They were joined in the round by local business angels, Sequoia scout and DeepMind PM Lukas Haas. 

Investors were drawn to the Maisa team and its development of the Vinci Knowledge Processing Unit (KPU), which delivers cutting-edge AI reasoning and execution for commercial use. This technology enables developers and businesses to harness state-of-the-art agentic AI, enhanced by the computational traceability of Maisa’s innovative Chain-of-Work approach.

The Vinci KPU works by leveraging existing LLMs – such as those from Anthropic, OpenAI and Google – in a completely new way to become a more powerful reasoning and execution AI system boasting world-leading benchmarks.

The money is to be used to continue developing the product and boost its already strong go-to-market strategy.

It comes as latest benchmarking research shows Maisa is the first and only agentic AI system to rank alongside OpenAI’s reasoning model o1 in graduate-level reasoning (GPQA) and that it outperforms latest frontier LLMs in coding (HumanEval), maths problem solving (Math), following procedures (ProcBench) and in accuracy. This level of performance is unrivalled by any other agentic AI on the market.

Despite the excitement AI has created through products like ChatGPT and o1, astonishing users with its apparent knowledge and capabilities, AI has actually had limited take-up in business automation to date.

A recent survey by Amazon Web Services revealed that although 25% of businesses have their staff experimenting with generative AI, only 6% have put its use into live production environments. And that number reduces in just three months as problems arise.

The reason is a number of serious and potentially dangerous flaws in AI to date, which Maisa’s Vinci KPU solves.

The key issue with generative AI is its inability to show and prove its work. Chain-of-Thought creates some feeling of traceability, but it’s actually still a predictive explanation of probabilistic results. Even when enhanced with RAGs, which itself have inherent flaws, or used in multi-agent setups, the final output is not the deterministic result of some computation, but a next word prediction. This lack of true ‘traceability’ or lack of proof of its work has made many use cases in a business context non-viable.

As a world first, the new Maisa system offers Chain-of-Work to prove the outcomes with full transparency and step-by-step traceability – meaning it can deal with open-ended and complex tasks that require auditability, radically increasing the value that AI can deliver for businesses.  

This could include performing supply chain analysis, credit loan approval processes or carrying out science-led research and development. These tasks move AI far beyond chatbots, copilots and agents which perform only limited support functions that don’t require deterministic execution, repeatability or full explainability and attribution.

David Villalón, Co-Founder and CEO of Maisa, said: “Our launch of the Vinci KPU is significant in the development of agentic AI for businesses and for developers of AI applications and automation solutions.

“This is the first time agentic AI becomes transparent and auditable, instead of the current ‘black box’ that obscures its outcomes and represents a huge risk for businesses.

“We are delighted to have gained the backing of such legends in global tech development and leaders of some of the largest firms in the world, not least for the recognition that their investment gives to the significance of what we are building.”

Anna Piñol, investor at NFX, said: “Maisa is going to be a major player in RPA 2.0 helping businesses across the world transform their core, business-critical functions through AI. It will allow them to work faster, more efficiently and achieve new and radical ways of operating.

“This launch will be seen as a pivotal moment in AI development this year and as AI's coming of age in practical, business-enhancing use.”

Max Kilberg, investor at Village Global, said: "David and the Maisa team are building a transformative technology to turn AI agents into actual workers that are capable of reasoning through complex workflows. We're super thrilled to be a part of their journey and are very excited to see the new benchmarks and enterprise traction."

The system is unique in taking a computational approach to reasoning, revealing and executing each step as it is taken, using internal capabilities and interacting with external tools in a deterministic way.

This is in marked contrast to existing AI ‘black box’ LLMs, which predict based on natural language what they deem as the most probable answer based only on the data they are trained on or retrieved from RAG into their context window, with no observability of how know-how was applied.     

The full auditability and unprecedented accuracy of the Vinci KPU is already transforming AI’s practical application – opening the door to its use in far more complex problem solving and execution, and in core, business-critical functions.