How we build with AI

This is part one of a three-part series on how HubSpot has changed with AI. Part two covers how we grow with Agent-first GTM. Part three looks at how we operate as an AI-first company.

Everything we build at HubSpot is designed to help our customers grow. When generative AI emerged, our engineering team saw more than just a productivity tool; We saw an opportunity to build better products and deliver more value to customers sooner.

And as off-the-shelf AI tools reached their limits, we didn’t just look for better ones. We built the platform underneath. This decision worsened faster than we expected. Because all of our AI is built on a common foundation, every new feature we ship makes the entire system more powerful and customers get a more consistent experience across all the devices they use.

Today we are able to drive innovation at a pace that was simply not possible before. 100% of our engineers use AI and we have seen a 73% increase in the lines of code written by our engineers.

We didn’t get here overnight. It took three phases: real infrastructure investment and a willingness to build something that didn’t yet exist. That’s how we did it.

Phase 1: Productivity with Co-Pilots (2023-2024)

In 2023, large language models had just crossed the threshold of being truly useful in a coding context. The best solution for using AI in engineering was to start with what was tried and tested. Back then, it was all about code completion: a human writes code, and AI co-pilots suggest what comes next.

We introduced a copilot for coding and quickly achieved 30% adoption. We then pulled the incident data, compared teams that were using Copilot with teams that weren’t using it, and proved that the introduction of AI had no negative impact on the reliability of the product.

With this data, we removed the guard rails and granted access to all co-pilots. Acceptance rose to over 50% overnight. This taught us a lesson about how we make decisions. Measure, prove and then scale.

At the end of Phase 1, 80% of engineers were using AI tools. We saw a 51% improvement in development speed, meaning engineers were pushing working code to production significantly faster, and a 7% increase in lines of code updated per engineer. We’ve proven that AI can make any engineer faster without compromising product reliability.

Phase 2: Scaling with Coding Agents (2024 – mid-2025)

The next step was autonomous coding with agents. Our teams could make the tools do end-to-end tasks. The agents could read the context, write code, run tests, and troubleshoot while the engineer reviewed and controlled the situation. We strongly believed that this was the future of engineering and were fully committed to it.

The real pressure came quickly. Standard coding agents could not access internal build systems and our libraries or verify that the code actually worked in our environment. That’s why we built these agent integrations ourselves using MCP, a standard that allows AI agents to connect to external tools and systems, and deployed them to every engineer. To drive adoption, we organized events to provide dedicated space for engineers to learn, experiment, and build confidence in new tools. Agent usage went from zero to 80% in a month.

The next challenge was scaling. The engineers wanted multiple agents to run in parallel overnight and without supervision. That’s why we built an agent execution platform on our Kubernetes infrastructure. Each agent runs in an isolated container that replicates a real HubSpot developer environment. Agents compile the code, run automated tests, read error output, and iterate independently until everything works. No human intervention required.

At the end of Phase 2, 96% of engineers were using AI tools, development speed increased by 60%, and lines of code updated per engineer had increased by 48%. We started using agents to ship better products faster. But that was just the beginning.

Phase 3: Scaling with our AI platform (mid-2025 – today)

HubSpot’s platform approach to product development has always enabled us to create greater customer value. When we built reporting and automation at the platform level, we didn’t just ship a feature; We distributed this feature to all hubs at the same time. This is how innovation happens.

We applied the same logic to our AI infrastructure in Phase 3. Instead of building each agent from scratch, we built the common foundation once: how agents access data, what actions they can take, how they connect to the rest of HubSpot. Everything runs over it.

The result is that all of our agents are interoperable. They speak the same language, share the same toolsets and draw from the same context. A customer receives a consistent experience regardless of which agent they use because they are all based on the same infrastructure. And because they’re all connected, every new feature we add makes the entire system more valuable. This is something that a collection of point solutions cannot reproduce.

This was made possible by scaling the technology with AI. Today, 100% of our engineers are using AI, lines of code updated per engineer have increased by 73%, and time to first feedback on pull requests has decreased by 90%. That means less waiting time and more time shipping things that customers actually use.

Why this is important: Increasing customer value

The right infrastructure accelerates the pace of innovation. For HubSpot, every agent we build makes the platform more powerful. Every context we add to the platform makes every agent more effective. For customers, this means that the product is becoming better, faster and more connected.

What once took months now takes weeks, and those weeks translate directly into new opportunities in the hands of marketers trying to reach the right audience, sales reps trying to close deals, and customer success managers trying to retain customers. You don’t have to think about the platform underneath. You can just experience the result.