A demonstration has a useful advantage over a working day: it knows when to end. The numbers appear, the answer sounds sensible, the audience nods. Then somebody has to put the system inside a business, where the data changes, permissions matter and yesterday’s sensible answer may be today’s expensive mistake. Rotem Alaluf has built much of his career around that awkward second act.
At Wand AI, the company he co-founded in 2022, he is pursuing a version of the problem that includes an organization chart. AI agents, in his account, will take on roles, execute work and collaborate with people. They will need management. The prospect gives software a rather human burden: someone will expect it to explain what it has been doing.
Alaluf’s public career makes the argument more interesting than a prediction delivered from a comfortable distance. He trained as an engineer, worked on difficult AI systems, founded an enterprise software company and left it. That company subsequently closed. He returned to enterprise AI with Wand. The vocabulary grew more ambitious. The concern with actual execution remained.
An education in things that go wrong
Alaluf studied electrical engineering and physics at Technion - Israel Institute of Technology, attending from 2013 to 2017. The combination supplies a fitting starting point for this story. Physics asks what a system does. Engineering asks whether it can keep doing it under conditions that somebody else has chosen.
His early professional work was in Israel’s defense technology industry. His public biographies describe research involving optimization, natural language processing and the vulnerabilities of AI systems. He also worked on time-series and speech processing. These are fields in which information arrives with complications attached. A system’s ability to handle the complications is part of its usefulness.
His career record includes algorithm engineering from 2014 and a later role leading deep learning and NLP work before his departure in 2018. This was a route into AI through implementation. The distinction matters to the questions he would ask as a founder: how does a model behave outside the conditions under which it was built, and what surrounds it when it does?
In 2018, Alaluf co-founded BeyondMinds with Or Kremer and Roey Mechrez. The company worked on enterprise AI, helping organizations move solutions into production. Mechrez later described the founding problem as a gap between what AI research could demonstrate and the value companies could obtain in practice. The founders were choosing to work inside that gap.
The first company does not vanish
By 2021, Alaluf was writing a four-part series about the market failure of AI. Its subjects included the appetite for proofs of concept, the messiness of real data and the need to deploy multiple solutions efficiently. The titles were hardly an invitation to an untroubled victory lap. He was writing about the machinery required after an enterprise decided AI sounded promising.
There is something revealing about that choice of subject. A founder selling AI had reasons to dwell on its capabilities. Alaluf also dwelt on the conditions that could prevent those capabilities from becoming useful. The argument gave maintenance, deployment and organizational fit a place beside the model.
He left BeyondMinds in September 2021. His departure followed a disagreement with the board about the company’s direction. In May 2022, BeyondMinds shut down after a possible sale fell through, with 65 employees losing their jobs. Alaluf had already moved toward a new venture; reporting at the time described Wand as being in stealth.
The sequence resists an easy founder’s fable. BeyondMinds had recognizable enterprise clients and had attracted outside funding. It still closed. Technical usefulness, financing and agreement over a company’s future are separate problems, even when they occupy the same meeting room. The ending belongs in Alaluf’s story because it puts a boundary around any account made entirely of ambition.
Co-founded
Leaves CEO role
New venture
Book published
A place in the organization chart
Wand carried forward his interest in making AI accessible to people who have business expertise without necessarily having technical expertise. In a February 2024 fireside conversation hosted by Rebellion Ventures, Alaluf and managing partner Jukka Alanen discussed business assistants, data insights and automation. The framing began with the user: what could a person get done with help from AI?
The company’s current language reaches further into the organization. Wand describes infrastructure for a hybrid workforce of people and agents, with oversight, interoperability and controls. Its pitch places the management of agents alongside their execution. An enterprise, in this account, needs a way to organize the activity it has authorized.
Calling an agent a member of a workforce changes the expectations attached to it. A role has a purpose. Authority has limits. Work has dependencies. The label invites questions about those things, which is useful even for a reader unconvinced by the larger prediction. Software that acts on behalf of a business will have to encounter the business’s rules.
It also exposes the distance between an engaging interface and a dependable working arrangement. A request can cross departments and tools. An answer can be plausible while an action is inappropriate. The interesting part of Alaluf’s proposition lives in the arrangements around the action: permission, coordination, review and the route back to a human decision.
