In a move aimed at closing the “Trust Gap” that continues to hinder enterprise adoption of artificial intelligence, Israel-based companies Tasq AI and BLEND announced on Monday a full merger, creating a global data infrastructure firm under the Tasq AI brand.
The combined entity merges advanced proprietary technology with a worldwide network of contributors and experts, positioning itself as a key player in the rapidly growing AI data market.
The merged company, which now employs 120 people and generates tens of millions of dollars in revenue, serves 200 enterprise clients — including Meta, PayPal, Reddit, iHerb, Puma and Payoneer. Its core offering is a unified platform that ensures AI systems are trained on high-quality, trustworthy data, a foundation increasingly recognized as critical to the success of production-grade AI.
“Data has become the critical defensive layer in the $1.5 trillion AI market, that safeguards the quality and reliability of AI models,” said Yoav Ziv, CEO of the merged Tasq AI and former executive at Amdocs and Checkmarx. “We are establishing a single entity that combines technology and people to tackle the greatest challenge in AI implementation: trust, at speed and quality levels no one thought possible.”
Ziv said Tasq AI’s platform enables AI models to be trained up to 10 times faster than traditional methods, eliminating the trade-off between scale and precision that has plagued enterprise AI projects. Its proprietary multi-layer system uses a just-in-time model that taps into a global pool of millions for high-volume tasks, and directs complex assignments to a curated network of more than 25,000 vetted domain experts in over 120 languages.
This architecture supports high-speed, high-quality data labeling and model evaluation in areas ranging from fraud detection to large language model (LLM) validation and video annotation.
The need for such a platform is growing as enterprises struggle with inconsistent data pipelines. A 2024 Salesforce survey found that 76% of business leaders feel pressure to deliver value from data, while 54% of AI users say they don’t trust the data used to train their models. MIT research estimates that 95% of generative AI projects fail — not because of flawed models, but due to poor or fragmented data inputs.
Erez Moscovich, founder and president of Tasq AI, compared the data problem to a broken supply chain: “If we compare the AI revolution to the automotive revolution, NVIDIA builds the highways, AI model companies build the engines and data is the fuel. Yet data-related challenges prevent many AI systems from reaching their destination, despite massive investments in infrastructure and models.Tasq AI's mission is to ensure AI models deliver real, production-grade value and that the adoption of AI models and GPUs continues to scale. We build the modern refinery that organizes data and human expertise to power the AI revolution.”