Now that we have demonstrated in this project that AI can be utilized safely and successfully, we can be more open to what else we can achieve with it.
GF Building Flow Solutions is a leading global provider of sustainable and innovative solutions, making water flow in buildings. The company was established when the international industrial corporation Georg Fischer acquired the Finnish company Uponor in 2023. With a rich history and a strong presence in both Europe and America, GF Building Flow Solutions' products are widely used in construction projects worldwide. The acquisition has further internationalized and expanded the business, as evidenced by the increase in sales orders from various parts of the globe.
"In Europe alone, we receive thousands of sales orders monthly", says Mika Haapala, Global Business Process Owner (Order-to-Cash) at GF Building Flow Solutions. As the volume of sales orders increased, so did the need to automate the process of recording them into the company's Oracle-based ERP system. "While some sales orders arrive through an electronic ordering system, many still come in PDF format, especially from certain countries and markets. These orders must be manually entered into the system line by line, which is time-consuming for our order processors – some orders can be as long as 26 pages, filled with numerous order lines", Haapala describes, noting that the number of manually processed orders has risen due to the organizational changes. As volumes and manual work grew, GF Building Flow Solutions realized their existing personnel resources were insufficient to meet the demand and began exploring automation possibilities to streamline the process.
AI as a solution for automation scalability
GF Building Flow Solutions had previous experiences with utilizing robotic process automation, but it did not prove to be a scalable solution for this need. Sales orders arrive in diverse formats from around the world, and creating rules for each unique order for the software robot would have been time-consuming and inefficient. The scalability issue was resolved with Efima's proposed solution, which enables the automatic reading of PDF-format sales orders into the ERP system using AI. This solution is based on an architecture model developed by Efima, utilizing Microsoft's Azure-based AI resources. With advanced text recognition, information from various PDF orders can be automatically extracted, converted into a structured format, and transferred to the ERP system as pre-prepared orders. "I still vividly remember the first demo, where the AI sucked all the information from the PDF into the system in one go", Haapala smiles.
The company's first AI project also commenced swiftly in the fall of 2024. "It took a total of two months before we were already in production in the first country", recalls Finance Development Manager Tiina Tuuri, who served as the project manager for the initiative. Due to the high pressure to streamline the process, a tight schedule was set for the pilot project, which was adhered to. Germany served as the pilot country, and soon after, the solution was successfully implemented in other European countries. In the near future, the goal is to bring all European countries under the AI implementation and then expand its use to North America. "A functional basic process has been created for the implementation, which has made the rollout efficient and controlled. Additionally, customer-specific characteristics are taken into account in each country to ensure the solution's functionality," Tuuri explains.
Implementing new technology in a critical business process required trust
Tuuri and Haapala are grateful for having an experienced partner in Efima for the company's first AI implementation, whose collaboration instilled trust. "From the very first meeting, Efima understood our schedule pressures and the criticality of the project, and they proactively ensured that everything was in order on our end. We never had to look over Efima's shoulder", Haapala says.
GF Building Flow Solutions also valued the transparency Efima provided regarding the possibilities and limitations of AI. "There was no traditional sales pitch of 'everything is possible.' Efima was honest from the start about what can be automated with AI and what cannot, and what the potential pitfalls are. This was very important so that we, as the customer, did not end up disappointed by empty promises during a project that was important to us", Tuuri emphasizes. "It's better to be cautious with AI implementation. Fortunately, we were able to trust Efima throughout the project and that they knew what they were doing."
Implementing new and unfamiliar technology also required significant change management. Eventually, every project member witnessed the benefits and functionality of the solution in practice. The international project team included Efima, Haapala, and Tuuri, as well as GF's developers, order processors, IT specialists, legal experts, and Oracle experts. "This has been, above all, a joint project that succeeded with good teamwork", Tuuri summarizes, thanking the entire project team for their commitment to the common goal.
Successful AI implementation has already sparked new ideas
With the AI solution, the amount of manual work has significantly decreased, and feedback from the field has been positive. "Previously, an order processor might have received dozens of long PDF orders weekly. Now that they can be automatically entered into the system and the order processors only need to check the orders, the time savings are significant", Haapala describes.
The successful AI project has generated a lot of interest across the organization, and the company is considering several possibilities for utilizing AI and similar technology to read necessary information from unstructured data. "Now that we have demonstrated in this project that AI can be utilized safely and successfully, we can be more open to what else we can achieve with it", Haapala says. "The first pilot required a lot of groundwork, which now rewards us with agile expansion of the solution and a lower threshold for applying AI in other use cases."
The duo advises other companies planning their first AI experiments to invest in groundwork first. "When AI compliance and security risks have been assessed and compatibility with the existing system architecture has been ensured, it is much easier to embark on AI experiments. There are so many possibilities related to utilizing AI that it would be good if AI experiments were not hindered by basic concerns", they summarize.
Could we help you too?
You can contact us via the form below.