Hi Jonas, please tell me about yourself.
Hi Robin! Thanks for the invite, very excited to do this interview with you. I’m Jonas, I currently live in Berlin and I’m the co-founder and CEO of the AI startup Beam. I’m in my 30s now, so I’ve been in the startup world for a while and have co-founded and helped scale a number of ventures in very different industries - from the railway industry to B2C e-commerce. I’m interested in a very broad range of topics in my personal life as well, but one recurring theme over the last years was the intersection between humans and AI. I’d already focused on behavioural economics and neuroscience during my studies and it’s been fascinating to see how, within only a few years, AI has become capable of taking over the first tasks only humans were previously able to do. That trend has never been more rapid or visible than now, so I couldn’t be more excited that we’re at the forefront at this with my new company Beam. Personally, I’m a bit of a maniac when it comes to work but I really enjoy life in general. I’ve always loved sports, read a lot and, since the pandemic, have also become a semi-professional pizza maker.
You are a serial entrepreneur and your latest gig, Razor Group, even achieved unicorn status in 2021. Can you tell us how you got started in the industry?
I’d say that the accelerating point of this journey was during my studies at the TU in Munich, where I focused on Business and Electrical Engineering, and especially when I joined CDTM - the Center of Digital Technology and Management, which not only sparked my interest in entrepreneurship but also led to amazing friendships and a network that I’m very grateful for. My co-founder Aqib and I both met through CDTM before we started Beam. Many of our business angels are also from that network.
As for my personal founding journey, I experienced both, success but also things not working out. The first startup I founded was as a student selling baseball caps in the US. It started as fun project by building a Shopify store on my couch in New York on a Friday night with my co-founder. But the caps got a lot of attention, and suddenly Lewis Hamilton and Paris Hilton were wearing our stuff. As I didn’t have a lot of exposure to entrepreneurship when I grew up, it was great for me to see that building a company can work incredibly well.
The first startup I joined after university, and after a detour to the consulting world at McKinsey, was Konux, which develops predictive maintenance solutions for the railway industry. I came on board in the very early days when we started with a sensor technology that ultimately didn’t work out and required us to pivot significantly. But that pivot was successful and we managed to raise additional funds with an outstanding team. As a management team, we spent quite some time in Silicon Valley and stayed at Andy Bechtolsheim’s place for a few weeks. Looking back, the time at Konux really shaped me and my former colleagues. And I remember that, back then, AI already played a big role in our daily discussions; from the Waitbutwhy post about the road to superintelligence (worth a read!) to the details of Neural Networks used to predict vertical displacement in infrastructure. So really this is when I started to slowly grasp the huge potential of AI. And yes, my last venture, Razor, was an absolutely insane and very successful ride, too. We started in Germany and after 6 weeks decided to go the UK and another 6 weeks later to the US. It was intense and challenging, and Razor eventually became a really big company with now $500m+ in revenue. But Beam really is the type of company I’ve always wanted to build. It’s very technical and I see incredible potential at this intersection of AI and humans. I can’t wait to see what the next 20 years in this space will look like.
When I first met you a year ago, I was impressed by your speed in building products. What kind of person are you, and what strength do you bring to building a company?
I’m a bit of a nerd in that I read a lot. At least an hour a day, every single day. Books, newspapers, or scientific magazines, about topics like the rise and fall of nations, to the Industrial Revolution or even trees, “The hidden life of trees” is still one of my favorite books. And I love that we have this culture at Beam as well, where my co-founder Aqib regularly sends me research papers and we discuss what potential these new developments can have for our company. The main topic is generative AI, of course, but we’re also curious about hardware, neuroscience and everything that is related to the brain. I like learning and I’m a very fast learner. So whatever the topic, I make sure I fully dive into it which usually helps me to get really good at it. And by doing that over quite some time now, it has helped me to gain a broad and deep understanding of a lot of topics and the effect of compounding growth is really kicking in. Another element I bring to the table is 100% dedication to whatever I do. I see building companies a bit like professional sports - it’s exhilarating and the best feeling in the world to achieve important milestones with a team, and to experience the moments when things fall into place and the business starts taking on more and more speed. But it can also get tough, emotionally as well as mentally. So executing is important, but recharging equally so. Only when I recharge can I perform at my best. There is actually some interesting research on recovery experiences, and I’ve experimented a lot to figure out what helps me, from measuring glucose levels to blood tests to breathwork. I don’t drink alcohol, try to avoid sugar as much as I can, go to the sauna regularly and take long walks. Most importantly though, it’s my friends who are absolutely vital to my well-being and shape me as a person. What I love about my closest circle is the absolute trust and support we have for each other. Overall, if I’m in a bad mood, tired or lose track, it has an effect on my team and I want to avoid that at all costs and be balanced, so I really believe that one of the most important aspects of performing at high speeds is recovery.
