When Twin got here out of stealth in January 2024, AI brokers have been extra of a theoretical idea than a actuality. Today, the Paris-based is releasing an automation agent in partnership with Qonto, the fintech startup that gives enterprise financial institution accounts to greater than 500,000 clients throughout Europe.
If you wish to automate repetitive duties, there are already a number of methods to sort out these issues. Some firms use API-based, no-code or low-code automation merchandise like Zapier. Others depend on RPA software program, comparable to UiPath.
With its crew of 9 individuals, Twin thinks there’s a way more environment friendly approach to deal with automation. As you might have guessed, it concerned synthetic intelligence and laptop use fashions.
Invoice Operator, Twin’s first product designed for Qonto, is an efficient instance of why it is sensible to make use of synthetic intelligence. Qonto handles hundreds of thousands of invoices per thirty days. But clients spend a number of hours per thirty days gathering invoices and importing them to Qonto.
Over the previous three months, Twin has created an automated bill retrieval device that may velocity up this course of. When customers launch Invoice Operator, Twin first fetches the record of transactions with lacking invoices. It then exhibits the record of companies that it must entry to obtain invoices subsequent to a browser window displaying the agent’s actions.
If you could log in to a service to obtain invoices, the browser pauses and asks you to enter your credentials manually. Once that is achieved, you may click on on a button to let the agent proceed its work.
After that, Twin’s Invoice Operator mechanically finds your record of previous transactions, downloads invoices and attaches the PDFs to the transactions in your Qonto account.
“When you do this on the scale of Qonto, you principally have to cowl a really, very lengthy tail of companies. Thousands, tens of hundreds and shortly a whole bunch of hundreds of various companies that everybody is utilizing,” Twin co-founder and CEO Hugo Mercier mentioned throughout a demo of the product.
“And that will be utterly not possible with RPA since you would wish to create a single customized script per web site after which each time the web site modifications, you would need to modify the script,” he added.
As for API-based automation merchandise like Zapier, Mercier mentioned that it took Zapier 10 years to assist 8,000 purposes on its platform. Twin already helps hundreds of purposes for its Invoice Operator simply a few months after beginning work on the product.
Behind the scenes, Twin runs a Chromium-based net browser on a server. The startup makes use of OpenAI’s CUA mannequin (CUA stands for Computer-Using Agent). In truth, Twin was one of many 15 firms that received to attempt CUA in beta.
CUA can also be the mannequin that powers OpenAI’s Operator, its prosumer product that permits you to enter a immediate to let an agent carry out an motion for you. In addition to higher efficiency, Twin believes it must be simpler to make use of brokers that browse the online for you.
“We labored so much on making the expertise very simple. We are actually focusing on the top consumer, perhaps people who find themselves not tech-friendly. They don’t need to immediate or configure something. You simply log into your accounts, launch it, and it navigates to seek out the invoices,” Mercier mentioned.
After bill retrieval, Twin thinks there are various industries that might profit from B2B agentic purposes. For instance, brokers may mechanically handle orders for an e-commerce firm, classify the catalog of a market or retrieve data for name heart brokers.
Twin is pitching a future during which AI brokers grow to be cheaper, quicker and extra correct throughout all kinds of duties. Now, let’s see if the startup can flip the core agent platform that powers Invoice Operator right into a product that builders can begin utilizing in their very own purposes.
