The Problem \/ Challenge<\/h2>\n\n\n\n
Given the size of our client’s property portfolio, the main pain point was understanding the costs for each property within the portfolio. Accurate cost tracking<\/strong> is important as each property generates monthly water, electricity and gas bills from different suppliers. The costs of the entire property portfolio form the basis for key financial metrics<\/strong> to understand performance within REIT and calculate returns.<\/p>\n\n\n\n Processing billing information and correct payments generate a huge amount of manual work. It is particularly important that they process all invoices timely and correct. A single mistake can have far-reaching consequences, such as the disconnection of a property from the grid or miscalculation of property portfolio performance. <\/p>\n\n\n\n In a competitive and fast-growing environment, the client wanted to automate the processing<\/strong> of utility bills as much as possible. This way, all bills would be processed on time and within budget, while minimising errors.<\/p>\n\n\n\n In previous attempts, the client experimented with Document AI and up training, without satisfactory results. Therefore, the S&P500 client contacted us.<\/p>\n\n\n\n Devoteam G Cloud has a lot of experience building solutions based on the Document AI <\/strong>suite. To speed up development processes, we have created our own accelerator <\/strong>that automatically batches calls and optimises resource usage. The resulting architecture is depicted below:<\/p>\n\n\n\n It is battle-tested on a daily basis across several of our customers and is able to process hundreds of millions of documents per month.<\/p>\n\n\n\n For this project, we needed to accommodate a typical Document AI workflow in order to keep manual verification inside of the processes through Human-in-the-Loop (HITL). Therefore, our custom machine learning solution uses documents labelled through HITL as training data and produces predictions that are compatible with the Document AI format. This requirement, in addition to bringing all of Document AI\u2019s features available, makes us able to switch to any other Document AI processor if ever it eventually reaches a better performance than our custom models.<\/p>\n\n\n\nThe Solution<\/h2>\n\n\n\n
![\"\"](\"https:\/\/devoteam.info\/wp-content\/uploads\/2024\/09\/Generic-Document-AI-pipeline.png\")
<\/figure>\n\n\n\n