Who are the people you work with? What would you say makes a great team?
I’m convinced that the defining characteristic of any company, especially a startup, is speed. So the people that we’ve hired at Beam, are all a bit crazy, very driven, very technical but still diverse and fun. My favorite interview question is: “What makes you get up in the morning?” and people who work at Beam all have this insane inner drive. We hire primarily for cognitive ability, passion, and a certain experience which obviously helps a lot.
The advantage of being a serial founder is that I’ve developed a better understanding of what personalities and skills we should look for, depending on where we are as a company. Right now at Beam, we’re not only hiring new talents but adding people to the team who we’ve worked with before, sometimes years ago. For example, Natalie joined us from AWS, where she led Executive Marketing for EMEA. But we already worked together back in Konux times. So it’s a bit like coming home and working together just feels very natural. Working with Aqib is also simply amazing. We’re different in a lot of aspects but share very similar views when it comes to aspiration, work ethics, and performance. It’s almost like there is an invisible connection that holds us together and guides us. It’s amazing how often we come to similar conclusions independently but also how fast we can both adapt and iterate if we have different views. Working with him just feels very natural and unbelievably fast.
Part of our team is based in Pakistan, as Aqib was born there. This has benefits in terms of costs but more importantly: We can hire the most impressive talent in a country of 220 million people. It’s a massive advantage for us and, as a consequence, we have absolute superstars in the team. I’m convinced that a great team needs to be diverse and bring in different experiences, on top of being insanely driven, very smart and very adaptive.
One aspect of great teams that is often a bit neglected is simply: Who shouldn’t be part of it? When we build a company, we need to grow the team and get our hiring right. But we also have to make sure that we put people in the correct role where they’re at their best, and that we recognize early if we’re not the right company and work environment for them to thrive. Individual performance follows a power law, and some people are contributing 100x more than others. They do things while others fail to get it done. We need to optimize for the 100x.
Now all of this is good but not excellent, if there is no strong glue that holds the team together and guides all of us. That’s through our vision, that’s because we have joint goals and it’s because of our shared experiences, values and also the fun we have as a team, in short, our culture. We spend too much time working for it to not be a lot of fun, too.
What exactly does your current company, Beam, focus on? What is the product that you’ve been building since you founded Beam last year?
At Beam, we develop generative AI agents for businesses. Right now, our customers primarily use these agents to automate internal processes, to support their existing workforce and significantly increase the productivity of their teams. Our Beam AI agents use LLMs, have their own databases, integrations as well as functions to get work done. They can take over complex workflows by chaining different steps together, similar to the way humans would approach a larger problem by completing one step after the next. AI agents are like virtual co-pilots that are very versatile and support you in your day-to-day. They can write emails on your behalf, analyse data and create spreadsheets or help you understand complex information. These agents also create code and variables, which means they can even interact directly with other business applications. So while ChatGPT provides information, AI agents can now plan complex tasks and also get actual work done. They fit seamlessly into an organization’s work environment and help teams automate repetitive tasks and free up time for more impactful projects. Users can set up agents in two different ways: They can use templates (pre-trained agents) that we’ve already created or they can use an orchestration AI that helps with setting up these agents - this currently works for up to three-step tasks but we are working on expanding the scope as fast as possible. The overarching goal of AI agents is always to help employees get work done faster and better as well as figuring out how to solve the task best.
What does this mean exactly? How are teams, or individual employees, using these AI agents in their daily work?
