Maisa AI gets $25M to fix enterprise AI’s 95% failure rate

A stunning 95% of generative AI pilots at companies failureAccording to a recent report Published by Mit’s Nanda Initiative. But instead of completely giving up the technology, the most advanced organizations with agent AI systems that can learn and are guided.

That’s true Maisa Ai Comes in. The year old Startup has built up its full approach around the starting point that Enterprise automation responsible for responsible AI agents, not opaque black boxes. With a new $ 25 million seed round under the leadership of the European VC company Creandum, it has now launched it Maisa StudioA model-agent self-service platform that helps users to implement digital employees who can be trained with natural language.

Although that may sound familiar, it is reminiscent of so-called atmospheric coding platforms such as Cursor and the Lovable with Creandum-Supported Maisa that his approach is fundamentally different. “Instead of using AI to build the answers, we use AI to build the process to be carried out to achieve the answer-what we call” work chain of work, “Maisa CEO David Villalón told WAN.

The most important architect behind this process is Maisa’s co-founder and head-scientific officer, Manuel Romero, who had previously worked with Villalón at the Spanish AI-Startup Clibrain. In 2024 the duo worked together to build a solution for hallucinations after he had seen first -hand that “you couldn’t trust AI,” Villalón said.

The couple is not skeptical about AI, but they think it will not be feasible for people to “judge three months of work in five minutes”. To tackle this, Maisa uses a system called Halp, standing for the processing of the Mens-Augmented LLM. This adjusted method works as students of the Blackboard – it asks users about their needs, while the digital employees sketch every step they will follow.

The startup also developed the knowledge processing unit (KPU), a deterministic system that is designed to limit hallucinations. Although Maisa started this technical challenge instead of a use case, it soon discovered that the bet on reliability and accountability resonated with companies that hope to apply to critical tasks. Customers who currently use corn in production include a large bank, as well as companies in car manufacturers and energy sectors.

By serving these business customers, Maisa hopes to position itself as a more advanced form of robot process automation (RPA) that unlocks productivity gain without having to rely on rigid pre -defined rules or extensive manual programming. To meet their needs, the startup also offers them either implementation in its secure cloud or via on-premise implementation.

This Enterprise-first approach means that the customer base of Maisa is still very small compared to the millions of current to Freemium-solid coding platforms. But because these platforms now investigate how they can win business customers, Maisa moves in the opposite direction with Maisa Studio, which is designed to make the customer’s funnel and it easier.

The startup is also planning to expand with existing customers who have activities in several countries. With a double headquarters in Valencia and San Francisco, Maisa himself already has a foothold in the US, as reflected in his dop table; are $ 5 million Pre-Seede Round Last December was led by Venture companies established in San Francisco NFX and Village Global.

In addition, WAN learned exclusively that the American company Forgepoint Capital International participated through its new round through his European joint venture With the Spanish Bank Banco Santander, who emphasizes his appeal to regulated sectors.

Focus on complex use cases that require accountability of non-technical users can be a distinctive factor for corn, whose competitors are crewai and many other AI-driven, business-oriented workflow automation products. In a LinkedIn Post, Villalón marked This “AI Framework Gold Rush”, warning that the “Quick Start” will be a long nightmare when you need reliability, auditability or the possibility to repair what went wrong. “

Maisa doubles on his goal to help AI scale, intend to use his financing to grow from 35 to 65 people by the first quarter of 2026 to meet demand. From the last quarter of this year, the startup expects rapid growth as its waiting list starts to serve. “We are going to show the market that there is a company that delivers what has been promised and that it works,” Villalón said.