What AI Companies Get Right About Culture
Reading Time: 12 Minutes
Image created by Superhumxn team.
1%
What separates the top-performing 1% of AI companies from the rest? Many would be tempted to say the product, the talent, or even the funding. Culture rarely comes up first, if ever.
But the data tells a different story. Deloitte research tracking companies over eleven years found that organisations with strong, strategy-aligned cultures showed up to a 50% performance advantage over those that didn't. Gallup's 2025 State of the Global Workplace Report found that disengagement cost the global economy $438 billion in lost productivity in 2024 alone.
In the age of AI and open-source, technology can be copied, but culture is much harder to replicate.
AI leaders like Anthropic, OpenAI, and Nvidia figured this out early. How a team communicates, makes decisions, and adapts is structural operational considerations, not soft skills. Even in a tight-knit start-up team, culture is a strategic tool. Here is what eight of the most forward-thinking companies are doing differently.
What can I expect from a subscription?
Real-world strategies for navigating the future of work.
AI AGENTS AT WORK
Everything you need to know about how AI agents are changing teams and ways of working.
SUPERHUMXN IRL
Real insights from investors and operators.
MEMOS ON PEOPLE, CULTURE & AI AT WORK
The latest use cases and strategies for change and transformation.
1. Anthropic - Operational safety
💰 Valuation: $61.5 billion (2025)
📍 Known for: Backed by Amazon, Google, and others. Creator of Claude AI.
While most of the AI industry optimises aggressively for speed, Anthropic has built its company culture around the opposite: restraint, governance, and strategic decision-making.
A great example is their Responsible Scaling Policy. Rather than functioning as a piece of branding, the framework outlines internal capability thresholds and identifies when additional review or safety processes should be triggered before development continues. In October 2024, Anthropic went further, appointing Jared Kaplan as its Responsible Scaling Officer, a dedicated role with formal authority to determine what safety assessments and precautions are required before any model ships.
In an industry where speed is often treated as the primary competitive advantage, slowing down is a notable differentiator. Many companies pay lip-service to ethical and responsible AI, but Anthropic appears to live by it. Their principles naturally slow down decisions, but the strategic long-term foundations are likely to outlive short-sighted competitors.
2. Nvidia - Learning at hyper-speed
Learning At Hyper-Speed
💰 Valuation: $2.5 trillion (2025)
📍 Known for: Publicly listed. One of the world's most valuable companies, providing hardware that powers much of today's AI.
“If you're not embarrassed by the first version of your product, you've launched too late.”
Jensen Huang has repeatedly described Nvidia as a company built around learning speed. The above quote is less about recklessness than iteration. But what makes Nvidia's culture interesting is how that philosophy is reinforced through their ways of working day-to-day.
Reportedly, Huang has around 60 direct reports and does not do one-on-ones with most of them. Instead, feedback happens in group settings where everyone is in the room at once. His reasoning: "Feedback is learning. For what reason are you the only person who should learn this?" Mistakes are shared openly across the leadership team under an internal motto: nobody fails alone.
He also avoids long-term planning, believing this stifles adaptability. In an industry where the competitive landscape can shift within a single product cycle, that is a deliberate cultural choice, not an oversight.
The result is a company where information flows without gatekeepers, learning happens publicly, and everyone moves fast in the same direction. Under Huang's leadership, Nvidia has achieved a 10-year compound annual growth rate of 70.82%. Culture is not incidental to that number.
3. OpenAI - Cross-team learning
💰 Valuation: $300 billion (2025)
📍 Known for: Developing ChatGPT. Microsoft also backs them.
OpenAI builds fast. When building Codex, they ran three to five teams working on the same problem in parallel, then consolidated at the right moment. The entire product launched in seven weeks.
That speed comes from how the organisation is structured. OpenAI has moved away from traditional functional departments toward fluid, project-driven teams where researchers, engineers, designers, and product people work as a single unit rather than passing work between separate departments.
A former engineer described the culture as having a strong bias toward action, "you can just do things", with the best ideas winning on merit rather than through seniority or politics. Accountability is maintained through a DRI framework where every project has a Directly Responsible Individual, which keeps decision-making speedy without creating ambiguity about ownership. The result is an organisation that learns through doing, where feedback feeds directly into the product faster than most companies can plan a sprint.