While developing the technology and training the models our AI agents are based on, we noticed that even the way they approach prompts and interact with employees already adds value. We put strong emphasis on allowing the user to control every individual step while the AI offers “food for thought”. This means it’ll dive deeper into a complex task and highlight aspects you might not have thought about. Say, for instance, you ask it to create a sales presentation without providing any additional information. The AI agents will then follow up with clarifying questions about your USP, value proposition, personas, technology, or even your brand identity. If you provide more context, it will come up with a better, tailored solution faster. In the process it will guide you and give you input and ask questions. Now you can ignore that input, but sometimes this input already helps you to think differently about the task you’re about to solve. Either way, the AI agent will create a structure and content for your presentation. And, if you give it access to Google Slides, it will build the slides directly in your workspace using the code and variables it has created. Unfortunately, it prefers to use Arial as font, but luckily this is something you can easily change as a user.
The use cases we are able to solve with these AI agents are literally expanding every day. They get more complex and we can now add integrations faster. Sometimes the output is not great and sometimes it’s simply astonishing. We’ve also starting to combine multiple agents, which is super exciting but very early.
One aspect that is important to many of our customers is that we pre-train agents for workflows or repetitive tasks that their teams encounter in their day-to-day and that we offer integrations with their main business applications. This means we build these pre-trained agents for very specific use cases. This can take the form of an account agent, a sales prospecting agent or a fundraising agent, for example, and it helps our customers to get started with only a few clicks. Unless they want to partner with us to create even more tailored solutions for their company.
What did your early sales calls and customer meetings look like? Did people get the concept of generative AI and the main value for their business immediately?
In the beginning, it took us a bit to figure out the best “starting point” for a generative AI pitch and we’re still spending a lot of time listening to our customers and understanding where they’re at in their own journey. But the exciting thing is that we’re having more and more conversations, in which senior executives tell us they’re already working on an “AI-first” strategy for their company. These customers grasp immediately where Beam could fit into that strategy and it’s always a nice surprise when our meetings become very technical and we suddenly find ourselves in the middle of an in-depth discussion of the latest AI research and studies with a CxO. Overall, there is a starting point in every company, sometimes it just takes a bit longer to figure out the best way to start with. Theoretically, there are many cases in which you could automate almost everything with AI. Practically, your options to get started might be much more limited. We had to find the right balance between being very broad or focusing on narrow use cases and it’s something we’re still optimizing for. Generally, we start relatively broad in use case identification but have a laser focus to show value when it comes to the first use cases. Further down the funnel, we’re also increasingly talking about data privacy and security, and addressing any concerns customers might have. Building systems that are fully compliant with GDPR and SOC2 luckily has become a bit easier in the past years. It is crucial for our customers that we comply with all regulations and guarantee their data's security and privacy, despite needing to be super fast and agile when conducting pilots. Our approach to building agents is very flexible, so depending on the customer’s needs we can actually use different modules and providers.
You’d already started worked on Beam before the current hype around ChatGPT and Generative AI started. How do you see this trend developing? And what role does AI play in your own day-to-day life at Beam?
The main hypothesis underlying everything we do at Beam is that the power of AI is growing exponentially. That sounds actually simple, but we as humans are very bad at understanding exponential growth. Let’s take knowledge doubling time as an example. Back in 2010, medical knowledge doubled every 3.5 years. In 2020, we saw doubling time go down to 73 days. So if I learned all of the medical knowledge available in the world in 2010, by 2015 I would know 37% of the knowledge then available. This is tough, but it is something we can still comprehend. Now if I learned everything in 2020, due to the decrease in doubling time by 2025 I would only know 0.000003% of all knowledge (assuming constant growth rates). It’s something we can barely comprehend. So if we reverse this, the power that AI will have in a few years is almost impossible to comprehend, unless we look at it super rationally.
When we started Beam we talked about it but didn’t fully understand this speed if I’m being honest. Even though we as a company operate very fast and agile, we actually underestimated this insanely fast development ourselves. So at a certain point we sat together and said, well, let’s forget everything we thought was possible and be much, much, much more radical in our thinking. Let’s think from the end, let’s think how we can be exponential ourselves. And this is where we changed a lot and are now starting to see massive gains from.
