Leya's $100M lesson: Skip fine-tuning, bet on better models

Max Junestrand, CEO of Leya, recounts building an agentic OS for lawyers from a Swedish archipelago startup to $100M ARR in 18 months. The key bet: don't fine-tune models, assume they'll improve, and focus on delivering value to the market. This allowed Leya to outpace competitors who spent millions on custom models.

The company's sales freeze for six months was a counterintuitive move that paid off. By pausing sales to fix product reliability before scaling, Leya avoided burning trust with law firms where one failure ends the relationship. This operational discipline, not technical wizardry, created the foundation for exponential growth.

Cultural tension is central to Leya's story. Swedish Jantelagen (humility) conflicts with the aggression needed to build a global enterprise. Junestrand's solution: a deliberate cocktail of US, European, and Asian cultural values, reinforced by requiring all employees to onboard in Stockholm.

The most practical signal for engineering leaders: Leya's shift from reactive to proactive agents. Connecting AI to data triggers so it acts without human prompting is where the real productivity multiplier lies. This isn't about better models; it's about workflow architecture.