Humans& Secures $480M Seed for AI Collaboration Models

A new artificial intelligence startup, Humans&, has emerged with a bold vision: to build the next generation of foundation models designed for collaboration and social intelligence, moving beyond the current focus on individual chat interactions. Founded by a stellar team of alumni from leading AI labs including Anthropic, Meta, OpenAI, xAI, and Google DeepMind, Humans& recently secured a substantial $480 million seed funding round to pursue this ambitious goal.

Beyond Chatbots: The Need for Collaborative AI

While today's AI chatbots excel at tasks like answering questions, summarizing documents, and solving equations, their utility largely remains confined to assisting a single user. They are not built to navigate the complexities of real-world collaboration, which involves coordinating individuals with diverse priorities, tracking long-term decisions, and maintaining team alignment over time. This gap, according to Humans&, represents the next major frontier for foundation models.

Andi Peng, a co-founder of Humans& and former Anthropic employee, highlights this shift. "It feels like we're ending the first paradigm of scaling, where question-answering models were trained to be very smart at particular verticals, and now we're entering what we believe to be the second wave of adoption where the average consumer or user is trying to figure out what to do with all these things," Peng told TechCrunch. The startup aims to guide people into this new AI era, shifting the narrative away from job displacement and towards empowerment.

Building a "Central Nervous System" for Human-AI Interaction

Humans& envisions its technology as a "central nervous system" for the human-plus-AI economy. Their core ambition is to develop a novel foundation model architecture specifically engineered for social intelligence, rather than merely information retrieval or code generation. This approach seeks to address the current disconnect where AI models are competent, but the workflows they integrate into are not, leaving the critical challenge of coordination largely unaddressed.

Eric Zelikman, co-founder and CEO of Humans& and a former xAI researcher, emphasized the company's focus. "We are building a product and a model that is centered on communication and collaboration," he stated, adding that the goal is to help people work together and communicate more effectively—both with each other and with AI tools. He cited the common struggle of reaching group consensus, like agreeing on a company logo, as a prime example of the "time-consuming tedium" their AI aims to alleviate.

Zelikman further explained that their model will be trained to ask questions in a manner akin to a friend or colleague, genuinely seeking to understand the user. This contrasts with current chatbots that often ask questions without grasping their underlying value, primarily optimizing for immediate user satisfaction or factual correctness.

A New Paradigm for AI Training

To achieve its vision, Humans& is fundamentally rethinking how AI models are trained. Yuchen He, a co-founder and former OpenAI researcher, explained that their approach will involve more intensive human-AI interaction and collaboration during the training process. The startup's model will also leverage long-horizon and multi-agent reinforcement learning (RL).

• Long-horizon RL: This trains the model to plan, execute, revise, and follow through on actions over extended periods, rather than merely generating a single, isolated good answer.
• Multi-agent RL: This prepares the model for environments where multiple AIs and/or humans are actively involved in the loop, fostering coordinated actions and optimized outcomes across numerous steps.

"The model needs to remember things about itself, about you, and the better its memory, the better its user understanding," He noted, underscoring the importance of contextual awareness and persistent learning.

Owning the Collaboration Layer

While Humans& has yet to unveil a specific product, the team hinted at applications that could replace existing multi-user communication and collaboration platforms like Slack, Google Docs, or Notion, targeting both enterprise and consumer markets. Crucially, Humans& isn't aiming to create a model that merely plugs into existing applications; it intends to own the entire collaboration layer.

This strategy aligns with a growing industry sentiment. The field of AI-powered team collaboration and productivity tools is rapidly expanding, exemplified by startups like Granola, which recently raised a $43 million round. Influential figures like LinkedIn founder Reid Hoffman also advocate for this shift, arguing that companies often misimplement AI by treating it as isolated pilots. Hoffman asserts that "the real leverage is in the coordination layer of work – that is, how teams share knowledge and run meetings." He wrote on social media, "AI lives at the workflow level, and the people closest to the work know where the friction actually is. They’re the ones who will discover what should be automated, compressed, or totally redesigned." This is precisely the domain Humans& seeks to inhabit, acting as the "connective tissue" across any organization.

Navigating Risks and Competition

Despite the formidable founding team and significant funding, Humans& faces considerable challenges. Training and scaling a new foundation model is an immensely expensive undertaking, requiring vast sums of capital and access to critical resources like computing power, placing them in direct competition with established tech giants.

The primary risk, however, extends beyond competing with collaboration tools like Notion or Slack. Humans& is positioning itself against the leading AI developers themselves. Major players such as Anthropic (with Claude Cowork), Google (integrating Gemini into Workspace), and OpenAI (pitching multi-agent orchestration) are actively enhancing human collaboration features within their own platforms. Yet, none of these giants appear poised to fundamentally rewrite their models based purely on social intelligence, potentially giving Humans& a unique advantage or, conversely, making it an attractive acquisition target for companies like Meta, OpenAI, and DeepMind, which are constantly seeking top AI talent.

Humans&, however, remains resolute. The startup has already declined acquisition offers, with Zelikman stating, "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. We trust ourselves to do that, and we have a lot of faith in the team that we’ve assembled here."