AI has the power to disrupt everything humans do, especially in the corporate world, and we as a company see a lot of white space ahead of us as a consequence. Yes, there are now LLMs and generally generative AI but there are many things not solved yet that will shape the future of AI and will rewrite the rules of the game. What we need to figure out, in order to leverage this immense potential, is where the limiting factors to this exponential growth of AI are. And then give AI the capacity to solve those. LLMs will continue to increase but AI can almost create everything already. If it can’t yet, it’s often a matter of months, not years. Toolformer, integrations, individual context and self-learning (multi) agent systems will drive this next wave. We see that the combination of context and smart chaining can be insanely powerful. At Beam, we’re building for this world - the “future of work” if you will.
As a consequence, we as a company need to become “AI-native” as well. Essentially an evolution of a “digital-native” company. This means that everything we do internally is driven by AI, everyone at Beam uses AI in their daily work, and we become an organization where humans are not the limiting factor anymore but help to shape the future with the use of AI.
I think right now, in 2023, it’s crucial to completely rethink the way you work. If you use AI to the fullest, you will be a minimum of 5-20x more productive, depending on what you do in your job. For other jobs it will make a 1000x increase possible - it’s a lot of potential power. If you don’t use it, you will fall behind because others will use that power to their advantage.
So what you’re saying is that the future of work, maybe even in a few months, will look very different from today’s corporate jobs. What do you tell people who are still skeptical and who’re maybe even afraid to lose their jobs?
The harsh reality is that yes, some types of jobs will disappear. But jobs have always disappeared for as long as humans have evolved. And new ones have always been popping up. This time, the new jobs are Prompt Engineers, Generative Agent Designers or others. Humans are incredibly adaptive - there have been countless tech disruptions in the past and still, unemployment rates are very low. If we look back, history teaches us that in every major wave, people were afraid of losing jobs, but eventually prosperity increased. It’s easy to paint doom scenarios with AI, and maybe in Germany we tend to do so more than others, but there are many bright scenarios of a future where humans use AI as well. In my view, the only path is: we have to shape the future with AI proactively. AI is there to stay and it will get better either way. We can try to regulate it and most likely will to a certain extent, but I think it’s much better to take an active role and shape the future with AI than to be afraid of it.
Also, I believe there is a lot of work that we do every day that is not a lot of fun. So I’m really looking forward to having all my presentations and emails automated, as well as having AI by my side to think through things faster and more comprehensively. There are a lot of new workflows emerging that previously weren’t possible. An easy example: I create voice notes in our tool, have them transcribed, then use functions to optimize the text or get feedback from agents (e.g. “ask Elon whether this is visionary enough”) and only then share it with my team. It’s much better than just directly sending a potentially confusing voice memo. These kinds of “new” workflows allow us to do better work and become better humans in a consequence. This is a small example, but what about these countless Excel sheets and endless preparation of presentations? When we now do all of this automatically, we can potentially focus more on the human connection, which is amazing.
On a macro level, we’ll need to see how this develops exactly - AI will be celebrated and hyped but there will also be phases of disillusionment, much like we saw with autonomous cars. But we can learn from the past and try to think about why autonomous cars are not ubiquitous yet, and then use these principles for building products with AI that don’t run into similar obstacles. For now, I don’t think that we really understand what AI can do eventually. We’re barely scratching the surface.
Now that we’ve talked about the future of AI, and the future of our work life as we know it, let’s go back to Beam. What’s next for your startup and where will it lead to?
When we started Beam, we essentially set two goals: Serving 100M users, and becoming a $10bn company. We still have quite some distance to cover to reach these goals, of course. But I’m growing more confident every week, and we are seeing the path ahead of us more clearly every day. I’m a very rational person, so I’m not “hoping” to get there. Instead, we need to keep optimizing our approach, and our own work, every single day to get there and then it will happen eventually. It needs to be a logical consequence of what we do, factoring in that there are always things we won’t see coming, and other challenges we can prepare for. Right now we feel that we are very much on the right path. We’ve been very quietly building in the past months but will go now a lot more in the offensive, so you’ll hear more about what we do soon.
Our short-term objectives are simple too: First, continue to build an amazing product while staying obsessively focused on what our customers need. Second, raise enough money and find amazing people, both employees as well as investors, to work with. Third, continue to build Beam as an AI-native organization that grows exponentially, fueled by our very own AI agents that learn on their own. This will make us as an organization faster than anybody else and it will allow us to serve our customers better.
#8 Jonas Diezun - Building Beam
Hi Jonas, please tell me about yourself.