficiency, virtually eliminated manual effort and improved the quality of data entry. The solution led to a noticeable reduction in the workload for employees and greater satisfaction in their day-to-day work.

Technical highlights – innovative power through AI

Microsoft Intelligent Documents:

Initial step to recognize and convert approx. 80% of the delivery note fields into structured data.

Azure AI for semantic analysis and language processing:

Use of Azure AI (OpenAI GPT-4 in the proprietary Azure Cloud) to interpret and semantically process the recognized fields and to process and understand complex language nuances in the delivery bills.

Finetunig the processing logic:

Development of specific finetuning prompts for the detection and classification of the remaining 20% of the fields.

Automated table processing:

Implementation of a special solution for extracting and processing data from complex table structures.

Provision of a web assistant:

The results of the AI are reviewed by the person in charge and there is an opportunity to provide feedback for further fine-tuning.

Integration into the ERP system:

Seamless connection of the processed data to the existing ERP system to optimize the workflow and increase data accuracy.