Humans& thinks coordination is the next frontier for AI, and they’re building a model to prove it

AI chatbots are getting better at answering questions, summarizing documents, and solving mathematical equations, but they still largely behave like helpful assistants for one user at a time. They’re not designed to manage the messier work of real collaboration: coordinating people with competing priorities, tracking long-running decisions, and keeping teams aligned over time. 

Humans&, a new startup founded by alumni of Anthropic, Meta, OpenAI, xAI, and Google DeepMind, thinks closing that gap is the next major frontier for foundation models. The company this week raised a $480 million seed round to build a “central nervous system” for the human-plus-AI economy. The startup’s “AI for empowering humans” framing has dominated early coverage, but the company’s actual ambition is more novel: building a new foundation model architecture designed for social intelligence, not just information retrieval or code generation.

“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,” Andi Peng, one of Humans&’s co-founders and a former Anthropic employee, told TechCrunch. 

Humans&’s pitch centers on helping usher people into the new era of AI, moving beyond the narrative that AI will take their jobs. Whether or not that’s just marketing speak, the timing is critical: Companies are transitioning from chat to agents. Models are competent, but workflows aren’t, and the coordination challenge remains largely unaddressed. And through it all, people feel threatened and overwhelmed by AI.

The three-month-old company, like several of its peers, has managed to raise its startling seed round off the back of this philosophy and the pedigree of its founding team. Humans& still doesn’t have a product, nor has it been clear about what exactly it might be, though the team said it could be a replacement for multiplayer or multi-user contexts like communication platforms (think Slack) or collaboration platforms (think Google Docs and Notion). As for use cases and target audience, the team hinted at both enterprise and consumer applications. 

“We are building a product and a model that is centered on communication and collaboration,” Eric Zelikman, co-founder and CEO of Humans& and former xAI researcher, told TechCrunch, adding that the focus is on getting the product to help people work together and communicate more effectively — both with each other and with AI tools. 

“Like when you have to make a large group decision, often it comes down to someone taking everyone into one room, getting everyone to express their different camps about, for example, what kind of logo they’d like,” Zelikman continued, chortling with his team as they recalled the time-consuming tedium of getting everyone to agree on a logo for the startup. 

Zelikman added that the new model will be trained to ask questions in a way that feels like interacting with a friend or a colleague, someone who is trying to get to know you. Chatbots today are programmed to ask questions constantly, but they do so without understanding the value of the question. He says this is because they’ve been optimized for two things: How much a user immediately likes a response they’re given, and how likely the model is to answer the question it receives correctly. 

Part of the lack of clarity around what the product is could be that Humans& doesn’t exactly have an answer for that yet. Peng said Humans& is designing the product in conjunction with the model.

“Part of what we’re doing here is also making sure that as the model improves, we’re able to co-evolve the interface and the behaviors that the model is capable of into a product that makes sense,” she said. 

What is clear, though, is that Humans& isn’t trying to make a new model that can plug into existing applications and collaboration tools. The startup wants to own the collaboration layer. 

AI plus team collaboration and productivity tools are an increasingly hot field — for example, the startup AI note-taking app Granola raised a $43 million round at a $250 million valuation as it launched more collaborative features. Several high-profile voices are also explicitly framing the next phase of AI as one of coordination and collaboration, not just automation. LinkedIn founder Reid Hoffman today argued that companies are implementing AI wrong by treating it like isolated pilots and that the real leverage is in the coordination layer of work — that is, how teams share knowledge and run meetings. 

“AI lives at the workflow level, and the people closest to the work know where the friction actually is,” Hoffman wrote on social media. “They’re the ones who will discover what should be automated, compressed, or totally redesigned.”

That’s the space where Humans& wants to live. The idea is that its model-slash-product would act as the “connective tissue” across any organization — be it a 10,000-person business or a family — that understands the skills, motivations, and needs of each person, as well as how all of those can be balanced for the good of the whole.

To get there requires rethinking how AI models are trained. 

“We’re trying to train the model in a different way that will involve more humans and AIs interacting and collaborating together,” Yuchen He, a Humans& co-founder and former OpenAI researcher, told TechCrunch, adding that the startup’s model will also be trained using long-horizon and multi-agent reinforcement learning (RL).

Long-horizon RL is meant to train the model to plan, act, revise, and follow through over time, rather than just generate a good one-off answer. Multi-agent RL trains for environments where multiple AIs and/or humans are in the loop. Both of these concepts are gaining momentum in recent academic work as researchers push LLMs beyond chatbot responses toward systems that can coordinate actions and optimize outcomes over many steps. 

“The model needs to remember things about itself, about you, and the better its memory, the better its user understanding,” He said. 

Despite the stellar crew running the show, there are plenty of risks ahead. Humans& will need endless large sums of cash to fund the expensive endeavor that is training and scaling a new model. That means it will be competing with the major established players for resources, including access to compute. 

The top risk, though, is that Humans& isn’t just competing with the Notions and Slacks of the world. It’s coming for the Top Dogs of AI. And those companies are actively working on better ways to enable human collaboration on their platforms, even as they swear AGI will soon replace economically viable work. Through Claude Cowork, Anthropic aims to optimize work-style collaboration; Gemini is embedded into Workspace so AI-enabled collaboration is already happening inside the tools people are already using; and OpenAI has lately been pitching developers on its multi-agent orchestration and workflows. 

Crucially, none of the major players seem poised to rewrite a model based on social intelligence, which either gives Humans& a leg up or makes it an acquisition target. And with companies like Meta, OpenAI, and DeepMind on the prowl for top AI talent, M&A is certainly a risk. 

Humans& told TechCrunch it has already turned away interested parties and is not interested in being acquired. 

“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 said. “We trust ourselves to do that, and we have a lot of faith in the team that we’ve assembled here.”