“AI agents are not just tools - they are a new class of labor.”
Rotem Alaluf, Nityo partnership announcement
The useful art of asking for help
At the World Policy Conference in Abu Dhabi in December 2024, Alaluf outlined agents that would work together, learn from execution and seek human assistance when they encountered a problem. The panel, chaired by Lucia Sinapi-Thomas of Capgemini Ventures, placed his enterprise argument in a conversation that also included policy and other approaches to AI.
One detail gives the vision a recognizable office texture: the agent that gets stuck should be able to approach a person. His ambition includes autonomy, but it also includes a return path to human judgment. He imagines people supplying feedback while agents gradually take on more execution, with human roles moving toward strategy and creativity. This is his forecast, rather than an established description of every workplace.
Asking for help is an unusually useful feature to put near the center of an ambitious technology story. Anyone who has worked in an organization recognizes the cost of an employee confidently misunderstanding a request. A machine can supply confidence without the office experience that makes a person hesitate. Building a reliable moment of hesitation is a substantial part of the challenge.
Alaluf has taken these questions into teaching and discussion. He has guest lectured at UC Berkeley, publicly describing a return to the Haas fintech program to discuss enterprise transformation. In February 2025, INSEAD hosted him with strategy professor Peter Zemsky for a discussion of agentic AI and autonomous organizations. Trust, explainability and human involvement were on the agenda.
There are lighter details around the lectures. After a Bank of America conference in London, he described a program containing a live quantum computer demonstration, a humanoid robot, flying cars and 3D-printed steaks for lunch. The future apparently had a catering plan. On LinkedIn, he also lists a deep-learning music-composition project, a small reminder that his interest in models has included making things as well as managing them.
The idea gets a working address
The practical tests have become more specific. In July 2024, Presight announced a partnership with Wand to bring generative AI assistants to enterprises in the United Arab Emirates. The planned applications spanned sectors including finance, retail, real estate, construction and legal work. Existing information resources and workflows were part of the deployment proposition.
In February 2025, Alaluf announced Wand’s acquisition of Accern, a no-code NLP platform. He described the purpose in terms of extracting context from unstructured information and enhancing what agents could do with data. He welcomed Kumesh Aroomoogan, Cristian Felix and the Accern team. For a workforce expected to execute tasks, access to usable information is a practical prerequisite.
Franklin Templeton supplied a more concrete enterprise milestone in November 2025. The investment management firm announced a multi-year strategic partnership with Wand, describing a collaboration that had moved from pilots to production deployments across multiple departments. It said further use in enterprise workflows was expected by 2026. Governance and controls were central to the announcement.
That distinction between deployed work and expected expansion matters. A partnership creates a setting in which an idea can face ordinary institutional demands. It does not settle every claim made about the future of work. The value of this milestone is its specificity: a named organization, a described move into production, and responsibilities around how the systems operate.
A separate Nityo Infotech partnership paired Wand’s technology with the services company’s international delivery network. Its announcement emphasized end-to-end workflows, escalation of difficult cases, auditability and centralized control. These are the prosaic requirements that return whenever Alaluf’s expansive vocabulary reaches an actual customer.
An office-sized argument grows
His activities also extend beyond Wand. Alaluf is a co-founder, director and technical advisor of GenAI Works. He describes investing in AI companies and serving on advisory boards. Those roles place him in a wider network of people trying to turn AI capability into products, businesses and education.
By September 2026, the argument had acquired a book. The Agentic Age, listed with a September 30 publication date, considers software as an economic actor and explores implications for companies, countries and institutions. Its scope includes governance, identity and accountability. The question that began inside an enterprise now reaches the rules around the enterprise.
There is a coherent line through the expansion. Alaluf keeps asking what must surround intelligence before it can do useful work. Earlier, the surrounding structure was deployment technology. At Wand, it includes the management of agents. In the book, it includes institutions. Each enlargement brings more people, decisions and obligations into the frame.
The ambition will be judged through execution, the subject he has kept returning to. A workforce needs to produce work someone can use. A business needs to survive while supplying it. And an agent, however fluent, needs a sensible way to reach the person who can answer the question it cannot. The next working day offers plenty of questions.