Processing purchase invoices is often labour-intensive in the construction industry. EG Jydacom is currently developing an AI-based service that automates invoice allocation. The aim is to reduce manual work, speed up invoice circulation and reduce the number of errors.
EG Jydacom has been a pioneer in ERP solutions for the construction industry in Finland for more than 45 years. We are now bringing new services to our customers that make use of the opportunities offered by AI. The first service to move into the pilot phase is designed to simplify one of the most labour-intensive stages of financial administration in the construction industry: purchase invoice processing.
A major pain point in financial administration
“According to the feedback we have received from customers, processing purchase invoices and allocating them correctly to projects is the biggest pain point in financial administration,” says Hanna Harra, Product Manager at EG Jydacom.
“Invoices are often routed to an invoice processor already at the first stage. The processor then has to manually extract the correct information from the invoice before sending it for verification and circulation. Invoice processing is made more difficult by the fact that suppliers in the construction industry use a wide range of different systems and posting practices, which means that the information may appear in random fields when received. In addition, a single invoice may include costs that need to be allocated to different construction sites. Manual data entry is slow and prone to errors, and invoice processing can easily become congested,” Harra continues.
EG Jydacom aims to use AI development to streamline purchase invoice processing by removing the manual stage altogether.
“We are currently developing purchase invoice automation, where AI automatically analyses purchase invoices. In the current model, the invoice is received from the e-invoicing operator into JIIRI and then sent to the invoice processor, who is typically the accounts payable clerk. Once purchase invoice automation has been implemented, this stage will be removed completely. The invoice will go directly from the operator to the analysis round and from there to the invoice reviewers for checking — not processing,” Harra explains.
How does purchase invoice automation work?
Purchase invoice automation uses machine learning and generative AI. The automation analyses the content of the purchase invoice and compares it with, for example, previous construction site, supplier and contract data. Based on this, it automatically suggests the correct construction site, transaction type and accounting account for the invoice lines — even when a single invoice includes costs for several different construction sites.
Purchase invoice automation reduces the workload for accounts payable and invoice processors, while also saving time for field workers, i.e. invoice reviewers. According to the Product Manager’s estimate, manual work can be reduced by up to 70% and the number of errors by 60%. At the same time, invoice circulation becomes three times faster. The pilot projects will measure the results through practical work, so that the achieved benefits can be verified.
“The strength of EG Jydacom’s purchase invoice automation lies in the fact that we have comprehensive access to all customer purchasing data from construction sites and supplier history. In addition, AI uses a continuous learning model, meaning that it learns from posting practices and improves its accuracy as it is used,” says Hanna Harra.
Customers involved in building the service
EG Jydacom’s invoice processing logic has now been renewed and serves as the foundation for purchase invoice automation. The service is being piloted with four customers in a production environment during the summer and early autumn. Once the learnings from the pilot project have been gathered and the further development needs that have emerged have been addressed, the service will be made available to all EG Jydacom JIIRI customers in October 2026. Purchase invoice automation does not require a separate integration, but is available directly in the JIIRI system.
“We first want to use the pilot project to ensure that the analysis works as expected and that customers find the functionality easy to use and practical,” Harra emphasises.
Four EG Jydacom customers were selected for the pilot, which began in spring 2026.
“There would have been plenty of customers interested in joining the pilot project. There is clearly demand for purchase invoice automation,” says Hanna Harra.
The project status, achieved results and feedback are reviewed with the customers every two weeks.
“User experience is important to us. In addition to the accuracy of the AI prediction, we collect feedback on how well the process and user interface serve the customer’s needs.”
EG Jydacom’s AI development continues
Purchase invoice automation is only the first step in Jydacom’s AI development. Next on the product development roadmap are a procurement service that complements purchase invoice automation, an assistant that suggests tasks based on user roles, and a tool that supports improved forecasting. These AI-powered services will be part of the JIIRI Työmaat service package, which will be released in 2027.