CONTEXT & CHALLENGE
Constraula is the urban maintenance and city services company of the Sorigué group. It is mainly based in the province of Barcelona. Constraula's growth is structured based on the ISO 9001, ISO 14001, and ISO 45001 standards (for quality, environmental management, and occupational health and safety, respectively). They were currently in the process of implementing the ISO 50001 standard for energy efficiency and verification of the ISO 14064 Carbon Footprint. This should help them improve metrics, evaluate, and implement more precise action plans for environmental objectives. Constraula's challenge is to reduce their CO2 footprint by 5% over the next 5 years based on 2022 data. However, the company faces challenges related to its limited capacity to extract and efficiently use information.
Specifically, the company aims for an innovative approach that allows better control over the CO2 footprint of products purchased during the project phase. The goal is to understand the real footprint of each project produced by the materials used. It is currently possible to associate an estimated carbon footprint with a volume of almost any type of material (m^3 of concrete, m^3 of sand, a pallet of tiles, etc.). However, the information on the materials used is not collected in a format that facilitates access to the information. In general, this data is found on the invoices issued by suppliers, usually in PDF format. Each supplier uses their own invoice model, which greatly complicates any attempt to automate the information from these invoices. This makes it more challenging to create a generic system that allows extracting the information, even from invoices in previously unseen formats (for example, from a new supplier).
This challenge is common in the sector and is exacerbated by the generally poor level of digitization in the construction industry. This results in two significant technical challenges: (a) the fragmentation of data sources due to the use of multiple independent digital tools; and (b) the inability to analyze this data cross-sectionally and consistently due to a lack of interoperability between these digital tools. Additionally, the data is often unstructured, and there's significant heterogeneity that prevents cross-sectional analysis of the information.
SOLUTION
Enter Beawre. With our revolutionary AI-driven technology, Beawre Automate can automatically extract data from diverse invoice formats. By integrating Optical Character Recognition (OCR) with advanced machine learning algorithms, Beawre's solution can decipher varying invoice structures and extract essential material data. Leveraging Large Language Models (LLM), our system classifies types of items in invoices, calculates their weights and quantities, and addresses the challenge of varied nomenclature by introducing a system for standardized data interpretation.
Beawre can manage hundreds of invoices per project automatically, and the processed data can be easily dumped into any format, whether it be a customer's specific system or simply Excel files stored in a SharePoint.
The primary objective of this contract was to automate the CO2 emissions calculation process, which Sorigué had previously performed manually. Beawre's platform, using Artificial Intelligence techniques, accurately classified 98% of the materials (such as cement, mortar, and others) present in a total of 172 invoices analyzed during the first test.
The system proved its ability to accurately calculate the weight of different materials. The complexity of this task is evident in the variety of criteria used for their classification, such as: calculation based on the weight contained in specific packaging (e.g., concrete bags); calculation derived from the physical dimensions of the material at the construction site (e.g., the area in m2 of a pavement like Stonegranit); calculation according to specific dimensions (such as the weight of concrete blocks); calculation based on unit weight (exemplified by bricks). Beawre Automate has also proven effective in discerning within invoices those elements that are not materials, such as services or tools, which do not require emission calculations. Technically, the approach implemented using Generative Artificial Intelligence is flexible. It can be adapted (with prompt engineering or even fine-tuning of the AI model) to produce the same excellent results for any type of invoices.
This project has also received support from ACCIO through grant ACE073/22/000108, contributing to its successful development.
With Beawre's AI-powered technology, our enterprise Constraula considerably improves the carbon footprint calculation process: manual calculations have been reduced by more than 90% and our processes can now be monitored in real time. The precision of the method implies reliable calculation results that will be reflected in emission reduction plans in line with reality. This pilot test with this technology is part of our environmental commitment and our permanent desire to improve as a company that forms part of the Sorigué group.
Francisco Moral
Deputy director of Quality, Prevention and Environment at Sorigué
RETURN OF INVESTMENT
Time Efficiency: Reduced time spent on manual calculations by over 90%.
Accurate Carbon Calculations: Increased accuracy in carbon footprint measurements by 50%.
Real-time Monitoring: Enabled dynamic monitoring and adjustments throughout project phases.
Standardization: Achieved greater consistency in data interpretation, regardless of the provider.
Get in Touch
C/ Esglesia 8, 1r, 08980
Sant Feliu de Llobregat, Spain
+34 93 688 61 10