At a company Hanukkah party in December 2022, the proposed entertainment involved children, story ideas and a chatbot. Each child would supply an idea to GEKO, and the system would turn it into a story. Oded Tahori, then CEO of ONE BI, described the plan while explaining a newly released Hebrew interface built around GPT-3. It was an unusually approachable introduction to a technology soon to acquire a formidable amount of corporate vocabulary. Before the governance meetings, there would be stories.
The team had taken platforms developed for robots and other customers and packaged them into a public website. The beta had already answered more than 10,000 questions. Tahori wanted people beyond the technically initiated to try it, and wanted his team to learn what they would do. Even the presentation belonged to the experiment: the engine chose the site's name, while AI tools supplied its logo and icons. The machine was helping decorate its own shop window.
That small episode offers a useful way into Tahori's career. He has worked on several ways of making technology usable: business intelligence, service robotics, conversational interfaces and now enterprise AI. The setting changes, as does the machinery. The question of how people actually use it keeps returning. At Jeen.ai, where he is founder and CEO, the people include employees inside organizations whose information, budgets and permissions cannot be treated as an invitation to improvise.
Leaving a company that felt like family
Tahori founded Sense BI before moving into ONE Technologies, which acquired the business in 2018 for NIS 26 million. He subsequently led the ONE BI division and served as ONE's head of innovation. The acquisition belongs to an earlier chapter of his story, when helping organizations work with data was the central business. It also gave him experience with the kinds of customers he would later pursue for generative AI.
His career included leading ONE Robotix, with a CEO tenure recorded from January 2020 through April 2024. In his account of that period, the work involved putting service robots into restaurants and public centers. Those are environments where a clever demonstration has to encounter people who have somewhere else to be. Robotics gives software a rather unforgiving audience: someone standing in front of it, expecting a response.
When he announced the move to Jeen in 2024, Tahori described closing a chapter of almost 20 years. He thanked the people who had shared its successes, challenges, ups and downs. He also described the former team as family and the farewell as difficult. That is a more personal account of a spin-off than the language of a transaction normally allows. A new business can be exciting while leaving the old one remains painful.
The transition carried people and accumulated work into the next venture. Jeen was built on the AI team that had developed within ONE BI. Tahori's account therefore places the founding inside an existing history of collaboration. The new name marked a departure; the team supplied continuity. His thank-you message gives the career its human scale, alongside the companies and titles.
A robot has to understand the room
During his robotics years, Tahori talked about giving people a sense of personal attention in public spaces. In a later post about the temi service robot, he described work on recognizing age and gender, alongside continuing improvements in language understanding, particularly natural Hebrew. Whatever the interface looked like, the ambition was to make its responses fit the person using it.
The emphasis on Hebrew also appeared in GEKO. Language was part of the effort to open the technology to a wider public. A person should be able to ask a question in familiar words and receive a usable answer. The same concern runs through those projects at different scales: a robot approaching someone, a chatbot accepting a child's idea, an organizational assistant working with internal information.
In April 2024, Tahori described a conference talk about the gap between the fantasy of implementing generative AI in organizations and the complicated, changing reality. His post was warm about meeting friends and grateful for the invitation. The subject was less carefree. Moving from an individual trying a tool to a business adopting it brings another set of obligations into the room.
July 2024: the experiment gets its own office
Jeen became independent in July 2024, following a spin-off of ONE Technologies' generative AI activities. In September 2024, it operated from Tel Aviv offices with around 25 employees, and ONE held a 25 percent stake. The launch figures show a business beginning with an existing team and a corporate connection.
Yossi Wolf joined the founding effort and was appointed chairman. His background included Roboteam and the temi robot business. Tahori's experience in data and enterprise implementation met Wolf's experience in robotics and international distribution. Jeen's early customers included Isracard, Delek, Reichman University, Amdocs and Cellcom. The company was entering organizations already accustomed to substantial information systems.
Its early offer included corporate chat, transcription and analysis, assistants using internal company information, and customer-facing bots. Installation could take place in the cloud or in a secure local environment on NVIDIA servers. The range matters because Tahori was selling a way to support several organizational uses. A single appealing conversation was only the opening act; the rest of the performance involved fitting into the customer's environment.
