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

A startling 95% of generative AI initiatives within enterprises are failing, as highlighted by a recent study from MIT’s NANDA initiative. In response to this alarming statistic, innovative firms are turning to agentic AI systems that combine learning with supervision. This is precisely where Maisa AI steps in. Founded just a year ago, the startup is centered on the belief that effective enterprise automation needs accountable AI agents rather than opaque systems. With a fresh $25 million seed funding round led by European venture capital firm Creandum, Maisa has introduced Maisa Studio, a versatile self-serve platform designed to empower users to deploy digital workers trained through natural language. While this concept may echo familiar platforms like Cursor and Lovable, Maisa distinguishes itself by asserting that its methodology is fundamentally unique. “We utilize AI not to generate responses but to construct the processes necessary to achieve those responses, which we refer to as ‘chain-of-work,’” shared CEO David Villalón with TechCrunch. The process is primarily crafted by co-founder and Chief Scientific Officer Manuel Romero, who previously collaborated with Villalón at the Spanish AI startup Clibrain. In 2024, they joined forces to tackle the issue of AI hallucinations after recognizing the impracticality of humans reviewing extensive work completed in mere minutes. To combat this, Maisa has developed HALP, or Human-Augmented LLM Processing, a system that engages users by identifying their needs while guiding digital workers through the steps they will execute. Additionally, the startup has introduced the Knowledge Processing Unit (KPU), a deterministic framework designed to minimize hallucinations. Although Maisa originated from a technical challenge rather than a specific use case, it quickly discovered that its commitment to trustworthiness and accountability resonated with companies eager to leverage AI for essential tasks. Current clients include large banks and organizations in the automotive and energy sectors. By catering to these enterprise clients, Maisa aims to establish itself as a more sophisticated variant of robotic process automation (RPA), facilitating productivity enhancements without necessitating rigid rules or extensive manual programming. The startup offers deployment options either through its secure cloud or on-premise, aligning with its enterprise-first approach. Despite its smaller customer base compared to the multitude flocking to freemium vibe-coding platforms, Maisa is strategically positioning itself as these platforms seek to attract enterprise customers. With Maisa Studio, the company aims to expand its customer funnel and simplify adoption. Moreover, it plans to grow alongside existing clients with international operations. Operating from dual headquarters in Valencia and San Francisco, Maisa has already established a presence in the U.S. market, as evidenced by its funding history; a $5 million pre-seed round last December was led by San Francisco-based firms NFX and Village Global. Notably, Forgepoint Capital International also participated in this latest funding via its European partnership with Banco Santander, underlining its appeal in regulated industries. By focusing on intricate use cases that require accountability from non-technical users, Maisa sets itself apart in a competitive landscape that includes CrewAI and other AI-driven workflow automation solutions. Villalón remarked on LinkedIn about the current “AI framework gold rush,” cautioning that what appears to be a quick start could devolve into a prolonged challenge when reliability and auditability become paramount. To accelerate its growth, Maisa intends to use its funding to expand its team from 35 to potentially 65 members by early 2026 to meet increasing demand. The startup anticipates a surge in growth beginning in the last quarter of this year as it starts serving its waiting list. “We are poised to prove to the market that we are fulfilling the promises made, and that we are successfully delivering results,” Villalón concluded.