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

A wonderful 95% of the generators AI pilots in agencies FailingIn accordance Report Published by the Nanda initiative of MIT. However, instead of quit the technology, the most advanced companies are experimenting with agents AI systems that can learn and oversee.

It From Comes It is now launched with a new, $ 25 million seed round led by European VC Farm Cradam Has been the studioA model-non-immune self-service platform that helps users to deploy digital staff that can be trained with natural language.

Although it may seem familiar-the so-called Vibb coding platforms, such as the cursor and the crandum-backed profitable, have argued that-its procedure is largely different. Mesa CEO David Vilalan told TechCrunch, “Instead of using AI to create response, we use AI to make the process that we need to get in response-which we call ‘chain-off-work’.

The main architect behind this process is Misa’s co-founder and Chief Scientific Officer Manuel Romero, who had previously worked with Vilalon Spanish AI Startup ClibrenThe In 2024, this pair for the first time this “You can’t rely on AI,” Vilalan said “after watching” after watching the hallucination solution came together.

This pair is not suspicious about AI, but they think that it would not be possible for people to review “doing three months in five minutes”. To solve it, Maisa stood up for human-urgent LLM processing and hired a system called HLP. This custom method acts like students on the blackboard – it outlines every step they will follow when the digital staff asks about their needs.

The startup has also developed the knowledge processing unit (KPU), A deterministic system designed to restrict hallucinations. Starting from this technical challenge, not in the use of Maisa, it soon found that it was resonated with companies in the hope of applying AI to AI critical work on faithfulness and accountability. For example, clients who currently use MASA in production include a large bank, as well as car manufacturing and energy sector companies.

This enterprise clients serve clients, Mesa is hopeful that the robotic process will position themselves as a more advanced form of automation (RPA), which companies unlock productivity without the need to depend on strict predetermined rules or broad manual programming. To meet their needs, the startup also provides the deployment of their protected clouds or by deploying on-primitives.

This Enterprise-Firm method means Mya’s customer base is still much smaller than shaking on Freemium Vibe-Coding Platforms. However, these platforms are now exploring how to win enterprise customers, Mesa Misa is moving in the opposite direction with the studio, which is designed to raise its customer funnel and acceptance.

The startup is also planned to extend with existing customers who have operations in multiple countries. With the dual headquarters in Valencia and San Francisco, Maisa itself has already set a foot in the United States, as its cap is reflected in the table; It Million 5 million pre-bees round Last December, San Francisco -based enterprise was NFX and Village Global.

Furthermore, TechCrunch has learned exclusively that the US Farmpoint Capital International has participated in this new round European joint venture With the Spanish Bank Banko Santander, it highlighted its application for controlled sectors.

Focusing on complex use in demanding accountability from non-technical users can lead to Mya’s defarator, whose competitors include Creely And many other AI-driven, business-centric workflow automation products. In a LinkedIn post, Villalan Highlight This “AI Framework is the gold Rush,” warns that “Quick Start” turns into a long nightmare when you need reliability, monitoring or what is wrong. “

Due to the AI ​​scale to assist, Mesa plans to use its funds to extend between 5 and 65 in the first quarter of 2026 to meet the needs. Starting at the end of this year, the startup is expected to develop rapidly as the startup starts to serve its waiting list. “We’re going to show the market there is a company that has been promised and it is working,” said Vilalan.