The assistant needs permission to be helpful
In September 2024, Tahori explained why large organizations presented a different problem from individual users. A bank or insurer has sensitive information and approval requirements. A new supplier cannot assume that this information will be handed over. He also emphasized the work needed to adapt a system to the customer's particular infrastructure. His experience integrating business intelligence systems gave him a concrete frame for that difficulty.
An internal assistant, in his description, needed to understand who the user was, what permissions applied and what obligations existed. Fluency could make an answer pleasant to read. Connecting it to the organization made it relevant to the job. The distinction is easy to appreciate: an assistant with an excellent vocabulary still needs to know which door it is allowed to open.
By November 2024, his forecast extended to the organizations themselves. He expected AI-assisted development to encourage companies to build more of their own systems, with consequences for the software they bought and the way IT was organized. That forecast helps explain the ambition behind Jeen: to participate in changes to how organizations produce and operate software.
Going abroad takes people on the ground
In May 2025, Tahori discussed Adani Electricity Mumbai working with Jeen's Indian distributor. He credited the local representatives, the ability to run the product inside the customer's environment and the work of adapting it to requirements. His explanation of an international opportunity was strikingly practical. Geography entered the story through partners able to sell and implement the system, alongside the technology itself.
He also set out his aspiration to build a substantial global generative AI company. He identified international distribution, keeping pace with technology, training implementation partners and growing steadily as challenges. It is an ambition with a considerable to-do list. The installer, the trainer and the local representative all have parts to play in a story that could otherwise be told entirely through a founder's forecasts.
His public description of the EY partnership added another connection. Jeen would provide capabilities including management of multiple models, voice and chat, workflow automation and no-code agent creation. EY brought experience with enterprise integration and regulated environments. Tahori presented the collaboration as a way to help large organizations move from pilots into production, with deployments suited to their architecture.
He also continued to test the technology. In a March 2025 post, he described connecting AI21's Jamba 1.6 to Jeen and examining its performance, efficiency and response speed. The post shared a small holiday-themed test. Here was a familiar combination: a new model, a practical trial and enough playfulness to make the experiment worth showing someone else.
When the internet connection is absent
By January 2026, Tahori described a shift from building and proving Jeen's capabilities toward larger customer commitments. He emphasized deeper connections to customers' infrastructure and explained why the company concentrated on large organizations: they could support the teams and longer investment periods that substantial implementations required. A signed contract mattered to him as evidence that a customer was choosing the technology.
A separate defense subsidiary reflected differences in staffing, security clearance, sales and delivery. In July, Tahori recalled an origin in civilian robotics and said defense customers had approached after seeing the platform. He put the operational requirement plainly: “AI without the internet.” Jeen's ability to run offline under security controls had become central to that part of the business.
An August 2026 announcement brought a NIS 14.9 million defense order, described as the largest received under the relevant agreement to that date. Tahori connected the repeat orders to customer confidence and expanding use. The repeat orders gave that expansion a visible commercial footing. The August contract marks a particular point in the development of this branch of his work.
AI without the internet.Oded Tahori / July 2026
The bill gets a vote
By 2026, Tahori was also discussing AI expenditure and the need to connect it to business outcomes. Jeen shared his comment from an interview about the enthusiasm for consuming tokens, the units used to measure model usage: “If you want to spend a million on tokens, you should do something good with that in your business.” It is a refreshingly legible sentence in a field fond of making simple things sound expensive.
His publicly stated interest in deployment choice fits the same practical concern. Organizations should be able to decide where models run, across public cloud, private cloud and their own infrastructure. In his account, flexibility, governance and control belong together. Choosing a model begins the work; deciding where it operates and under whose authority brings it into the organization.
Return to the Hanukkah party and the distance becomes clear. A child offered an idea and received a story. The business customer has to ask further questions about information, permissions, installation and spending. Tahori's career has followed that widening circle of responsibility. The curiosity is still visible in his tests and demonstrations. The latest chapter gives it a budget, an implementation partner and a set of keys.