HubSpot Says All Engineers Now Use AI After Phased Rollout

HubSpot has reached full adoption of AI among its engineering staff without issuing mandates to employees. The software-as-a-service company said 100% of engineers now use AI in their workflows, resulting in a 73% increase in the lines of code they update.

Across the company, 94% of all employees use AI after the firm followed a similar approach of pilots, training events and sharing measurable results. Duncan Lennox, chief product and technology officer at HubSpot, said the milestone followed a three-part phased rollout that began in 2023.

“We found mandates not to be effective,” Lennox said. He has led the engineering, product, IT and security teams at HubSpot since September after holding leadership roles at Google and Amazon Web Services. The initial barrier to adoption was not fear of job loss but concerns about errors, quality and reliability when using large language models for coding.

After the company achieved 30% adoption of a coding copilot, officials shared incident data showing that teams using the tool maintained the same reliability levels as those that did not. Usage rose immediately to 50%. By the time autonomous coding agents became available, 80% of HubSpot’s engineers were already using AI.

Senior engineers were among the first to adopt coding copilots, which created internal competitiveness that encouraged more junior staff to participate. The company also held hackathon events to train workers, including one scheduled soon on building custom AI agents.

External advances in large language models contributed to higher usage. Anthropic’s upgraded Opus model, released in November, produced a noticeable improvement in coding capabilities. “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,” Lennox said.

HubSpot built customized internal infrastructure that allows autonomous coding agents to read context, write code, run tests and fix errors. The company works closely with OpenAI, Anthropic and Google, and encourages teams to switch between models based on performance for different tasks.

“We didn’t want 17 different frontier models,” Lennox said.

The spread of AI has begun to change how teams operate at HubSpot. Engineers, product managers and user experience professionals previously worked in clearly defined roles, but those boundaries are starting to blur. HubSpot continues to hire engineers while increasing total headcount at a slower rate than before the introduction of generative AI tools.

Lennox attributed the slower hiring pace primarily to an industry-wide correction after pandemic-era overstaffing at many technology companies. The company invested more than $900 million in research and development last year, equal to about 20% of revenue.

Lennox declined to disclose the exact size of the engineering team. The company is also seeking AI-native engineers who learned coding with tools such as Cursor or Claude Code, while still requiring grounding in computer science and software engineering principles.

Investors have pressured HubSpot’s shares amid broader industry concerns about potential effects of AI on software-as-a-service pricing models. The company reported solid first-quarter earnings last week but its shares fell 20% after issuing a revenue forecast slightly below analysts’ expectations.