Maisa AI Raises $25M to Tackle AI Failure in Enterprises

The High Failure Rate of Generative AI and a New Approach

Recent findings from MIT’s NANDA initiative indicate a concerning trend: a substantial 95% of generative AI pilot programs within companies are not achieving success. However, leading organizations aren’t abandoning the technology; instead, they are focusing on experimenting with agentic AI systems capable of learning and being effectively supervised.

Introducing Maisa AI and Maisa Studio

Maisa AI is a year-old startup addressing this challenge. The company’s core philosophy centers on the necessity of accountable AI agents for enterprise automation, moving away from the ambiguity of “black box” systems. With a newly secured $25 million seed round, spearheaded by European VC firm Creandum, Maisa has launched Maisa Studio.

This platform is designed to be model-agnostic, offering a self-service environment for deploying digital workers that can be trained using natural language.

Differentiating from Existing Platforms

While similar to platforms like Cursor and Lovable, Maisa distinguishes itself through a fundamentally different approach. According to Maisa CEO David Villalón, the company focuses on building the process required to achieve a desired outcome, rather than directly generating responses. This is referred to as a “chain-of-work.”

The Origins of Maisa: Addressing AI Hallucinations

The foundation of this process lies with Maisa’s co-founder and chief scientific officer, Manuel Romero, who previously collaborated with Villalón at Clibrain, a Spanish AI startup. In 2024, the pair recognized the critical need for a solution to AI hallucinations, realizing the unreliability of AI outputs.

They acknowledge the potential of AI but believe that manually reviewing vast amounts of AI-generated work is impractical.

HALP: Human-Augmented LLM Processing

To overcome this, Maisa developed a system called HALP (human-augmented LLM processing). This method facilitates collaboration between users and digital workers, with the workers outlining each step of the process while users provide their requirements – akin to a collaborative problem-solving session.

The Knowledge Processing Unit (KPU) and Enterprise Adoption

Maisa also created the Knowledge Processing Unit (KPU), a deterministic system specifically engineered to minimize hallucinations. Initially focused on this technical hurdle, Maisa discovered that its emphasis on trustworthiness and accountability resonated strongly with companies seeking to apply AI to crucial operations.

Current Maisa clients include a major bank, alongside organizations in the automotive and energy industries.

Maisa as Advanced Robotic Process Automation

Maisa aims to evolve beyond traditional robotic process automation (RPA), delivering productivity improvements without the constraints of rigid rules or extensive manual coding. The startup offers flexible deployment options, including secure cloud hosting and on-premise installations.

Strategic Growth and Customer Focus

While Maisa’s customer base remains relatively small compared to freemium platforms, the company is strategically positioned to capture enterprise clients. Maisa Studio is designed to expand its customer reach and simplify the adoption process.

The company also intends to support existing customers with operations across multiple countries, leveraging its dual headquarters in Valencia and San Francisco.

Investment and Future Expansion

A recent funding round included participation from U.S. firm Forgepoint Capital International, through a joint venture with Banco Santander, demonstrating its appeal within regulated industries. The startup’s pre-seed round was led by NFX and Village Global.

Differentiating in a Competitive Landscape

Maisa’s focus on complex use cases requiring accountability from non-technical users sets it apart from competitors like CrewAI and other AI-powered workflow automation tools. CEO Villalón cautions against the pitfalls of prioritizing speed over reliability and auditability in the current “AI framework gold rush.”

Scaling for Demand

To facilitate the scaling of AI solutions, Maisa plans to expand its team from 35 to 65 employees by the first quarter of 2026. The company anticipates significant growth starting in the final quarter of this year as it addresses its existing waitlist, aiming to demonstrate the practical delivery of promised AI capabilities.