Humans& believes that the future of AI lies in mastering coordination, and they are developing a model to demonstrate this concept

Redefining AI Collaboration: The Vision Behind Humans&

While AI chatbots have made significant strides in tasks like answering queries, summarizing content, and solving equations, their primary function remains as single-user assistants. These systems are not yet equipped to handle the complexities of group collaboration, such as managing conflicting priorities, overseeing ongoing decisions, or ensuring team cohesion over extended periods.

Humans&, a startup launched by former employees of Anthropic, Meta, OpenAI, xAI, and Google DeepMind, aims to bridge this gap. The company recently secured a $480 million seed investment to develop what they describe as a “central nervous system” for a future where humans and AI work side by side. While much of the initial buzz has focused on their “AI for empowering humans” approach, the real innovation lies in their goal to create a foundational AI model built for social intelligence, moving beyond simple information retrieval or code generation.

Moving Beyond the First Wave of AI

“We’re witnessing the close of an era where AI models excelled at answering questions within specific domains. Now, we’re entering a phase where everyday users are seeking meaningful ways to leverage these technologies,” explained Andi Peng, co-founder of Humans& and former Anthropic team member, in an interview.

The company’s mission is to guide people into this new AI-driven age, shifting the conversation from fears of job displacement to opportunities for collaboration. As businesses transition from basic chatbots to more sophisticated AI agents, the challenge of orchestrating effective teamwork remains largely unsolved. Many individuals still feel anxious and uncertain about the rapid evolution of AI.

Ambitious Goals and Early Momentum

Despite being just three months old, Humans& has attracted significant funding, thanks in part to the impressive backgrounds of its founders. The company has yet to unveil a concrete product, but hints suggest it could eventually serve as an alternative to current multi-user platforms like Slack, Google Docs, or Notion. Both enterprise and consumer applications are on the table.

“Our focus is on building both a product and an AI model that prioritize communication and collaboration,” said Eric Zelikman, CEO and co-founder of Humans& and a former xAI researcher. The aim is to help people interact and cooperate more effectively, whether with each other or with AI-powered tools.

“When a group needs to make a major decision, it often requires gathering everyone together to discuss their preferences—like choosing a company logo,” Zelikman recalled, sharing a laugh with his team about their own lengthy logo selection process.

Designing AI for Human-Like Interaction

Zelikman emphasized that their model will be trained to ask questions in a manner reminiscent of a thoughtful friend or colleague, rather than simply firing off queries without context. Current chatbots tend to focus on immediate user satisfaction and accurate answers, but often lack an understanding of the deeper purpose behind their questions.

The company is still refining its product vision, with Peng noting that the development of the user interface and the AI model are happening in tandem. “We want the product to evolve alongside improvements in the model, ensuring the interface and AI behaviors make sense together,” she said.

Owning the Collaboration Layer

Unlike many AI startups aiming to integrate with existing collaboration tools, Humans& intends to create its own platform for teamwork and communication.

The intersection of AI and productivity tools is rapidly gaining attention. For instance, the AI note-taking app Granola recently raised $43 million at a $250 million valuation as it introduced new collaborative features. Industry leaders like LinkedIn founder Reid Hoffman have also highlighted the importance of AI in enhancing team coordination, rather than treating it as a series of isolated pilots.

“AI operates at the workflow level, and those closest to the work are best positioned to identify what should be automated or improved,” Hoffman shared on social media.

Building a Socially Intelligent AI

Humans& envisions its product as the connective tissue within organizations of any size, understanding each individual’s skills, motivations, and needs, and balancing them for collective benefit.

Achieving this requires a new approach to AI training. “We’re developing our model through collaborative interactions between humans and AIs,” said Yuchen He, co-founder and former OpenAI researcher. The training process will leverage both long-horizon and multi-agent reinforcement learning (RL), techniques that are gaining traction in recent research as AI moves beyond simple chatbot tasks to more complex, multi-step coordination.

• Long-horizon RL enables the AI to plan, adapt, and follow through on tasks over time, rather than just providing one-off answers.
• Multi-agent RL prepares the model for environments where multiple humans and AIs interact and collaborate.

“A model that can remember details about itself and its users will be far better at understanding and supporting people,” He added.

Challenges and Competition Ahead

Despite a strong founding team, Humans& faces significant hurdles. Developing and scaling a new AI model is resource-intensive, requiring substantial funding and access to computing power. The company will be competing not only with established collaboration platforms like Notion and Slack, but also with leading AI firms. Major players are already integrating advanced collaboration features into their products—Anthropic with Claude Cowork, Google’s Gemini in Workspace, and OpenAI’s multi-agent orchestration tools.

However, none of these giants appear to be prioritizing models built specifically for social intelligence, which could give Humans& a unique advantage—or make it an attractive acquisition target. With tech giants constantly seeking top AI talent, the possibility of mergers and acquisitions looms large.

Humans& has stated it is not interested in being acquired, having already declined offers from interested parties.

“We’re building a company for the long term, one that could transform how people interact with AI,” Zelikman said. “We have confidence in our vision and in the team we’ve brought together.”

Originally published on January 22, 2026.