20VC: 15 Term Sheets in 7 Days and Choosing Benchmark | Harvey vs Legora: Who Wins Legal and How to Play When You Have $600M Less Funding | Are AI Models Plateauing Today | Building a 9-9-6 Culture From Stockholm with Max Junestrand
Summary
The episode of The Twenty Minute VC features Max Junestrand, co-founder and CEO of Legora, an AI-powered legal tech startup based in Stockholm. Max's unique journey from a promising multi-million dollar gaming career to founding an AI legal startup highlights a blend of competitive gaming skills, strategic thinking, and entrepreneurship. The discussion covers Legora's approach to leveraging large language models (LLMs), particularly GPT-based models, to build collaborative AI tools that empower lawyers, contrasting with legacy AI systems built on older models like BERT. Max emphasizes a significant gap between the capabilities of modern AI models and their current product usage, suggesting vast opportunities lie in building application frameworks, including swarms of LLMs executing thousands of parallel calls to enhance output quality, despite increased computational costs. One critical theme is how AI is transforming the legal industry, shifting from traditional billable hour models towards project-based pricing fueled by AI efficiency gains. Legora's strategy involves deep partnerships with top law firms, embedding AI software within firms rather than purely competing on service or legacy product offers. The episode also covers the competitive fundraising landscape, with Legora raising $120 million from top investors like Benchmark and IVP, efficiently scaling with far less capital than competitors such as Harvey which raised $800 million, and details on the rapid closing of multiple term sheets. Challenges discussed include managing expectations around AI capabilities, overcoming conservative legal culture's technology gap, and transitioning spend from human labor to technology budgets to realize AI's full disruptive potential. Additionally, Max shares insights on startup culture, including adopting a disciplined 9-9-6 work mentality adapted in Europe, and remote operational challenges, especially being outside Silicon Valley during Y Combinator. The episode intertwines broad AI industry debates on whether foundational models are plateauing or just beginning while stressing that ecosystem and application layer innovation will drive future value. Overall, it combines founder storytelling, AI technical depth, operational scaling, fundraising tactics, and industry transformation, offering a rich, nuanced view of legal AI startup building today.
Key Takeaways
- 1Max Junestrand's background in competitive gaming, including strategic games like World of Warcraft and Dota, fundamentally influenced his entrepreneurial mindset and approach to building Legora. His gaming experience fostered complex problem-solving skills, rapid decision-making, teamwork, and resilience, which translated well into the startup environment. Additionally, early work creating statistical models for esports betting introduced him to data-driven approaches underpinning AI innovation.
- 2Legora made a decisive technological pivot from legacy AI systems built on early models like BERT and proprietary legal document platforms to a GPT-based architecture for its legal AI tools. This bold move involved discarding cumbersome and poorly performing legacy infrastructure and building the product anew around generative large language models to deliver clear lawyer benefits.
- 3There exists a substantial and underexplored gap between the theoretical capabilities of state-of-the-art AI models like GPT-4 and their practical deployment in products, with many current applications utilizing less than 20% of model functionality. Major innovation opportunities lie in developing frameworks, orchestration tools, and ecosystems around these models rather than in raw model improvements alone.
- 4Legora employs a strategy of running swarms of large language model API calls in parallel, sometimes in the thousands, to improve output quality for complex legal tasks like due diligence by iteratively refining and aggregating outputs. While this method drastically enhances accuracy and depth, it increases computational costs by roughly 100 times compared to single API calls.
- 5Despite the high compute costs introduced by swarms of LLM calls, these expenses remain substantially lower than the billable hourly rates of top-tier lawyers, meaning AI-driven legal tech solutions still hold massive economic leverage to disrupt the legal services industry. However, model provisioning is rapidly commoditized, making compute providers the primary economic winners, while startups must focus on application-layer innovation to capture value.
- 6The legal market presents a monumental opportunity for AI companies because while the legal software market is approximately $20 billion, the legal services market is near $1 trillion. This large disparity shows most legal work remains labor-intensive, underscoring AI’s potential to transform how legal services are delivered, not just sell software.
- 7AI adoption in legal services critically depends on the industry's ability to reallocate budget from human labor toward technology spending. If spending shifts accordingly, AI tools can truly disrupt cost structures and operational models; if not, AI's impact may remain marginal. This potential budget transition signals an exciting prospect for founders and venture capitalists.
- 8Legora’s strategic approach hinges on becoming 'software-powered' by deeply embedding AI tools within top law firms and empowering lawyers rather than replacing them with pure service models. This hybridization balances leveraging legal expertise with cutting-edge AI augmentation, viewing software as fundamental to building a multi-billion dollar legal tech enterprise.
- 9The episode provides compelling evidence of the vast challenges and opportunities in AI startup fundraising, highlighting Legora’s experience closing 15 term sheets in seven days and choosing Benchmark as lead investor. Despite significantly less capital than competitors, Legora's efficient capital deployment and strategic fundraising enabled rapid product development and scaling. The episode details differences in VC behavior between Europe and the US, the importance of investor fit, and balancing dilution with runway.
Notable Quotes
"It explains a lot. Because I was juggling school on the side, and the people I was competing against were not."
"First, Max turned down a multi-million dollar career in gaming to build Legora. Second, he's raised from the best in the business, including Benchmark and IVP, to name a few. And third, he's scaled the firm with one-sixth of the capital of his closest competitor, Harvey, who've raised a reported $800 million, while Legora has raised just $120 million."
