When Twin emerged from stealth in January 2024, the idea of AI agents was still largely theoretical. Fast forward to today, and the Paris-based company has introduced its first AI-powered automation agent in collaboration with Qonto, the fintech startup that provides business bank accounts to over 500,000 customers across Europe.
While there are numerous ways to automate repetitive tasks, such as API-based tools like Zapier or RPA software like UiPath, Twin believes there’s a more efficient method for handling automation. With a team of just nine, the company is leveraging artificial intelligence and computer-using models to create cutting-edge solutions.
The Invoice Operator, Twin’s debut product designed for Qonto, serves as an excellent demonstration of AI’s potential in automation. Qonto processes millions of invoices each month, and customers spend significant time gathering and uploading these invoices to their accounts. In just three months, Twin developed an automatic invoice-retrieval tool that drastically accelerates this process.
“For Qonto, this solution tackles a long tail of services. At scale, this would be unmanageable with RPA because it would require creating a custom script for every website,” explained Twin co-founder and CEO, Hugo Mercier, during a product demo.
While competitors like Zapier have taken years to support thousands of applications, Twin’s Invoice Operator is already supporting a wide range of services within just a few months.
Twin uses a Chromium-based web browser running on a server and OpenAI’s CUA (computer-using agent) model for this task. In fact, Twin was one of the 15 companies chosen to beta-test CUA, which also powers OpenAI’s Operator product.
The goal is to provide a seamless user experience, even for those who may not be tech-savvy. Users simply log into their accounts and let the agent do the work, without needing to configure or prompt the system.
Looking ahead, Twin sees great potential for AI agents across various industries. Future applications could include managing orders for e-commerce businesses, classifying product catalogs for marketplaces, or assisting call center agents by automating routine tasks.
Twin is positioning itself for a future where AI agents become increasingly affordable, efficient, and accurate, with hopes of expanding its platform to allow developers to integrate these capabilities into their own applications.
Also Read : OpenAI Adopts Anthropic’s MCP Standard to Enhance AI Connectivity
