Twin’s Debut AI Agent Assists Qonto Customers with Invoice Retrieval

In January 2024, Twin emerged from stealth at a time when AI agents were more theoretical than practical. Now, the Paris-based company is launching an automation agent in collaboration with Qonto, a fintech startup serving over 500,000 European businesses with banking solutions.

For automating repetitive tasks, existing solutions include API-based no-code and low-code platforms like Zapier or RPA software such as UiPath. However, Twin, with its nine-person team, believes AI-driven automation is a more efficient approach.

Twin’s first product, Invoice Operator, demonstrates AI’s potential by streamlining invoice retrieval for Qonto users. Managing millions of invoices monthly, Qonto users often spend hours collecting and uploading them. Over the past three months, Twin has developed a tool to accelerate this process.

Invoice Operator Automates Invoice Retrieval with Seamless Integration

Invoice Operator first identifies transactions with missing invoices and displays the required services alongside a browser window showing the agent’s actions. If login credentials are needed, the browser pauses for manual input before continuing. Once authenticated, the agent retrieves past transactions, downloads invoices, and attaches the PDFs directly to Qonto accounts, simplifying invoice management.

“When operating at Qonto’s scale, you must account for an immense range of services—eventually reaching hundreds of thousands,” explained Twin co-founder and CEO Hugo Mercier during a product demo.

He emphasized that RPA isn’t a viable solution since it requires creating a custom script for each website and updating it whenever the site changes.

Twin’s Invoice Operator Achieves Rapid Integration, Surpassing Zapier’s Pace

Regarding API-based automation like Zapier, Mercier noted that it took Zapier a decade to integrate 8,000 applications, whereas Twin’s Invoice Operator has already supported thousands within just a few months.

Twin’s system runs a Chromium-based web browser on a server and leverages OpenAI’s CUA (computer-using agent) model. The company was among 15 selected to beta-test CUA, the same model powering OpenAI’s Operator, which enables users to prompt an AI agent to complete tasks.

Twin prioritizes ease of use, aiming to make AI agents accessible to non-technical users. “We focused on simplifying the experience—users don’t need to configure anything. Just log in, launch it, and the agent finds invoices,” Mercier said.

Beyond invoice retrieval, Twin sees potential for AI agents in B2B applications, such as automating e-commerce order management, organizing marketplace catalogs, or assisting call centers. However, the company envisions AI agents becoming faster, more affordable, and highly versatile. The next challenge is turning the core platform behind Invoice Operator into a tool that developers can integrate into their own applications.

Read the original article on: TechCrunch

Read more: Spain Moves To Outlaw AI-generated Sexual Images.