New data from Jellyfish, a software engineering management service, reveals a sharp uptick in how companies are moving beyond assistive AI to embrace tools that independently take action in the coding process.
Between December 2024 and May 2025, Jellyfish analyzed data from more than 400 companies to gauge adoption of agentic AI in software coding. This is technology that doesn't just suggest code, but performs tasks such as reviewing, authoring, and submitting code.
At the start of 2025, just over half of companies were using agentic AI for these tasks. By May, that figure had jumped to 82%, according to Jellyfish data.
AI-powered code reviews grew even more, from 39% in January to 76% in May. Popular tools for this task include GitHub Copilot Reviewer, Cursor BugBot, and CodeRabbit, with up-and-comers being Graphite and Greptile, according to Jellyfish data.
There's still a long way to go before AI fully takes over software coding. This can be seen in a third set of data from the same period. Just under 8% of companies analyzed by Jellyfish are piloting fully agentic coding workflows, where AI is not just reviewing, but actually writing and submitting code. That's up from less than 2% in January, but still pretty low adoption.
Sign up for BI's Tech Memo newsletter here. Reach out to me via email at abarr@jkmperu.com.
