While AI chatbots have become increasingly proficient at answering queries, summarizing texts, and solving math problems, they primarily function as solo assistants for individual users. They are not built to handle the complex realities of true teamwork: aligning people with different goals, monitoring ongoing decisions, and maintaining team cohesion over extended periods. The startup Humans&, founded by veterans from Anthropic, Meta, OpenAI, xAI, and Google DeepMind, believes bridging this divide represents the next significant evolution for foundational AI models. This week, the company secured a $480 million seed investment to develop what it calls a "central nervous system" for an economy that integrates both people and artificial intelligence.

Early discussions have focused on the company's mission of "AI for empowering humans," but its core objective is more innovative: to construct a new foundational model architecture engineered for social intelligence, rather than merely retrieving information or generating code. The central proposition from Humans& is to guide society into a new AI age, shifting the conversation away from fears of job displacement. Regardless of whether this is seen as marketing, the moment is pivotal: businesses are evolving from simple chat interfaces to autonomous AI agents. While the models themselves are capable, organizational workflows are not, and the fundamental problem of coordination is still largely unsolved. Amid this shift, many people feel both anxious and overwhelmed by the rapid advance of AI.

Despite being only three months old, Humans& has attracted substantial seed funding based on its vision and the esteemed background of its founders. The company does not yet have a finished product and has been somewhat vague about its precise form, though the team has suggested it could serve as a new solution for multi-user environments such as communication tools (like Slack) or collaborative platforms (such as Google Docs and Notion). Regarding potential applications and customers, the team indicated interest in both corporate and general consumer markets.

Co-founder Zelikman illustrated the need by describing the often tedious process of reaching a consensus within a large group, like choosing a company logo, which typically requires gathering everyone to express conflicting opinions. He added that their new model will be trained to ask questions in a manner that feels natural, akin to a friend or colleague genuinely seeking to understand you. He noted that current chatbots are programmed to ask frequent questions, but they do so without grasping the purpose or value behind the inquiry, as they are primarily optimized for user satisfaction with an immediate reply and for providing a correct answer.

Some of the ambiguity surrounding the product may stem from the fact that Humans& is still determining the final details. Co-founder Peng explained that the product is being designed in tandem with the model's development. She emphasized that part of their work involves ensuring the user interface and the model's capabilities evolve together into a coherent final product. What is evident, however, is that Humans& is not aiming to create just another model that can be integrated into existing apps and tools. The startup intends to own the entire collaboration layer itself.

The intersection of AI with team collaboration and productivity tools is a rapidly growing area. For instance, the AI note-taking app Granola recently raised $43 million at a $250 million valuation as it expanded its collaborative functionalities. Furthermore, several prominent thinkers are now defining the next stage of AI as one focused on coordination and collaboration, not merely automation. LinkedIn founder Reid Hoffman recently argued that many companies are misapplying AI by treating it as a series of isolated experiments, asserting that the true power lies in enhancing the coordination layer of work—how teams share knowledge and conduct meetings. Hoffman stated on social media that AI operates at the workflow level, and the individuals closest to the actual work are best positioned to identify what should be automated, streamlined, or completely reimagined.

This is precisely the domain Humans& aims to occupy. The vision is for its combined model and product to serve as the "connective tissue" within any organization, whether a large corporation or a small family, by comprehending each person's abilities, motivations, and needs, and then balancing these for the collective benefit. Achieving this requires a fundamental rethinking of AI training methodologies. Techniques like long-horizon reinforcement learning are intended to train models to plan, act, adjust, and execute over time, rather than just producing a single good response. Meanwhile, multi-agent reinforcement learning prepares models for environments involving multiple AI systems and human participants. Both approaches are gaining traction in academic research as scientists push large language models beyond simple chatbot replies toward systems capable of coordinating actions and optimizing outcomes across multiple steps.

As co-founder He explained, the model must be able to remember details about itself and the user, with the quality of its understanding improving alongside its memory. Despite the highly experienced team leading the project, significant challenges lie ahead. Humans& will require continuous and substantial funding to support the costly process of training and scaling a new model. This means competing with well-established industry giants for essential resources, particularly computing power.

The foremost risk, however, is that Humans& is not only competing with collaboration platforms like Notion and Slack. It is entering the arena against the leading AI companies themselves. These major players are also actively developing better methods to facilitate human collaboration on their own platforms, even as they discuss the potential for artificial general intelligence to reshape the workforce. For example, Anthropic's Claude Cowork is focused on enhancing work-style collaboration; Google's Gemini is integrated into Workspace, bringing AI-powered collaboration directly into widely-used tools; and OpenAI has been promoting its capabilities in multi-agent orchestration and workflow management to developers.

A critical differentiator is that none of the major incumbents appear to be rebuilding a model from the ground up based on social intelligence. This could provide Humans& with a unique advantage or, alternatively, make it an attractive acquisition target. With firms like Meta, OpenAI, and DeepMind continuously seeking top AI talent, the possibility of a merger or acquisition is a real consideration.

"We believe this is going to be a generational company, and we think that this has the potential to fundamentally change the future of how we interact with these models," Zelikman stated. "We trust ourselves to do that, and we have a lot of faith in the team that we’ve assembled here."