New data report: Big code in the AI era

Big Code has long been a challenge for enterprise dev teams. Codebases today are made up of millions (sometimes billions) of lines of code written by thousands of devs over the span of two to three decades. The complexity that comes from these massive codebases creates massive tech debt, and causes issues around code maintenance, collaboration, and scalability. And with the overnight boom of AI, Big Code pains are about to get much worse.

Today, Sourcegraph is releasing a full data report on AI's impact on Big Code in the enterprise, "Big Code in the AI era" and the risk companies face if they fail to standardize AI adoption for their teams.

• Big Code's getting bigger. 77% of devs say their codebase grew 5x over the past three years.
• AI is taking software by storm. 95% of devs report using AI tools to write code, and 76% say they are worried about being able to manage the new code AI will generate.
• This is a "do or die" moment for companies. 72% of devs say Big Code presents a real risk to their company's ability to innovate and compete.

The kicker? Only 65% of devs say their company has a plan to address Big Code prior to the rise of AI.

View the full report: Big Code in the AI era

Survey highlights

The AI era is here, ready or not.

• 76% of developers are excited about the rise of dev tools powered by AI, but there are still big concerns around AI and its impact on Big Code.
  • 61% are concerned about AI's impact on tech debt.
  • 67% expressed concern about code sprawl due to the growth of AI.
  • 76% are worried about the amount of new code created and the need to manage it.

Innovation is at risk.

• What have those devs ID'd as the biggest threats Big Code and AI present in their day-to-day?
  • 95% need help getting up to speed on the codebase significantly faster.
  • 91% want to be able to ID and resolve code issues more efficiently.
  • 91% would save a significant amount of time if their codebase was fully searchable across all sources and repos.
  • 88% want a tool that would allow them to have a greater output with fewer people and resources.

We can't ignore the Big Code problem anymore.

• Devs spend almost as much time reading, interpreting, and fixing code as they do writing it. Devs told us that they spend only 20% of their time in the codebase writing new code for core products.
  • That 20% drops down to 14% when you take non-code activities into account like meetings and documentation.
• 87% of devs and engineering leaders are already struggling to keep their knowledge of new code up to date.
• 74% of developers and engineering leaders say they already rely too heavily on other people to understand their code base and product architecture in order to develop new code.
• Developer satisfaction is taking a big hit because of Big Code:
  • 73% of devs are blocked more frequently due to the size of their codebase.
  • 85% of devs struggle to maintain efficiency in their day-to-day.
  • 82% wish they could spend less time looking for information or old code and more time actually coding.

How can you get ahead of big code + AI pains?

Are we headed into a future where companies declare bankruptcy on Big Code in hopes of getting back to innovation, or are there strategies and solutions dev teams can adopt to get ahead of Big Code pains in the rise of AI?

• Form an AI advisory committee with leaders from your engineering, legal, and security teams and task them with ID-ing the risks and opportunities AI presents for your company, and plan ahead to address them.
• Streamline your use of AI across teams and consider a "Bring Your Own LLM". Security standards, retention policies, and documentation for AI tools vary, so pick one tool you're confident in and standardize on it.
• Put AI to work on Big Code. Yes, AI can make Big Code pains worse, but many devtool companies are embedding AI into their workflows to help devs get back to building cool stuff. Use AI to gain clearer insight into code ownership, find vulnerabilities faster by using AI to search across your entire codebase, and rely on AI coding assistants like you would a senior dev: to onboard team members and provide valuable context that's easily lost in growing codebases and changing teams.