4. Google DeepMind - Curiosity as infrastructure
💰 Estimated Value: While part of Alphabet/Google, its speculated valuation is anywhere between $3 billion and $700 billion if independent.
📍 Known for: Leading AI research and for breakthroughs like AlphaFold.
AlphaFold won Demis Hassabis and John Jumper the 2024 Nobel Prize in Chemistry. It also solved a problem that had resisted biology for 50 years. It did not come from a single discipline.
DeepMind brought together structural biologists, physicists, and machine learning researchers to work as one team on one problem. The biologists had to learn how the AI model worked. The engineers had to develop enough understanding of biology to make the model useful. DeepMind says that picking the right problems then giving teams enough time to build a real working solution is the key to success.
That patience, protecting long-term research from short-term commercial pressure, is what most organisations struggle to hold onto as they scale. DeepMind has cemented this into the organisation.
5. Perplexity AI - Simplicity=speed
💰 Valuation: $18B+ (2025)
📍Known for: A hyper-growth AI startup known for its real-time answers
Perplexity processed 780 million search queries in May 2025, up from 230 million in mid-2024. It did that with a team of under 50 people.
That ratio is incredibly telling. Features go from idea to production in days, not quarters. Each engineer carries enormous individual impact because there are no layers of coordination slowing decisions down. The company competes directly with Google Search and moves faster precisely because it has refused to build like Google.
Perplexity is not lean by accident, but rather by design. Fewer layers means information moves faster, and product adjustments happen in real-time. As most organisations scale, complexity becomes the biggest threat to speed, and Perplexity's culture resists precisely that.
6. Notion - Scaling knowledge
Culture Built On Community And Code
💰 Valuation: $10B+ (2025)
📍 Known for: Productivity platform blending docs, wikis, and AI
Notion has 100 million users and an estimated $300 million in annual revenue. It got there partly by taking documentation more seriously than most companies do.
When a product area recently transferred ownership between teams, the process took several months. Engineers hosted workshops, flew in for in-person pair programming, and ran sprints specifically to write documentation before handing anything over. One employee described the approach as: "we know when to slow down so that we can move fast over the long term."
At Notion, knowledge does not live in one employee's head or one particular team's inbox. It gets written down, shared, and kept up to date. This continuous documentation supports scale.
7. Hugging Face - Open collaboration
💰 Valuation: $5.5 billion (2025)
📍 Known for: Language model APIs and enterprise LLMs
Hugging Face started as a teenage chatbot. When the team open-sourced the model behind it and invited a small community of friends to contribute features on an open platform, something unexpected happened. The community grew beyond expectation. Researchers, developers, and companies started building on top of each other's work publicly.
That early choice shaped everything about how the company operates today. Hugging Face now hosts over 2 million public models, 13 million users, and more than 500,000 public datasets, with 1,000 to 2,000 new models uploaded every single day. Their 250-person internal team did not produce all that. A global community did, built on infrastructure Hugging Face originally created.
CEO Clément Delangue describes the mission as empowering everyone from individual researchers to large organisations to host, share, collaborate on, and use models for their own purposes. The platform works because it does not try to own innovations, but instead it enables them.
8. GitLab - Systems reduce friction
💰 Valuation: $10B+ (2025)
📍 Known for: Productivity platform blending docs, wikis, and AI
GitLab has 2,300 employees across 60 countries and has never had a physical office. But how?
The company runs on a 2,700-page public handbook that covers everything from HR policy to engineering guidelines. Meetings require a shared agenda and need to be justified by the organiser. Zoom calls are recorded and posted publicly.
GitLab calls this approach "handbook-first." What that means in practice is that a new hire in Seoul and a senior engineer in Berlin are working from the same information on day one. Nobody is waiting on a reply from someone in a different time zone to know what to do next. The coordination happens through the systems, not through the people.
Perspective
What these companies just get
The technology these companies create gets most of the attention, but the way they work is more interesting. One bets on openness, another on restraint, another on radical flatness, another on documentation that runs to 2,700 pages.
What connects them is harder to copy than any individual practice. They have all figured out, in their own way, that how a company works on the inside shapes what it can do on the outside. Communication, knowledge, decision-making, feedback are all strategic considerations and not things to outsource to HR.
Culture is not soft, and in the age of AI it directly determines the future value of any given company.
