Back office ops · Production

HubSpot scales AI coding tool adoption to near-universal across its engineering organization

The problem

HubSpot's engineering organization needed to scale AI coding tool adoption beyond early POCs and early adopters, but faced exploding demand causing backlogs, conservative guardrails requiring individual license requests, cost concerns, and internal skepticism about AI's safety impact on production.

First attempt

HubSpot initially relied on teams already managing their GitHub setup to drive AI adoption, but as demand exploded the backlog became unmanageable; procurement processes built for long-term contracts also slowed rapid evaluation of new tools.

Workflow diagram · grounded in source

Executive push to evaluate Copilot

trigger

“Dharmesh had recently used GitHub Copilot to build ChatSpot and had a good experience with it, so he and Brian pushed us to evaluate it”

GitHub Copilot pilot with teams

ai_action

“Our strategy was to include entire teams so they could adopt and learn together that had different experience levels, different missions, and worked in different domains. We gave teams over two months to try it.”

Measure productivity and incident impact

validation

“We pulled metrics on code review burden, cycle time, velocity comparisons before and after adoption, and production incident rates.”

Universal license rollout

output

“In May 2024, we ditched the restrictions. Then we proactively gave everyone a seat, making it as easy as possible to get started. Adoption shot above 50% overnight.”

Curated MCP server per engineer

integration

“We transparently set up a local MCP server on every machine with default rules and configuration optimized for our development environment.”

Sidekick AI assistant launch

output

“creating Sidekick (our AI assistant that answers platform questions, creates issues, implements changes, and reviews PRs)”

Peer video sharing feedback loop

feedback_loop

“Whenever we heard someone did something interesting with AI, we asked them to record a video and share it”

Reported outcome

HubSpot achieved near-universal AI coding tool adoption across engineering, surpassing 90% adoption, with data showing no correlation between AI adoption and production incidents, and built Sidekick plus 400+ tools for agents.

Reported metrics

AI coding tool adoption (immediate milestone)above 50%

AI coding tool adoption (peak milestone)90%

AI coding tool adoption levelnear universal adoption

Agent tools available400+

Show all 6 reported metrics

AI coding tool adoption (immediate milestone)above 50%

AI coding tool adoption (peak milestone)90%

AI coding tool adoption levelnear universal adoption

agent tools available400+

AI adoption vs production incidents correlationno correlation

engineer productivity improvementmeasurable but modest productivity improvements

Reported stack

GitHub CopilotCursorMCP serverSidekickOpenAIAnthropic

Source

https://product.hubspot.com/blog/context-is-key-how-hubspot-scaled-ai-adoption

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

HubSpot achieved near-universal AI coding tool adoption across engineering, surpassing 90% adoption, with data showing no correlation between AI adoption and production incidents, and built Sidekick plus 400+ tools fo…

What tools did this team use?

GitHub Copilot, Cursor, MCP server, Sidekick, OpenAI, Anthropic.

What results were reported?

AI coding tool adoption (immediate milestone): above 50%; AI coding tool adoption (peak milestone): 90%; AI coding tool adoption level: near universal adoption; Agent tools available: 400+ (source-reported, not independently verified).

What failed first in this deployment?

How is this back office ops AI workflow structured?

Executive push to evaluate Copilot → GitHub Copilot pilot with teams → Measure productivity and incident impact → Universal license rollout → Curated MCP server per engineer → Sidekick AI assistant launch → Peer video sharing feedback loop.