Developers: Trust But Verify | Global Finance Magazine
Despite its broad adoption, AI raises questions amongst the coding community.
Artificial intelligence has gone from a novelty to widespread adoption among software developers, with 90% of developers using the technology in their workflows, up 14% from a year earlier, according to a study by Google Cloud’s DevOps Research and Assessment (DORA) team. However, the same study finds a trust gap with the technology.
“While 24% of respondents report a ‘great deal’ (4%) or ‘a lot’ (20%) of trust in AI, 30% trust it ‘a little’ (23%) or ‘not at all’ (7%),” wrote Ryan Salva, senior director, product management at Google, in a DORA blog post. “This indicates that AI outputs are perceived as useful and valuable by many of this year’s survey respondents, despite a lack of complete trust in them.”
Such adoption findings do not come as a surprise to Matt Kropp, managing director and senior partner at the Boston Consulting Group.
“AI is already in the flow of work for many developers and inside the integrated development environment (IDE) for code suggestions, in code search, test generation, documentation, and even basic refactoring,” he says. “That said, adoption is still ‘wide but shallow.’”
Still, more than 80% of the DORA study’s 5,000 respondents also noted that AI has enhanced their productivity, and 59% report a positive impact on code quality.
Global banking giant Citi has seen dramatic productivity gains enabled by the technology in the past few years. According to Citi chair and CEO Jane Fraser, AI-driven automated code reviews have exceeded 1 million in 2025. “This innovation alone saves considerable time and creates around 100,000 hours of weekly capacity as a very meaningful productivity uplift,” she said during the bank’s third-quarter earnings call.
AI has taken much of the toil out of developing and implementing code, but it still has much more potential to address additional tasks. “There’s still headroom in areas like structured refactoring, better test coverage, and smoother migrations. AI is strongest on new code paths,” says Kropp. “It’s less reliable on legacy systems without context. Guardrails—secure patterns, repo rules, and review discipline—are what turn the remaining ‘easy wins’ into real gains.”