Maisa AI offers Maisa Studio, a model-agnostic, self-serve platform for deploying trainable digital workers that users can instruct with natural language. The company emphasizes accountable, agentic systems (HALP: Human-Augmented LLM Processing) and a deterministic Knowledge Processing Unit (KPU) to limit hallucinations. Its product is targeted at enterprise use cases that require reliability, auditability and the option of secure cloud or on-premise deployment. Current customers include a large bank and companies in the car manufacturing and energy sectors, though the startup’s enterprise customer base remains relatively small. Maisa positions itself as a more advanced form of RPA focused on trustworthiness rather than rigid rules or heavy manual programming. The company plans to scale headcount from 35 to as many as 65 people by Q1 2026 to meet anticipated demand. Maisa is a Spanish deeptech startup developing Vinci KPU, a system that it says "executes reasoning" computationally, step by step, using attributable sources rather than next-word prediction. The company leverages existing LLMs (Anthropic, OpenAI and Google) differently to reduce hallucinations and provide traceability and auditability for business outcomes. Maisa targets open-ended, complex tasks that require audit trails, including supply chain analysis, credit loan approval processes and science-led R&D. Early customers include an international car manufacturer, an international oil and gas company, and several financial services and cybersecurity firms. Latest benchmarking cited in the article places Maisa alongside OpenAI’s reasoning model o1 on graduate-level reasoning (GPQA) and shows it outperforming frontier LLMs on coding (HumanEval), maths (Math), procedure following (ProcBench) and accuracy. The company raised $5M in a Pre-Seed round and will use the proceeds to continue product development and accelerate its go-to-market strategy.

Total raised

$30M

Funding rounds

2

Latest round

Seed

Latest activity

Aug 2025