This is part of a three-part series on how HubSpot evolved with AI. The second part covers how we are growing with Agent-first GTM. The third part is how we work as an AI startup company.
Everything we build at HubSpot is there to help our customers grow. So when productivity AI came out, our engineering team didn’t just see a productivity tool; we saw an opportunity to build better products and get more value into the hands of customers faster.
And when off-the-shelf AI tools reached their peak, we didn’t just want better. We built a platform under it. That decision grew faster than we expected. Because all of our AI is built on a shared foundation, every new capability we deploy makes the entire system stronger and customers get a consistent experience across everything they use.
Today, we are able to innovate at a speed that was not possible before. 100% of our developers use AI, and we’ve seen a 73% increase in lines of code written by our developers.
We didn’t get here overnight. It took three phases, real investment in infrastructure, and a willingness to build what didn’t exist yet. Here’s how we did it.
Phase 1: Production with Co-pilots (2023-2024)
By 2023, large model languages had just crossed the threshold of being really useful in a coding context. The best solution for using AI in engineering was to start with what was proven. Back then, it was code completion: a person writes code, and AI copies suggest what comes next.
We released a copy of the codes and reached 30% adoption quickly. We then extracted incident data, compared teams that used the pilot to teams that did not, and proved that AI adoption had no negative impact on product reliability.
With that data in hand, we removed the guardrails and gave everyone copilot access. Acquisitions exceeded 50% overnight. This taught us a lesson in how we make decisions. Measure, prove, and measure.
By the end of Phase 1, 80% of developers were using AI tools. We saw a 51% improvement in engineering speed, meaning engineers were sending working code to production much faster, and a 7% increase in lines of code reviewed per engineer. We’ve shown that AI can make every developer faster without compromising product reliability.
Phase 2: Scaling Through Code Agents (2024-Mid 2025)
The next step was to code automation and agents. Our teams can introduce tools to complete end-to-end tasks. Agents can read context, write code, run tests, and debug, all while the developer reviews and directs. We felt strongly that this was the future of engineering and we were fully committed.
The real reason came quickly. Agents writing off-the-shelf code couldn’t access our internal build systems, our libraries, or verify that that code actually ran in our environment. So, we built that agent integration ourselves using MCP, a standard that allows AI agents to connect to external tools and systems, and we sent them to all developers. To promote adoption, we’ve organized events to provide developers with a dedicated environment to learn, test, and build confidence with new tools. Agent utilization went from zero to 80% of monthly acquisitions.
The next challenge was scale. Developers wanted multiple agents to work in parallel, around the clock, without supervision. So we built an agent platform on top of our Kubernetes infrastructure. Every agent runs within a single container that replicates the real HubSpot developer environment. Agents compile code, run automated tests, read error output, and iterate until everything works. No human intervention is required.
By the end of Phase 2, 96% of engineers were using AI tools, engineering speed had increased by 60%, and lines of code reviewed per engineer had increased by 48%. We were starting to send better products quickly through agents. But that was just the beginning.
Phase 3: Scaling with our AI Platform (Mid 2025-Present)
HubSpot’s platform approach to product development has always been how we’ve created more customer value. When we built reporting and automation at the platform level, we didn’t just ship one feature; we sent that capability to all hubs at the same time. That’s how innovation comes together.
We applied that same logic to our AI infrastructure in Phase 3. Instead of building every agent from scratch, we built a shared foundation: how agents access data, what actions they can take, how they connect to the rest of HubSpot. Everything rushes over you.
The result is that all our agents work together. They speak the same language, share the same toolsets, and take the same context. The customer gets a consistent experience regardless of which agent they use because, at bottom, they are all built on the same infrastructure. And because they’re all connected, every new capability we add makes the whole system more valuable. That object is a collection of point solutions that cannot be repeated.
And it’s made possible by the way we’ve measured engineering with AI. Today, 100% of our developers are using AI, the lines of code reviewed per developer have increased by 73%, and the initial response to pull requests has decreased by 90%. That means less waiting time and more time to ship items to customers.
Why this is important: Integrating customer value
Having the right infrastructure accelerates the pace of innovation. At HubSpot, every agent we build makes the platform even more powerful. Every piece of context we add to the platform makes each agent more effective. For customers, that means the product keeps getting better, faster, and more connected.
What used to take months now takes weeks, and those weeks translate directly into new energy in the hands of marketers trying to reach the right audience, reps trying to close deals, and Customer Success Managers trying to retain customers. They don’t have to think about the bottom stage. They just get to feel the effect.
