Parseur initially experimented with a high-profile generative AI solution to augment the functionality of its platform. However, this solution was limited in the volume of data it could process each time it was used. This forced users to restrict data extraction to two- or three-page documents or undertake the cumbersome process of breaking large documents down into digestible chunks. For Parseur, this limited the available use cases of its platform and restricted its revenue potential. 

With Gemini 1.5 Flash featuring a 1 million token context window, Parseur can now support use cases such as data extraction from annual bank statements running into thousands of pages, enabling clients to properly leverage their accounting software to deliver the insights and reporting required to support decision-making.

Parseur remains a small business with four team members managing a platform that supports more than 50,000 users processing over 2 million documents a month, making cost management a key priority. With Compute Engine delivering the virtual machines supporting its applications, databases, and staging, and Cloud Storage storing documents and large binary data from users, the organization is complementing its use of Document AI and Gemini 1.5 Flash with a scalable, cost-efficient, and reliable cloud infrastructure. The business plans to move to C3D range servers to achieve a 30% increase in performance at an equivalent core count.