Built on top of Microsoft’s Visual Studio Code, Cursor set out with a clear ambition: to go beyond existing AI coding tools like GitHub Copilot, which had already gained popularity since its official launch in 2022.
California-based go-to-market (GTM) platform Apollo.io swiftly moved its engineering team from GitHub Copilot to Cursor, which seemed to have “gotten it right”.
In an interview with AIM, Himanshu Gahlot, VP of engineering at Apollo.io, and Saravana Kumar, head of machine learning at Apollo.io, discussed the reason behind their engineering team’s switch from GitHub Copilot to Cursor.
California-based go-to-market (GTM) platform Apollo.io is a B2B sales platform powered by AI. It is designed to empower revenue teams with cutting-edge sales intelligence and engagement tools.
“We started using GitHub Copilot early last year when it had just launched. We noticed it and quickly started using it,” Gahlot shared. “Slowly, we realised that there are better tools out there, or at least the ones we could use even more effectively within our company.”
Is Cursor Really So Special?
Explaining in simpler terms, Gahlot said that Cursor uses a newer approach in AI in which you can ask it to do several things at once, and it takes care of them in one go.
The team ran a pilot program using Cursor with their engineers, and the response was overwhelmingly positive. “We got a 90% plus satisfaction rate. Almost every engineer said positive things about being able to understand the whole code base and generate the right things,” Gahlot added.
But while the tool showed promise, Gahlot warned that there is a lot of hype around AI tools that don’t always match reality.
“It did come with a caveat. You’d often hear people hyping these tools, claiming productivity gains of 25x or even 50x — but it’s very nuanced,” he said.
Gahlot added that these tools work really well if starting from scratch—what he calls “0 to 1” use cases. They can be incredibly helpful when building something new or just putting together a prototype or demo.
But things get tricky when dealing with large, complex code bases that have been developed over many years. “When it comes to a 10-year-old code base with millions of lines of code and like 30-40,000 individual files, then that is not how you would use it,” Gahlot said.
In such cases, Gahlot said, teams need time to figure out the right way to use the tool, and engineers need proper training.
Adding to this, Kumar pointed out that people tend to either overestimate or underestimate the capability of AI tools.
“I would say that is not even an important aspect,” he said, referring to converting natural language into code. “The important aspect is to understand what it can and can’t do.
He explained that if somebody is able to clearly explain what they want, including all the details, assumptions, and context, turning that into a working code is mostly handled by AI tools now.
“What is actually not done [by them] is figuring out how we solve the problems. That’s where humans come in,” Kumar said.
Reason Behind the Shift
Gahlot said they moved from GitHub Copilot to Cursor because, though the former was doing pretty well in terms of auto-completion and small code generation, their engineers were not finding much success.
“There wasn’t an “aha!” moment there. It wasn’t like, you know, you want to get something done, and it would just do it for you.”
However, when the team tried out Windsurf and Cursor, they found their “aha!” moment. “It is like, you can chat with your code, write anything you want done, especially in a zero-to-one use case, and it just does it for you, rather than completing part of your code or suggesting a few things,” he explained.
Apollo.io began adopting Cursor more widely and started seeing higher satisfaction among engineers using the IDE. However, as he pointed out, it’s not a one-size-fits-all solution.
He explained that different roles within engineering teams have different needs. The same applies to machine learning engineers, back-end developers, and front-end teams.
Other Big-Tech Collaborations
Apollo.io has now onboarded three major AI providers, OpenAI, Anthropic, and Google, and it is constantly experimenting with their new models. Gahlot believes the future will see businesses relying on multiple AI models for different tasks.
Gahlot also spoke about the company’s early collaboration with Anthropic. “We have been early partners with Anthropic on multiple things, specifically on the model context protocol (MCP) initiative that they recently launched,” he said.
“We were one of the first companies they launched MCP with. I think initially there were about 10 startups, and we were one of them.”
Currently, out of the 700 employees at Apollo.io, about 200 are spread across engineering, product, recruiting, sales, and support in India. Out of those, about 160 are in engineering and 40 in other departments, constituting 65% of its engineering team in India.
