This article is the first in a three-part series on HubSpot’s AI transformation. Part two explores how we scale with an Agent-first GTM strategy, and part three dives into how we run the business as an AI-first company. Everything we create at HubSpot is designed to help our customers grow. So when generative AI arrived, our engineering team didn’t just see a way to work faster; we saw a chance to build better products and deliver value to customers much earlier. And when out-of-the-box AI tools reached their limits, we didn’t just search for stronger alternatives—we built the underlying platform ourselves.
That choice paid off more quickly than we anticipated. Because all of our AI capabilities sit on a common foundation, every new feature we release strengthens the entire system, and customers experience a more unified, predictable experience across our products. Today, we’re innovating at a speed that simply wasn’t achievable before. Every one of our engineers uses AI, and we’ve seen a 73% increase in the volume of code they produce.
Reaching this point didn’t happen instantly. It required moving through three distinct phases, making serious infrastructure investments, and being willing to create what didn’t yet exist. Here’s the path we took.
Phase 1: Productivity with Co-pilots (2023–2024)
In 2023, large language models had just become reliably effective for coding tasks. The most practical way to bring AI into engineering was to lean on what already worked. At that moment, the leading use case was code completion: humans write the core logic, and AI copilots propose what comes next…