Sana: Harnessing the power of AI to curate your company’s internal knowledge

Sana founder & CEO Joel Hellermark is helping companies unlock the utility of their information so employees can work faster and think bigger

From Claude to ChatGPT, the most powerful large language models (LLMs)  have sparked a once-in-a-century revolution in technology. But these models are only as good as the information we give them and the context we provide about the tasks we need to accomplish.

For organizations to truly harness the power of these models, here at NEA we believe you need a way to connect them to the vast repositories of knowledge that exist in disparate places — everything that sits in Notion, Outlook, or Google Drive — and create tools for people to manage, curate, and update this knowledge base. Sana, a Stockholm-based company, is doing just that, transforming the way companies organize, access, and build their internal knowledge, so employees can work faster and think bigger — whether they’re in sales or R&D. Along the way, Sana is turning Stockholm into a hub of AI research and innovation. 

Scott Sandell, Philip Chopin and Luke Pappas are the NEA partners working closely with Sana, and Luke recently spoke to Sana founder Joel Hellermark about why he created the company, how it can enhance organizational performance, and the importance of making knowledge easily accessible — both at your company and around the world. 

Luke: What sparked your interest in AI systems and when did it begin?

Joel: I was 13 and developing a program to play chess with my friends. I was intrigued by this idea that you could take these concepts that were historically done by humans and create rules that subsequently the computer could act on.

Around that same time, Stanford started publishing these online courses. I ended up binging most of them, and that’s ultimately how I learned to code. I was really obsessed with autodidacticism, the idea that you could just teach yourself anything. I read a lot of biographies: Edwin Land and Buckminster Fuller and Da Vinci and Michelangelo. And I found their interdisciplinarism really intriguing. I thought if I could teach myself a lot of different disciplines, I could do meaningful work. Something that I’d be proud to tell my grandchildren about.

Luke: How does Sana change the way we use AI?

Joel: Buckminster Fuller, a hero of mine, was out for a walk with another very clever guy named Albert Einstein, and they started talking about the most important problems they should be working on. They concluded that the most important problems are the meta problems, the problems that if you solve them, they cascade and solve a lot of others.

If you look back in history, every time we’ve changed access to knowledge, everything has changed for the better. Sana is the first system that really has indexed all knowledge that sits within a company and makes that universally accessible for every employee. So whether you’re a researcher at Merck accessing clinical data or you work at Svea Solar developing solar panels, you can have the knowledge that ultimately enables you to achieve that mission.

Luke: How has that system helped companies and people in the real world?

Joel: During Covid, our entire team was locked down in Stockholm, and we really wanted to find a way to help people. We set up a conference call with the presidents of a lot of the different hospitals. They said the biggest challenge was actually training all of these new folks who were showing up in the ICUs and onboarding volunteers and upskilling their nurses. This was done in legacy systems. If you were getting onboarded, you got sent 20 PDFs that were completely scattered, and you had to just browse through them. 

What we decided to do was use AI to create a very personalized program for every single nurse, so we could upskill them in a matter of days rather than months. Ultimately that program supported over 150,000 nurses at over 2,000 hospitals.

Luke: So Sana is an application layer company and people will say, “Oh, it’s a wrapper around GPT or it’s a wrapper around this model.” How do you respond to that?

Joel: I’m very excited to be a wrapper. A lot of our customers, they’ll just care about speed and accuracy. Whether it’s Claude or GPT-4 or Mistral. 

The legacy models are becoming incredibly capable. They have an IQ of 155. But we really aren’t using much of the models’ capabilities. And they’re only as good as their context.

What we’re working on now is embedding reasoning into this model. So they don’t just output their first word that comes to mind, but they lay out the plan when you ask it a question, they execute that plan, they use any relevant tools and knowledge that they need for that plan to self-reflect on whether the answer they generated was any good or if they need to run another loop. And when you do that, you can get GPT-4 to go from 67% accuracy to 96% accuracy on a lot of tasks.

Luke: Back when you were a 13-year-old programmer, did you know that being a founder would be your life’s work?

Joel: I don’t believe in the concept of founders, and that’s something I want my team to know. A company’s continuously founded, and you have so many people who joined Sana throughout the years that I would definitely consider founders of the company.

Luke: How do you think about building that team?

Joel: We want to find people that have fire in their eyes, who get very excited whenever you give them that impossible task. Every six months we reorganize the entire company around our top missions, what we’re looking to achieve. There’s not a single person at Sana that’s not working on the most important thing.

Luke: Stockholm, Sweden, hasn’t always been known as the AI capital of the world. But you’re changing that perception through Sana and your AI Summit events.

Joel: We’ve always wanted to bring people from all different aspects of AI under one roof to discuss what’s next and inspire interdisciplinary discussions. And that’s what we did at the summit. 

We were excited to bring them to Stockholm, because we think we’re well positioned to create a great ecosystem there. When it comes to product and engineering, Stockholm is second to none. Swedes are also very humble, and we think that’s core to Sana’s DNA.

Luke: Finally, what made you want to work with NEA?

Joel: We talk at Sana about pragmatic dreamers. What really inspired me was you’re really trying to build something generational. And I was just struck by how you went deep on the product, but also deep on effectively everything else while being extremely ambitious in the long term.

Luke: That’s amazing. I think when you meet a special company and a truly special founder, you do everything you can to partner with them.