Optimizing Inbound Processes at DSV

Powerhouse AI background image
Powerhouse AI hero element

Optimizing Inbound Processes at DSV

Powerhouse AI background image

Optimizing Inbound Processes at DSV

Summary & Results

DSV, a Danish logistics multinational needed to enhance its complex, error-prone inbound process. Powerhouse AI was customised to simplify their operations into a two-step image capturing process. Despite challenges such as pallet wrapping and inaccurate box dimensions, the application improved accountability and operational efficiency. A controlled trial found the AI application to be 29% quicker than manual checks. Although accuracy from both methods was under-target (93% manual, 90% AI), the AI detected a manual error, showcasing its potential. Integrated with a learning model, Powerhouse AI promises ongoing refinement and improvements.

29% quicker than manual process

Positive feedback regarding ease of use

Logistics (consumer goods)
Company Size
75,000 employees
Key Features
AI counting
Key Processes


The client receives approximately 19 trucks daily, each loaded with loose boxes that their inbound team had to stack onto about 600 pallets. Each SKU requires a unique stacking pattern to ensure that each pallet for a specific SKU contains the exact same number of boxes.

The complexity of these stacking patterns often involves placing boxes both horizontally and vertically, adding a further layer of difficulty to the process. The checking procedure was conducted manually, a time-consuming and error-prone activity.

The client had clear objectives for the project: to accelerate their current checking process by a minimum of 25%, maintain or increase accuracy, and ensure that the warehouse operators had a positive perception of the application. 


To address these operational hurdles, we tailored the Powerhouse AI application to support the company's inbound process, delivering a solution centred around a user-friendly workflow. Counting and checking became a two-step procedure: firstly, the operator would capture an image of the pallet ID (due to its relatively small size). Secondly, the operator would take pictures of two sides of the pallet to determine the box count. Following these steps, the application would automatically cross-check the captured data with the system records.

Our solution was adapted to meet the unique challenges of the warehouse environment:

  • It was designed to handle pallet wrapping, a common issue in warehouse operations.
  • The application provided access to the captured image as concrete proof of goods received to the client’s supplier, adding an extra layer of transparency and accountability to the process.
  • It delivered quick results, enabling the correct pallets to be put-away promptly, preventing interruptions of the happy flow. In case of any discrepancies, pallets would be sent to quality control.
  • Despite the inaccurate box dimensions data in the client’s WMS, the application was capable of handling this issue by taking images from two sides of the pallet, ensuring accuracy in counts and checks.

Overall, our Powerhouse AI application offered a comprehensive solution tailored to the company's operational needs, significantly enhancing their inbound process's efficiency and accuracy.


To evaluate the new solution's effectiveness, the customer initiated a controlled trial. They selected 50 pallets and proceeded to count and check them, measuring both speed and accuracy. The entire process, from counting to administrative reconciliation, was then replicated using the Powerhouse AI application.

The feedback from users and their manager regarding the ease of use was positive. They appreciated how the application simplified the operators' tasks to mere picture-taking, rendering the job nearly "mindless”

In terms of speed, the Powerhouse AI application demonstrated clear advantages, proving to be 29% quicker than the manual process. This efficiency gain significantly sped up the checking process and improved overall productivity.

As for accuracy, while both the manual and automated processes didn't hit the intended goals, it was important to highlight the room for growth and future improvements. Manual checking resulted in 93% accuracy while the automated process came close at 90%. The shortfall in accuracy was primarily due to the challenges posed by heavy pallet wrapping. However, indicating the potential of the solution, the app managed to spot an error in a manually counted pallet.

To drive improvements in accuracy, we integrated a learning model into the application. As the system continues to be used and learn, it's designed to continuously refine its counting and checking processes.

Want to eliminate manual work?