France\’s H Company unveils 3 groundbreaking agentic models, aiming to become the heartbeat of AI
H Company Highlights AI Evolution in Euronews Next Interview
Interview Overview
The CEO and co-founder of H Company engaged in an in-depth conversation with Euronews Next, shedding light on the organization’s latest AI breakthroughs and its strategic outlook.
Key Themes Discussed
- Cutting‑Edge Models: A showcase of new AI models that blend advanced machine learning with real‑world practicality.
- Human Oversight: Emphasis on the role humans play in guiding and supervising AI decisions to safeguard ethics and accountability.
- Future Roadmap: Insight into forthcoming developments, partnerships, and the broader impact of AI on society.
Takeaway Message
With human judgment at the helm, H Company envisions AI not as a replacement but as a powerful, responsible ally that augments human capabilities across industries.
French AI Start‑Up H Company Launches Three New Autonomous Agents
Paris, France – H Company, a young artificial intelligence venture founded last year, has unveiled a trio of cutting‑edge AI agents on Tuesday. The startup, which secured several hundred million euros in funding before launching its first product, has positioned itself firmly in the race toward agentic AI.
Company Vision and Human-Centric Approach
Charles Kantor, CEO and co‑founder, emphasized that all human users remain “at the centre” of the technology. “Even if the agent drafts an email, only a person can send it,” he said, underscoring the company’s commitment to safe and responsible deployment.
Three Agent Models
- Runner H – Designed for rapid decision‑making and task execution.
- Tester H – Focuses on validation and quality assurance of solutions.
- Surfer H – Optimised for navigating complex information streams.
H Company claims that these agents represent a “trusted, action‑oriented AI” capable of performing tasks beyond the scope of conventional chatbots.
Performance Highlights
According to the firm, the new agents achieve a 92.2 % success rate while cutting operating costs by up to 5.5 times compared to competitors such as OpenAI, Anthropic, and Google.
Future Outlook
As AI continues to evolve, H Company plans to refine these models further, aiming to maintain a leadership position in the agentic AI space.
What are the three models?
Introducing the Next Generation of AI Assistants
Runner H is designed to empower users by automating routine workflows and enhancing productivity. According to Kantor, it represents a sophisticated evolution of the earlier Manus platform, specifically tailored for consumer applications.
Surfer H: Web‑Minded Intelligence
- Enables seamless browsing and interaction within web environments.
- Attains a 92.2% completion rate on the WebVoyager benchmark.
- Charges only $0.13 (€0.11) per task, making it cost‑effective compared to rivals like OpenAI’s Operator agent at $200 (€175) monthly.
Kantor described Surfer H as “an almost autonomous agent” capable of planning, visual perception, interface navigation, and data retrieval—all without the need for manual clicks.
Tester H: Enterprise‑Focused Testing
Designed primarily for corporate use, Tester H excels in software testing but also offers utility in generating smart email responses. The system can draft replies based on recent correspondence, yet the final send action remains controlled by the user.
Core Vision: Trusted, Action‑Oriented AI
All three models embody Photon´s commitment to “a trusted, action‑oriented AI that delivers task execution beyond traditional chatbots.” By leveraging synthetic data—artificially crafted datasets that emulate real scenarios—the company adheres to Europe’s GDPR regulations.
Personal Applications by Kantor
He reports using the AI suite to integrate tools such as email and document systems for billing calculations, drafting communications, and producing internal content. Reviewing the agent’s output reveals significant productivity gains.
The ‘clear’ agentic vision
H Company’s Launch and Early Challenges
When H Company first entered the market last year, it quickly became a focal point in the tech community. Its founding team was a blend of academic and industry talent: campus professor Robert Kantor of Stanford, joined by colleagues from the renowned research lab DeepMind.
High-Profile Investment
The start‑up attracted a diversified portfolio of investors, notably including:
- Bernard Arnault – CEO of LVMH
- Xavier Niel – Founder of Iliad
- Amazon, Samsung, and other corporate giants
- Former Google chief Eric Schmidt
Key Leadership Departures
Only a few months after its debut, three co‑founders — Daan Wierstra, Karl Tuyls, and Julien Perolat — stepped down citing “operational differences.” The company confirmed the exit on LinkedIn and outlined a vision moving forward.
CEO’s Vision and Strategic Alignment
Executive Director Robert Kantor reflected on the departures:
“The recent departures are a thing of the past. We had strategic misalignments, but now those individuals are ambassadors for the company. H’s vision is crystal clear.”
“Our goal is to lead the frontier of action‑modeling and agentic AI, ensuring the technology serves humanity.”
Organizational Structure
H Company currently employs roughly 70 people, split between hubs in Paris and the United Kingdom. Its core strength lies in a robust research and engineering department.
Symbiosis of Core Functions
Kantor underscored the importance of a unified effort:
“We need to orchestrate product, research, and engineering in perfect harmony to build AI that defines a new category.”
Agentic AI – The Next Frontier
According to Kantor, the evolution toward agentic AI is pivotal for real‑world applications. He highlighted recent discussions at NVIDIA about physical AI and outlined the following prospects:
- Robotics — especially in France, where partner companies are investing heavily.
- Computer games and simulation environments.
- Advanced AI agents that drive autonomous systems.
“Software foundations for robotics will increasingly rely on agentics and superintelligence. The opportunities are vast,” Kantor explained.
