A staggering 95% of generative AI pilots in enterprises are failing, according to a recent report from MIT’s NANDA initiative. The numbers are grim, but rather than abandoning the promise of AI altogether, forward-thinking organizations are shifting their focus toward a new frontier: agentic AI systems — models that don’t just generate outputs, but can learn, reason, and be supervised like real employees.
This is precisely the space where Maisa AI is planting its flag. The year-old startup believes the future of enterprise automation doesn’t lie in opaque black boxes spitting out unpredictable answers, but in accountable AI agents that work in full transparency. Now, with a fresh $25 million seed round led by European VC firm Creandum, Maisa is officially launching Maisa Studio — a model-agnostic, self-serve platform that allows companies to deploy trainable digital workers powered by natural language instructions.
At first glance, Maisa Studio might sound like other AI “vibe coding” platforms — such as Cursor or Creandum’s other portfolio darling, Lovable. But Maisa insists its methodology is fundamentally different. “Instead of using AI to build the responses, we use AI to build the process that needs to be executed to get to the response — what we call ‘chain-of-work,’” co-founder and CEO David Villalón explained to TechCrunch.
The brains behind this approach is Maisa’s co-founder and Chief Scientific Officer, Manuel Romero, who had previously worked with Villalón at Spanish AI startup Clibrain. In 2024, after repeatedly witnessing AI’s shortcomings in real-world use cases, the two set out to design something radically different. “We saw firsthand that you could not rely on AI,” Villalón said. Their solution: build systems that put humans firmly in the loop while giving AI clear structures to follow.
To do that, Maisa developed HALP (Human-Augmented LLM Processing), a proprietary method designed to keep AI accountable. Think of it like a classroom exercise: users describe what they want, and the AI agents sketch out the steps they plan to take — just as a student might show their work on the blackboard before delivering the answer. The goal isn’t just accuracy, but auditability — ensuring companies know how and why decisions are made.
Maisa also introduced the Knowledge Processing Unit (KPU), a deterministic system designed to dramatically reduce hallucinations and keep AI outputs tethered to reality. While the company originally began with the technical challenge of trustworthiness rather than a specific business use case, its timing proved impeccable. Corporations desperate to apply AI to critical workflows — but wary of errors — quickly found Maisa’s approach compelling. Today, the startup already counts a major European bank and companies in the automotive and energy sectors among its production clients.
By taking this route, Maisa is carving out a position as an evolution of robotic process automation (RPA) — the software bots enterprises have long used to streamline repetitive work. Unlike legacy RPA systems, which rely on rigid rules and manual programming, Maisa’s AI agents adapt, learn, and explain themselves. Clients can choose between on-premise deployment for maximum security or Maisa’s secure cloud, depending on their compliance needs.
This enterprise-first focus has kept Maisa’s user base smaller than the millions flocking to consumer-friendly platforms like Cursor or Lovable. But the startup sees that as a strength, not a weakness. While vibe-coding platforms are only now exploring how to sell into enterprise accounts, Maisa has been enterprise-native from day one. Its new product, Maisa Studio, is specifically designed to scale adoption by lowering barriers to entry for businesses that want to try AI automation without risking reliability or compliance.
And the company is growing quickly. With dual headquarters in Valencia and San Francisco, Maisa already has one foot firmly planted in the U.S. Its $5 million pre-seed round, raised in December 2024, was led by San Francisco-based venture firms NFX and Village Global. The latest $25 million seed round adds even more firepower, with Forgepoint Capital International participating via its joint venture with Spanish bank Banco Santander — a strong signal of Maisa’s appeal in regulated sectors like finance.
Maisa is also clear-eyed about its competition. The company is entering an increasingly crowded space, with players such as CrewAI and a growing swarm of AI-powered workflow automation startups fighting for market share. Villalón isn’t shy about the stakes: in a recent LinkedIn post, he warned that the current “AI framework gold rush” often leads companies down the wrong path — where a quick proof-of-concept turns into a nightmare when systems can’t scale or errors can’t be fixed. Maisa’s pitch, in contrast, is simple: reliability, auditability, and trustworthiness from day one.
To deliver on this promise, Maisa plans to nearly double its team from 35 employees today to between 60 and 65 by the first quarter of 2026. The company also has a waitlist of eager customers ready to be onboarded starting in late 2025, which Villalón says will kickstart a new phase of rapid growth.
His message to the market is bold: “We are going to show that there is a company delivering on what’s been promised — and that it’s working.”
With $25 million in fresh capital, a growing roster of enterprise clients, and a differentiated approach that emphasizes trust, Maisa AI is betting it can do what most generative AI pilots so far have failed to achieve: turn the hype into actual, measurable business impact.