ST Logistics, a Singapore-based supply chain solution provider faced inefficiencies in manually inputting up to 2,500 medicine order lines daily. They enlisted Powerhouse AI to automate this laborious task. The solution involved scanning order labels, which Powerhouse AI processed to extract essential data and compiled into an easy-to-upload Excel sheet. To ensure accuracy amidst challenges like poor label print quality, the system alerted users to manually verify difficult-to-read labels. The results transformed the laborious, 10-hour task into a few minutes' job, substantially increasing operational efficiency and maintaining high data accuracy.
Reduction of label processing time from 10 hours per day to a couple of minutes
Improved workflow integration: tailored to their warehouse processes
Robust accuracy maintenance: system alerts in case not fully certain
The client is offering medicine repacking services. Each day, they prepared approximately 600 medicine orders, destined to be dispatched to patients throughout Singapore. The process started with their supplier sending physical labels, each one meticulously detailing the type and quantity of medicine required for each specific order.
However, the challenge arose with the manual transfer of information from these labels into the client's systems. Given that each label usually contained between 3 to 6 line items, the team found themselves manually inputting an average of 2,500 lines each day. This laborious task was not only time-consuming but also offered significant scope for human error. With each label requiring around a minute to be keyed into the system, the team was investing about ten hours of labor each day on this single task alone.
Recognizing the inefficiencies of this manual system, the client sought an automated solution to streamline the process, with the ultimate goal of saving time and reducing the scope for potential errors.
In response to the client's unique challenges, our goal was to design a simple and effective solution that could eliminate the need for manual data entry. To this end, we configured the Powerhouse AI system to perfectly suit their specific process.
Users were asked to affix ten labels to both sides of an A4 template. These templates were then scanned using a standard scanning machine, with the images being sent automatically to our Powerhouse AI system.
Powerhouse AI, would then extract the necessary information from these scanned images. This extracted data was compiled into an Excel sheet and sent back to the users via email. The layout of this Excel sheet was designed to be compatible with the format requirements of their Warehouse Management System (WMS), thereby enabling the team to upload it easily and efficiently.
This innovative solution, however, was not without its challenges. We were not permitted to connect directly to their systems, and the quality of the label printing varied, with some labels being hard to read. Moreover, the labels often contained handwritten text or pen markings, which our system was not intended to process.
To address these hurdles, we incorporated additional measures. First, we chose to provide output via email, bypassing the need for direct system integration. If the AI encountered labels that were hard to read due to poor print quality, we set up notifications to prompt the user to manually verify the data, ensuring accuracy. Lastly, our system was fine-tuned to disregard any handwritten text or markings.
The client chose our solution because it not only streamlined their process but was also specifically crafted to suit the unique demands of warehousing operations, ensuring seamless integration with their existing workflow.
With the newly implemented solution, the customer experienced a substantial increase in operational efficiency. The daily application of our system transformed the time-consuming label checking and keying process. A task that used to consume 10 hours each day was cut down to just a few minutes.
Despite the continued presence of misprinted labels, our solution effectively mitigates this issue by alerting the user when it encounters labels it's not completely sure about, ensuring the high accuracy of data capture. This review process, performed on the computer, typically involves editing just a few characters.
