How HubSpot got all engineers to use AI without any mandates

Across Big Tech, Wall Street, and other major industries, employers have pushed their workforce to embrace artificial intelligence through a mix of tactics that go beyond mere nudging, including tracking how often they are logging into these workplace tools and linking annual performance reviews to AI usage.

But at software-as-a-service company HubSpot, 100% of engineers are using AI in their workflows, which the company says has resulted in a 73% increase in the lines of code updated by its engineers. Duncan Lennox, chief product and technology officer at HubSpot, says the milestone was achieved following a three-part phased rollout that began in 2023 and didn’t depend on a mandate.

“We found mandates not to be effective,” says Lennox, who has steered the engineering, product, IT, and security teams at HubSpot since September, after holding multiple leadership roles at Google and Amazon Web Services.

The proliferation of agentic AI coding tools like Claude Code and OpenAI Codex, which are both used at HubSpot, has led to growing fears about the future of work for technologists. Even HubSpot CEO Yamini Rangan weighed in this year, sharing that she still encouraged her son to pursue a career in the field. Technology industry layoffs are up 33% for the first four months of 2026 compared with the prior-year period, according to outplacement firm Challenger, Gray & Christmas. Major tech industry job cuts have been announced at large employers including Amazon and Meta. Employment trends have been particularly bleak for computer programming roles.

But Lennox says that the biggest barrier he had to overcome when initially seeking to leverage large language models in coding was less about fear of job replacement and more about concerns around introducing errors, quality issues, and reliability.

After achieving 30% adoption for a coding copilot that was rolled out by HubSpot, Lennox and his team shared incident data that showed engineers that the work being done by teams using that tool wasn’t negatively affecting reliability versus those that weren’t. This led usage to immediately increase to 50%.

By the time autonomous coding agents gained traction, 80% of HubSpot’s engineers were using AI at work. A few other factors helped further lift interest. The company’s more distinguished engineers were the first to notice how effective coding copilots could be, leading to some internal competitiveness that inspired junior employees who wanted to keep up. HubSpot also implemented hackathon events for training, including one that’s coming up soon to train workers on how to build their own AI agents.

LLM advancements were also an external influence. Anthropic’s upgraded Opus model, which debuted in November, was particularly heralded. “That was a meaningful step forward for coding, and that causes a spike because now you can do a bunch of things you couldn’t do before because the models have gotten better,” says Lennox. 

To further support HubSpot’s AI aspirations, Lennox says the company built its own customized infrastructure so that the team could read context, write code, run tests, and fix errors with autonomous coding agents. OpenAI, Anthropic, and Google are the three AI companies that HubSpot works closely with, and the team is encouraged to make swaps based on performance for various use cases. “We didn’t want 17 different frontier models,” says Lennox.

With all engineers now using AI, there’s been a spillover effect across........

© Fortune