AI coding mandates are driving developers to the brink

Estimated reading time: 7 minutes

Almost half of C-suite executives said in a recent survey that AI adoption is “tearing their company apart” as a rift emerges between leadership and the employees adopting such tools.

While 75% of company leaders thought their AI rollout over the past 12 months has been successful, only 45% of employees said the same.

“Even those C-suite leaders who believe their AI integration is proceeding smoothly are handing down policies and tools to a workforce that is more frustrated than they are,” Axios reporter Megan Morrone wrote of the findings.

For software developers specifically, there are concerns that AI coding tools are introducing errors into their code, failing at many tasks, and compounding technical debt. But they also feel that misguided mandates are inhibiting the successful adoption of AI tools. While AI coding assistants can be helpful, it’s clear that how leaders approach and support engineering teams makes all the difference.

FOMO drives adoption

The most attractive aspect of these tools to business leaders is their potential to automate repetitive coding tasks, which can make teams more efficient, help them ship faster, and increase revenue. 

The quiet implication also means employing fewer expensive developers, which Meta’s Mark Zuckerberg, Salesforce CEO Marc Benioff, and AWS chief Matt Garman are happy to say out loud. 

The creators of these tools are publishing all kinds of stats touting how much more productive they make developers, and many developers agree. In Stack Overflow’s 2024 annual survey of 65,000 developers, 58% cited improved efficiency as the biggest benefit, and 81% cited increased productivity.

There’s also a strong feeling that companies need to adopt these tools or risk being left behind, a popular talking point for enterprise software companies. GitHub’s Copilot has been adopted by 77,000 organizations since it was released in October 2021, Microsoft reported in its fourth quarter earnings. And that’s just one offering. Y Combinator’s managing partner, Jared Friedman, said that a quarter of startups in the accelerator’s current cohort have codebases that are almost entirely AI-generated.

“AI is something that helps us, and it is also helping our competitor as well, right? So if we are not utilizing this, we are not leveling the playing field with our competitor,” Simon Lau, an engineering manager at ChargeLab, a software company focused on electric vehicle chargers, said. 

Mo’ code, mo’ errors 

As time has gone on, developer faith in AI coding tools has quickly declined. In 2024, 72% of respondents in the Stack Overflow report said they held a favorable or very favorable attitude toward the tools, down from 77% in 2023.

Overall, developers describe a slew of technical issues and headaches associated with AI coding tools, from how they frequently suggest incorrect code and even delete existing code to the many issues they cause with deployments.

In a survey of 500 engineering leaders and practitioners conducted by Harness, 59% of respondents said they experience problems with deployments at least half of the time when using AI coding tools, and 67% said they spend more time debugging AI-generated code than before. Many also say the use of AI tools is causing an increase in incidents, with 68% of Harness respondents saying they spend more time resolving AI-related security vulnerabilities now compared to before they used AI coding tools.

“I tried GitHub Copilot for a while, and while some parts of it were impressive, at most it was an unnecessary convenience that saved only a few seconds of actual work. And it was wrong as many times as it was right. The time I spent correcting its wrong code I could have spent writing the right code myself,” said one developer in a Reddit discussion about company mandates dictating how developers use AI tools. 

While AI tools can certainly increase the pace and volume at which code is shipped into production, more code isn’t necessarily always a good thing, especially because bad code can generate tech debt. 

These tools also have limitations and often don’t perform well on more complex coding tasks – at least right now. “I think AI is really good at code generation for prototype-level or low complexity features, but when you talk about production systems, that’s when all these small things actually fail. And it takes a lot of time for someone to review something they actually didn’t fully write and understand,” said Alejandro Castellano, cofounder and CEO of automation company Caddi.

Executives mismanage their expectations 

While developers have learned to use AI tools thoughtfully, company leaders aren’t always exercising that same degree of intention.

Developer frustrations with AI mandates often surface due to their being handed down by company leaders who don’t have close visibility into engineering workflows. Developers describe executives instituting OKRs and tracking AI usage without any regard for whether it’s actually helping, let alone where it may be making things worse.

Code acceptance rate (how often developers accept the code suggestions an AI tool makes) is a popular adoption metric, but some argue it’s a poor measure because it counts people accepting suggestions that may be problematic. Sometimes, it’s just constant pushing to use AI “more” or to “be more efficient” without any intention around what that actually means or should look like. “It’s a constant battle to bring execs back to earth on their expectations of what Gen AI can do,” wrote one Reddit commenter. 

In what feels like a massive step backwards, some even describe that their companies have started publicizing how much code developers are individually generating with AI in an effort to push everyone to output more with the tools, either through weekly reports or public leaderboards.

The power of empowering the developers

AI coding tools are still new, and figuring out how to use them effectively requires a concerted effort. While tactics that essentially boil down to “more AI now” sow resentment and can hinder adoption efforts, leaders can foster enthusiasm and find positive results when they come from a place of understanding developers’ day-to-day and trust that they know what’s best for their work. 

“It’s not someone in an ivory tower or in a glass house who should say, ‘these are the best tools for this person,’” said Rajesh Jethwa, CTO of software engineering consultancy Digiterre, where he not only leads engineering teams, but also has insight into the engineering challenges various companies are dealing with. “People closest to the problems are the ones who have all the context, and context is what really matters when it comes to AI tools.”

---

London • June 16 & 17, 2025

Speakers Gergely Orosz, Camille Fournier and Lara Hogan confirmed.

See who else is speaking

Empowering engineers has been the modus operandi for ChargeLab, where every one of the company’s 40 engineers is now using AI coding tools in their work every day and reporting benefits. An internal survey revealed a roughly 40% productivity increase among the engineering team, according to CTO Ehsan Mokhtari. But it didn’t start out that way: Lau said that “initially, there was some pushback among our developers.”

Lau credits the team’s embrace of AI coding tools to the company’s more flexible approach, where they’ve empowered everyone to use whichever tools they want, however they see fit. 

“We are not enforcing the developer to use one tool or the other. We gave them resources to explore what works best for them,” he said, explaining how ChargeLab makes a wide variety of tools available to its developers, including Windsurf, Cursor, Copilot, ChatGPT, and Claude.

Mokhtari also put in the work by getting hands-on with the tools, earning trust and avoiding the common frustration of top-down directives from those with no real experience. There’s also a clear and transparent mandate from the top to save $1 million across the company this year using AI, which Mokhtari argues shows that AI is a priority for the company while maintaining flexibility to experiment and only pursue what’s working. Specific teams like Lau’s may have more granular metrics for tracking the adoption of AI tools, but they’re put in place by managers who have direct insight into each developer’s specific tasks and workflow.

The acceleration of AI technology and proliferation of AI coding tools is a major inflection point in the world of engineering. Company leaders are clear that they need to embrace it to stay competitive, but adoption doesn’t automatically equal success. In times of change, collaboration, empowering workers, and fostering a culture of innovation prove more important than ever.

“You cannot really pull people forward and make them innovative,” said Mokhtari. “You have to foster the culture.”