Senaatti

The document management system used at Senate Properties was in need of an update. With Zefort, the work of seven people-years was completed in just 30 calendar days and the quality of the contracts improved significantly.

Senate Properties provides facility services to public administration clients and manages real estate property worth 4.3 billion euros in total, consisting of over 9,000 buildings.

The company had a long tradition of scanning contracts to an electronic system, but along the years, IT systems had been replaced several times and metadata collection was not up to speed. The lack of metadata made it harder to find contracts and, at worst, led to sieving through a hefty paper archive.

”Due to Senate’s line of business, the contract base includes items that originate from the 1960s and are yet still valid. Due to deficient metadata, finding the relevant contract has sometimes required an unreasonable amount of time.”– Jussi Hirvelä, CIO at Senate Properties

AI processed 75,000 files

Senate Properties wanted to improve the discoverability of its electronic contract documents. They had around 4,000 currently valid contracts, but especially the oldest documents were hard to find as their metadata was often insufficient.

Since it would have been a massive manual workload to go through the documents that were missing metadata – approximately seven person-years, in fact – Senate decided to save both time and money by employing artificial intelligence.

The project was realized with Zefort’s Contract Takeover model. The model involves a project-type workflow in which an extensive contract mass is brought up to date without investments in traditional IT systems. The project was implemented in two phases.

At the first phase, Zefort taught its own DocAI engine by using a limited sample of 200 different types of contracts. The purpose of this phase was to teach the AI to identify the desired metadata from a large contract pool.

During the second phase, roughly 75,000 documents with deficient metadata were submitted to the AI engine, which then gathered all the required metadata, such as contract type, parties, heading and target of contract, from the mass. While doing so, the Zefort system was also able to identify text from PDF files and scanned TIFF documents.

High-quality data facilitates contract management

As an outcome from the project, Senate Properties received a list of its erroneously saved contracts. Once the metadata of these contracts has been fixed, the quality of contract data will improve and everyday contract management will become a lot easier.

”You hear people talking about AI, but not many have been able to prove its tangible benefits. This project with Zefort was a perfect example of the significant benefits AI has to offer when used wisely.”  – Jussi Hirvelä, CIO at Senaatti Properties