H, the artificial intelligence startup that secured $220 million, presents its first product: Runner H for 'agent' applications.
The Parisian startup H, founded by former Google employees, surprised everyone last summer by unexpectedly revealing that it had secured an initial funding round of 220 million dollars.
The startup H, founded by former Google employees, garnered attention last summer by announcing a $220 million seed funding round without having launched a product yet. However, three months later, the situation became complicated when three of its five co-founders left the company due to "operational and business disagreements." Despite these setbacks, the company has persevered and today presents its first product: Runner H, an "agency" artificial intelligence designed for businesses and developers, which encompasses tasks such as quality assurance and process automation. This product is based on the startup's proprietary language model, which utilizes 2 billion parameters.
H has opened a waitlist for Runner H on its website. According to CEO Charles Kantor, in the coming days, APIs will begin to be released for those on the list, allowing the use of pre-built agents by H and offering options for developers to create their own agents. Initially, the use of these APIs will be free, although a paid model may be considered in the future. Building and operating AI, even with compact models, is not economical, particularly in an increasingly competitive environment.
Kantor also reports that H is in the process of raising a Series A funding round, aiming to establish what he describes as the second era of artificial intelligence, in contrast to the first era dominated by companies like OpenAI. H's previous funding, totaling $230 million, was a combination of equity and convertible debt, attracting notable investors such as Eric Schmidt, Yuri Milner, and Xavier Niel, as well as strategic investors like Amazon, Samsung, and UiPath.
The company has been discreetly working with a small number of clients in sectors such as e-commerce, banking, insurance, and outsourcing, using feedback from these clients to refine the product. Runner H will initially focus on three specific use cases: robotic process automation (RPA), quality assurance, and business process outsourcing (BPO). Robotic process automation has been used for years to automate repetitive tasks through basic scripts; however, Runner H aims to overcome limitations even when forms and sites have been modified, broadening the range of sources it can cover.
Regarding quality assurance, Kantor mentioned that one of the most popular approaches is reducing maintenance burdens in web testing, ensuring page availability and simulating real user actions. BPO, on the other hand, will not only improve billing processes but also optimize how agents access data from different sources.
Unlike other AI companies that focus their development on models with billions of parameters, Runner H opts for a more limited approach with only 2 billion parameters, which, according to Kantor, results in greater efficiency in both costs and operations. Additionally, the company claims that its compact models outperform others like Anthropic's "Computer Use" and models from Mistral and Meta.
