RPT, part of Hubexo, specializes in information services for the real estate and construction sectors. They are Finland's leading processor of real estate and construction industry data, providing companies in the sector with tools for data-driven management.

RPT sought to transform their vast collection of unstructured data into structured data products. We leveraged AI and large language models (LLMs) to convert this unstructured data into structured formats, creating relevant and useful classifications for RPT's clients.

‍

The need for which a solution was sought

RPT had developed a new service for their portfolio, aiming to provide their customers with information about renovations in housing companies across Finland. They had already collected a substantial amount of data in free text format, which needed to be structured and classified.

Another dataset that required organizing included details about housing managers, which needed to be structured into company and contact information formats.

Previously, RPT had sought to use traditional machine learning classifiers for structuring the data, but the results were inadequate and unsatisfactory.

‍

Our solution

The project began with processing the collected data using OpenAI's APIs, experimenting with different workflows, and refining prompts. Leveraging Generative AI quickly proved highly effective for separating the data into desired sections, achieving excellent quality with limited effort.

The data was structured in stages, progressing from high-level categories to more detailed classifications. A key advantage of this approach is that while AI handles the structuring, business operations retain full control over data sets and classifications, allowing for adjustments without involving developers or AI specialists.

To safeguard RPT's business data from potential changes in OpenAI services, such as pricing or availability, we also developed a local model tailored to produce similar results without relying on external services. This model serves as a backup, ready to be activated in production if necessary.

‍

Cooperation

RPT played a key role in validating the results and providing valuable feedback as we fine-tuned the workflow and models. Their involvement highlights the importance of customer collaboration in ensuring that the final outcome aligns as closely as possible with their specific needs and expectations.

“This project was of great importance to us, as we had sought a solution for categorizing the vast amount of data we had gathered already before. Therefore, it was great to see how fluent the advancement of the project was, and how quickly Fluentia was able to tackle the different obstacles in the way. We’re really pleased with the end result – the quality and cost efficiency. This new solution enables us to harness cutting-edge AI technology in a cost-effective and strategic manner for a service crucial for our business”, says Kimmo Karhemo, Data Lead, NE Europe, Hubexo.

‍

‍

RPT Byggfakta Oy

RPT Byggfakta Oy specializes in information services for the real estate and construction sectors. Established in 1975, it has become Finland's leading provider of construction-related data, offering tools that facilitate informed decision-making within the industry. By using RPT’s information services, the building materials industry, contractors, designers, and other companies offering real estate and construction services can reach the decision-makers of construction projects at precisely the right time.

RPT Byggfakta Oy is part of Hubexo, formerly known as Byggfakta Group, which operates in over 20 countries and employs approximately 2,500 people. Hubexo provides top-tier data, insights, and software solutions globally to the construction industry.