Maisa AI, a Valencia- and San Francisco-based startup building accountable AI agents for enterprises, has raised $25 million in seed funding led by Creandum, with participation from Forgepoint Capital International (via Banco Santander JV), NFX, and Village Global.
Founded in 2024 by David Villalón and Manuel Romero, Maisa AI addresses a major pain point: according to MIT’s NANDA initiative, 95% of generative AI pilots in enterprises fail due to hallucinations, lack of auditability, and opaque outputs.
A new framework for reliable automation
Maisa AI’s flagship product, Maisa Studio, is a model-agnostic self-serve platform for deploying digital workers trained in natural language. Instead of simply generating answers, its system builds a “chain-of-work” process to ensure transparency and accountability.
Key innovations include:
- HALP (Human-Augmented LLM Processing): where digital workers explain each step they’ll follow, letting users supervise workflows.
- Knowledge Processing Unit (KPU): a deterministic system that reduces hallucinations and enforces reliability.
“We’re not just chasing vibe coding. We’re building accountable AI agents that enterprises can trust for critical tasks,” said Villalón, CEO.
Advertisement
Early traction and enterprise focus
Maisa already counts a major bank, an automaker, and energy companies among its first production customers, positioning itself as a next-gen alternative to rigid robotic process automation (RPA). Customers can deploy via Maisa’s secure cloud or on-premises for regulated sectors.
The company currently employs 35 staff and plans to double headcount by Q1 2026 as it serves a growing waiting list of enterprise clients.
“Quick-start AI frameworks often collapse when reliability or auditability is required,” Villalón noted. “We are showing the market that accountable AI for enterprises is not just possible — it’s working.”
