Israeli AI Pioneer AI21 Pivots to Enterprise Agent Optimization After Restructuring
Israeli AI pioneer AI21 shifts strategy to enterprise agent optimization and post-training after cutting 60% of its workforce, securing public backing from Nvidia.
After years of strategic shifts and a severe 60% workforce reduction, Israeli artificial intelligence pioneer AI21 is refocusing on enterprise AI agent optimization, securing a public endorsement from Nvidia CEO Jensen Huang.
Founded nearly a decade ago by Prof. Amnon Shashua, Prof. Yoav Shoham, and CEO Ori Goshen, AI21 was among the early trailblazers attempting to rival global giants like Anthropic and OpenAI by building proprietary large language models. Despite raising $336 million from prominent investors including Google, Nvidia, Intel, and 8VC, the company struggled to break out in the hyper-competitive LLM market and underwent at least four business pivots.
In May, the company reached a critical crossroads, laying off 110 of its 180 employees after acquisition talks fell through. The remaining team of 70 core employees has now shifted focus toward post-training, model optimization, and managing custom enterprise AI agents. According to Goshen, this new vector has dramatically reduced enterprise operating costs by up to 50% while leveraging open-source language models from Nvidia, Meta, and DeepSeek.
Addressing the departure of co-founder Amnon Shashua, Shoham described it as a natural progression after years of diminishing involvement. Shashua previously founded Mobileye and launched a new AI venture, doubleAI, backed by Nvidia. Meanwhile, Goshen and Shoham express renewed optimism, noting that market demand for their current optimization infrastructure surpasses anything the company has experienced in its history.