"Let's just delete all of it and let's build on GPT and let's like build a version of it that at the end of the day, lawyers are benefiting from."
"The gap from the BERT models to GPT was the biggest gap ever. And I think the gaps since then have been much more incremental. I actually think the biggest things allowing for new products to be built is not the models themselves, but the frameworks around them."
"So I actually think that the product to model gap today is massive. Where the products are today, I don't think we've utilized 10, 20% of what the models can do. We have a huge sort of gap and way to go. And I think we could, just given more time, with the models we have today, given no incremental improvement in them, still have years and years to years to build on."
"We've started playing around with this idea of swarms of LLMs being thrown at a problem. So swarms of LLMs are another problem is when you take Grok, Anthropic, and OpenAI model. Well, and you throw, instead of doing one API call, you do 100 or 1,000. And the quality of the output is drastically improved."
"It literally is 100x. So the unit cost is the same for every increase. Yeah, basically. Wow."
"The compute providers are for sure winners. But I also think that there's a huge piece of the cake in the application layer. And in particular, there's a cake in the application layer, which is not only on the software side, but on the service side."
"The legal software market is 20 billion. Legal service market is a trillion. The way you should think about it is not how do you play for the software market? It's how do you work with the best firms in the world to play for the service market?"
"You are paying hundreds and hundreds of millions in legal fees and you're thinking about how do I work with my law firm to get the most amount of value here? The opportunity for the law firms are massive, but they have to be forward-leaning and say, all right, in this new world where there's a new equilibrium and there's a new demand and supply function for our services, we need to get majority market share and we need to own the conversation because right now, I'm finding that it's sometimes quite reactive."
"The law firms and the best attorneys in the world will go software-powered, and we help them do that. For me, it's like, if you want to build like a hundred billion dollar business, you have to be software-powered."
"We had pre-access to GPT-4, which was an incredibly unfair advantage in some ways. But we started in a much smaller market. We started in the Nordics. And I quite quickly realized that nobody was going to take three software guys very seriously in this world of law."
"Somebody told me it's harder to break into his calendar than the Fort Knox. But somehow we managed to do it. I go to their offices, we sit down like this for 30 minutes and I realize we're sort of a pre-seed company, we have no product, we have nothing to show for, so I need to make the impression that AI is going to be a very important part of their business and that I'm a guy that he's going to want to work with."
"Dude, I don't know law, you don't know AI but this is going to be a thing and we're going to need each other."
""Every day, I’d sit next to somebody new in the lunch canteen and sort of bother them... Hey, we’re building this demo, do you want to look at Leia? And they’re like, ‘Oh Leia! Oh yeah, you guys are the AI guys, right?’ And I’d show them and watch them play around with it, see what they try to do. Very quickly you realize these are the buckets of things people want to do." This quote from Max Junestrand illustrates the grassroots approach Legora took to gather user feedback by organically engaging lawyers in informal settings. It conveys how direct observation helped the team identify core categories of AI uses and shaped their product development."
Episode questions
How did Max's background in competitive gaming contribute to his approach to building Legora?
Max’s competitive gaming experience, especially with strategic games like World of Warcraft and Dota, helped develop critical skills such as complex problem-solving, rapid decision-making, and teamwork. These innate qualities translated well into the entrepreneurial challenges of founding Legora, fostering discipline and a resilience mindset. Moreover, exposure to building statistical models for esports betting introduced him to data-driven approaches and predictive analytics, foundational to AI innovation. This intersection exemplifies how non-traditional backgrounds can provide unique startup advantages, particularly in AI domains demanding strategic thinking.
What role did the pandemic play in shaping the founding journey of Legora?
The pandemic created an unusual social and operational context where remote working and virtual interactions became the norm. Max met his co-founders at a volleyball event held during the pandemic, highlighting the increased importance of informal gatherings and flexible community structures even amid restrictions. Additionally, the Swedish education system’s remote learning and exam flexibility facilitated Max completing parallel degrees during this period, supporting his diverse skill acquisition. The pandemic thereby acted as both a constraint and enabler, reshaping networking opportunities and educational pathways that contributed to Legora’s founding.
Why is Max skeptical about recent model improvements being as impactful as the jump from BERT to GPT?
Max notes that the transition from BERT to GPT was a monumental shift in AI capabilities, setting a high bar for subsequent improvements. Since then, he observes that improvements have been incremental, making ecosystem and application-layer innovations the real drivers of new product success. His point is that raw model advances alone no longer guarantee revolutionary products; rather, developers must build robust frameworks around these models to unlock scalable and usable solutions. This reflects a broader AI industry maturation where infrastructure and tooling become central.
How did Legora’s team approach the integration of AI into legal technology compared to legacy legal AI players?
Legora’s team chose to abandon legacy AI tech based on early, poorly performing models like BERT variants and cumbersome proprietary systems. Instead, they rebuilt their platform from the ground up using GPT-based architectures, leveraging the superior language understanding and generative capabilities of modern LLMs. This approach allowed them to create collaborative AI tools that meaningfully benefit lawyers by focusing on usability and practical utility rather than incremental enhancements of outdated platforms. The strategy signifies a break from the past, embracing generative AI's potential to radically transform legal workflows.