Paris-based startup Twin has officially launched its first AI-powered product—an invoice retrieval agent—in partnership with European fintech giant Qonto. This marks a big leap forward in practical automation, showing how AI agents can handle tedious business tasks faster and smarter.
Back in January 2024, when Twin emerged from stealth mode, AI agents were mostly a futuristic concept. Fast-forward to today, and Twin is proving that automation tools powered by AI are not only real but ready for action. Their first tool, dubbed Invoice Operator, is designed to help Qonto users automatically retrieve and upload invoices directly into their business banking accounts—cutting hours off repetitive monthly admin work.
Why Twin Thinks AI Beats Traditional Automation Tools
Today, many businesses already use no-code automation platforms like Yavi or robotic process automation (RPA) tools like UiPath to streamline repetitive tasks. But Twin believes there’s a more scalable and intelligent way forward.
With a small but mighty team of just nine, Twin leverages advanced computer-using agent (CUA) models to offer smarter automation. Their new product shows how AI can go beyond scripted workflows and actually understand, navigate, and interact with web interfaces.
Qonto processes millions of invoices each month, and its users often spend hours manually collecting them. With Twin’s AI agent, that time-consuming task is drastically simplified.
Here’s How Twin’s Invoice Operator Works
Once a Qonto user launches the Invoice Operator, the system scans their account to detect any transactions that are missing invoices. It then pulls up a list of services—like SaaS platforms or online vendors—where those invoices need to be fetched.
On the screen, the agent opens a browser window showing its activity in real time. If the tool encounters a login page, it pauses and prompts the user to input their credentials securely. After that, the agent takes over—navigating through websites, locating past invoices, downloading them, and automatically attaching the PDFs to the corresponding Qonto transactions.
This entire process is seamless and doesn’t require any technical setup. “We wanted something that even non-technical users could run without any prompting or configuration,” said Hugo Mercier, Twin’s co-founder and CEO.
Thousands of Services Supported—And Growing Fast
What makes Twin’s approach impressive is its speed and scalability. Most RPA tools would need a custom script for each individual service—and every time a site’s interface changes, that script would need updates. That’s not sustainable at scale.
“Supporting tens of thousands of services with RPA just isn’t practical,” Mercier explained. “But our AI can adapt dynamically to different websites.”
In comparison, it took Zapier nearly a decade to integrate 8,000 apps. Twin’s platform already works with thousands—and it’s only been in development for a few months.
Built on OpenAI’s Cutting-Edge CUA Technology
Behind the scenes, Twin runs a Chromium-based browser on remote servers and uses OpenAI’s CUA model, the same technology that powers OpenAI’s Operator product. Twin was among the first 15 companies granted early access to CUA, allowing them to build an agent that can navigate the web, take action, and retrieve data—much like a human would.
By combining this AI muscle with a user-friendly interface, Twin is making agentic AI accessible to everyone—not just developers or tech-savvy users.
What’s Next: AI Agents for More B2B Workflows
While invoice retrieval is Twin’s first step, the startup envisions AI agents revolutionizing many more B2B workflows. Think order processing for e-commerce, automated data classification for marketplaces, or intelligent support for call centers.
The long-term vision? A world where AI agents are fast, affordable, and reliable enough to tackle a wide range of tasks—without complex setup or coding.
Twin isn’t just building a single tool. It’s creating a scalable agent platform that could power automation across industries. The big question now is whether developers will adopt Twin’s technology and embed it into their own products.
