Salesforce ships higher-quality code across 20,000 developers with Cursor
Salesforce has been building software for more than 25 years. It maintains a codebase that stretches across decades featuring every kind of system a modern tech company can accumulate. Twenty thousand engineers ship new products on top of it every day.
That's what makes the rapid adoption of Cursor inside Salesforce so notable. When an organization this central to the industry shifts how it builds software, it points at where the whole field is heading.
Today more than 90 percent of Salesforce's engineers use Cursor as part of their daily workflow. Salesforce has already seen double-digit gains in velocity and code quality, allowing it to ship products like Agentforce that are setting the pace for how AI is incorporated into software.
I would say that it's 0 to 1 in terms of how Cursor has transformed the way our developers use tools to improve quality of the product.
Two paths toward Cursor adoption
Before Cursor arrived, Salesforce had already invested in its own internal AI tools, including an open-sourced coding model called Code Genie that thousands of its developers relied on. But Salesforce wanted its engineers to have a range of options, so it made Cursor available.
Junior engineers were the first adopters. Many had started their careers during the pandemic, when remote work made standard ways of learning a codebase unavailable. Cursor helped them catch up.
They didn't have senior engineers sitting with them and explaining a lot of things. Cursor helps them actually understand existing code better so that they are able to contribute more effectively.
For senior engineers, Appajodu observed that they first proved out Cursor on the boring, tedious tasks that are inefficient to tackle manually. And as they experienced the value of automating these tasks and built trust in the tool, these senior engineers expanded their use cases quickly to higher value tasks.
Senior engineers start with the places that feel boring, and once they see the value there, they're ready to use AI coding for other use cases.
Adoption followed the same pattern across teams: a small group would try Cursor, see the impact, and the rest would follow. Within a few months, Cursor went from a new tool at Salesforce to one that nearly every single engineer at the company was using. And over time, AI proliferated into every aspect of the SDLC, not just code writing, according to Appajodu.
Three metrics that matter most
Salesforce is careful about how it evaluates engineering work. The company measures pace and reliability using three key metrics:
- Cycle time
- Quality (bug count)
- Throughput
Salesforce already had dashboards that tracked these metrics across thousands of engineers, making the impact of Cursor visible almost immediately. Cursor helped Salesforce make large improvements in all three areas while also reducing legacy code coverage time by 85 percent.
We have seen more than double-digit increases in all of these metrics.
Shipping better products than ever
We have seen a huge improvement in the quality of products.
To take one measure, engineers are now generating far more unit tests with Cursor than before, enhancing the reliability of what Salesforce ships.
Some challenges remain. As more code is written with Cursor's help, Salesforce is still refining how that code should be reviewed and how to maintain the same level of trust in each change. Even so, the direction is clear. AI is already reshaping how software is planned, built, and maintained, and Appajodu believes the real transformation is still ahead.
We are at the starting point of this journey. It's only going to get better.
If you're interested in leveraging AI to ship higher-quality software, please reach out to our team to get started with a Cursor trial.