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

A staggering 95% of generative AI pilots at firms are failing, in keeping with a current report printed by MIT’s NANDA initiative. However somewhat than giving up on the expertise altogether, essentially the most superior organizations are experimenting with agentic AI techniques that may study and be supervised.

That’s the place Maisa AI is available in. The year-old startup has constructed its whole strategy across the premise that enterprise automation requires accountable AI brokers, not opaque black packing containers. With a brand new, $25 million seed spherical led by European VC agency Creandum, it has now launched Maisa Studio, a model-agnostic self-serve platform that helps customers deploy digital staff that may be educated with pure language.

Whereas that may sound acquainted — harking back to so-called vibe coding platforms like Cursor and the Creandum-backed Lovable — Maisa argues that its strategy is essentially totally different. “As an alternative of utilizing AI to construct the responses, we use AI to construct the method that must be executed to get to the response — what we name ‘chain-of-work,” Maisa CEO David Villalón informed TechCrunch.

The principal architect behind this course of is Maisa’s co-founder and Chief Scientific Officer, Manuel Romero, who had beforehand labored with Villalón at Spanish AI startup Clibrain. In 2024, the duo teamed as much as construct an answer to hallucinations after seeing firsthand that “you might not depend on AI,” Villalón stated.

The pair isn’t skeptical about AI, however they suppose it received’t be possible for people to overview “three months of labor carried out in 5 minutes.” To handle this, Maisa employs a system known as HALP, standing for Human-Augmented LLM Processing. This tradition methodology works like college students on the blackboard — it asks customers about their wants whereas the digital staff define every step they’ll observe.

The startup additionally developed the Information Processing Unit (KPU), a deterministic system designed to restrict hallucinations. Whereas Maisa began out from this technical problem somewhat than a use case, it quickly discovered that its wager on trustworthiness and accountability resonated with firms hoping to use AI to important duties. As an illustration, purchasers that at the moment use Maisa in manufacturing embody a big financial institution, in addition to firms within the automobile manufacturing and vitality sectors.

By serving these enterprise purchasers, Maisa hopes to place itself as a extra superior type of robotic course of automation (RPA) that unlocks productiveness positive factors with out requiring firms to depend on inflexible predefined guidelines or intensive handbook programming. To satisfy their wants, the startup additionally affords them both deployment in its safe cloud or by on-premise deployment. 

This enterprise-first strategy means Maisa’s buyer base remains to be very small in comparison with the hundreds of thousands flocking to freemium vibe-coding platforms. However as these platforms at the moment are exploring tips on how to win enterprise prospects, Maisa is transferring in the wrong way with Maisa Studio, which is designed to develop its buyer funnel and ease adoption.

The startup additionally plans to broaden with current prospects which have operations in a number of nations. With twin headquarters in Valencia and San Francisco, Maisa itself already has a foothold within the U.S., as mirrored in its cap desk; its $5 million pre-seed round final December was led by the San Francisco-based enterprise companies NFX and Village International. 

As well as, TechCrunch realized solely that U.S. agency Forgepoint Capital Worldwide participated on this new spherical through its European joint venture with Spanish financial institution Banco Santander, highlighting its enchantment for regulated sectors.

Specializing in complicated use circumstances demanding accountability from non-technical customers could possibly be a differentiator for Maisa, whose opponents embody CrewAI and lots of different AI-powered, business-focused workflow automation merchandise. In a LinkedIn submit, Villalón highlighted this “AI framework gold rush,” warning that the “fast begin” turns into a protracted nightmare while you want reliability, auditability, or the flexibility to repair what went flawed.”

Doubling down on its objective to assist AI scale, Maisa plans to make use of its funding to develop from 35 to as many as 65 folks by the primary quarter of 2026 as a way to meet demand. Beginning within the final quarter of this 12 months, the startup anticipates speedy progress because it begins serving its ready record. “We’re going to present the market that there’s a firm that’s delivering what has been promised, and that it’s working,” Villalón stated.