Their jobs were among the first to be changed by AI. Now some are trying to avoid it

Emilija Manevska/Moment RF/Getty Images

In 2024, software engineer Laura Housh began using AI as a helper for coding and other tasks. Two years later, Housh says she has a love-hate relationship with AI.

“I’ve tried to use it for coding, and I just can’t get behind it,” she said.

AI can provide direction or serve as a second set of eyes, she said, but added that it lacks the context to be truly useful.

Housh isn’t the only one who thinks AI isn’t all that it’s cracked up to be.

Six in 10 tech workers who participated in a new survey say they try to avoid using AI, according to Stack Overflow, a popular question-and-answer site for programmers. The results of Stack Overflow’s 100-question survey, which includes responses from 30,000 software developers in 169 countries, were shared exclusively with CNN ahead of their Tuesday publication.

Software engineers have been at the forefront of the AI shift as one of the first industries to embrace the technology. Both tech giants and startups are increasingly using AI to write code, and “software developer” was ranked as the job most poised to be transformed by AI in an Indeed Hiring Lab report last year.

The survey’s results, like Housh’s experience, provide a glimpse at what that transition has been like for some developers, painting a mixed picture of both rising adoption and uncertainty about how the technology is reshaping their professions.

The Stack Overflow survey showed that 65% of responders use coding assistants or agents. Last year, Google’s research division published a survey of 5,000 tech professionals from around the world and found that 90% said they use AI in their job — a 14% increase from 2024.

But job unhappiness among respondents in Stack Overflow’s survey has also grown compared to last year. Nearly one in three said they are not happy in their current roles, compared to 28.4% last year. About 45% described themselves as “complacent” with their jobs and only 22.3% are “happy.”

The biggest factors among those who are complacent or unhappy are burnout, tech fatigue, economic uncertainty and a lack of clarity about upward mobility in their role.

Many of those anxieties would exist regardless of AI. But the fast pace of AI advancement has “augmented” some of those concerns, according to Stack Overflow CEO Prashanth Chandrasekar.

“People have to completely retool their jobs, and so it is quite unnerving for folks to know exactly how much of what they’re doing will be automated,” Chandrasekar told CNN.

Software developers are grappling with concerns that are more nuanced than whether their jobs will be replaced, according to Chandrasekar. Senior programmers may feel the skills they’ve honed are no longer as relevant, he said. Entry-level workers are expected to operate as if they have three years of experience since “the bar has been raised” because of AI. Tech workers often use their spare time to learn the latest coding tools, he said.

Mark Felix/AFP/Getty Images

Housh says she’s familiar with the pressures of learning and keeping up with AI. She said she watches AI-focused videos and reads platform documentation in her personal time — and she still struggles to keep up.

“It’s taxing because you have a lot of noise saying, ‘Do this, do this, do this, or you’ll get left behind,’” she said.

For programmers like Housh and Audrey Eschright, a software engineer with 20 years of experience, it’s not about a lack of desire to embrace the new technology. Rather, the problem is that AI agents have neither the broad general knowledge of an experienced developer nor the job-specific context of a longtime employee.

That can cause friction and more fatigue for developers, said Housh. What’s more, using AI instead of writing code manually can leave developers out of the loop of the solution, Eschright said .

“Writing code is how you solve problems,” said Eschright. “As a software engineer, the process of figuring out what a system does is to change it.”

This sentiment is also captured in Stack Overflow’s survey results. A little more than 74% of respondents said AI is useful for implementing code, and about 72% said it’s helpful for debugging — but only 34.4% say AI is useful for other job functions like communication and design.

The survey showed that just 6.6% trust AI with important decisions.

And about 78% said that source attribution is important when checking AI-generated code.

Developers are particularly careful about trusting AI today because the stakes are higher as more companies move AI projects from the pilot phase to mainstream use, according to Chandrasekar. A January report from Deloitte found 54% of the 3,200-plus business and IT leaders it surveyed expected to move 40% or more of their AI experiments into production in the next three to six months.

“I don’t like losing control; I like critical thinking and curiosity,” said Housh. “It’s just so important to be more curious than ever.”

Emilija Manevska/Moment RF/Getty Images

2024年，软件工程师劳拉·豪什开始使用人工智能辅助编程和其他任务。两年后，豪什表示她对人工智能又爱又恨。

“我试着用它来编程，但我就是学不会，”她说。

她表示，人工智能可以提供指导或充当第二双眼睛，但她补充说，人工智能缺乏真正有用的背景信息。

并非只有 Housh 认为人工智能并没有人们吹捧的那么好。

据程序​​员问答网站 Stack Overflow 的一项最新调查显示，参与调查的科技工作者中有六成表示他们会尽量避免使用人工智能。这项包含 100 个问题的调查收集了来自 169 个国家/地区的 3 万名软件开发人员的反馈，调查结果在周二正式发布前已独家提供给 CNN。

软件工程师一直走在人工智能变革的前沿，他们是最早拥抱这项技术的行业之一。无论是科技巨头还是初创公司，都在越来越多地使用人工智能编写代码。去年，Indeed招聘实验室的一份报告将“软件开发人员”评为最有可能受到人工智能变革影响的职业。

调查结果与 Housh 的经历一样，让我们​​得以一窥一些开发人员的转型历程，描绘出一幅喜忧参半的景象：一方面是技术普及率不断提高，另一方面是人们对这项技术如何重塑其职业感到不确定。

Stack Overflow 的一项调查显示，65% 的受访者使用代码助手或代理。去年，谷歌的研究部门发布了一项针对全球 5000 名科技专业人士的调查，发现 90% 的受访者表示他们在工作中使用人工智能——比 2024 年的预期增长了 14%。

但与去年相比，Stack Overflow 的调查显示，受访者对工作的不满情绪也有所上升。近三分之一的受访者表示对目前的工作不满意，而去年这一比例为 28.4%。约 45% 的受访者表示对自己的工作“安于现状”，只有 22.3% 的受访者表示“满意”。

导致人们感到自满或不快乐的最大因素是职业倦怠、技术疲劳、经济不确定性以及对自身职业晋升机会缺乏清晰的认识。

即使没有人工智能，许多这类焦虑依然存在。但据 Stack Overflow 首席执行官 Prashanth Chandrasekar 称，人工智能的快速发展“加剧”了其中一些担忧。

“人们不得不彻底改变他们的工作方式，因此，当人们确切地知道自己正在做的事情有多少会被自动化时，这让他们感到非常不安，”钱德拉塞卡尔告诉 CNN。

钱德拉塞卡尔表示，软件开发人员面临的担忧远比工作是否会被取代更为复杂。他说，资深程序员可能会觉得他们苦练的技能已经不再那么重要。由于人工智能的出现，“门槛提高了”，入门级员工现在必须具备三年工作经验才能胜任工作。他还提到，科技工作者经常利用业余时间学习最新的编程工具。

Mark Felix/AFP/Getty Images

胡什表示，她很了解学习和跟上人工智能发展步伐的压力。她说，她会在业余时间观看人工智能相关的视频，阅读平台文档——即便如此，她仍然觉得难以跟上。

“这让人很疲惫，因为到处都是这样的声音，‘做这个，做那个，做这个，否则你就会被落下，’”她说。

对于像Housh和拥有20年经验的软件工程师Audrey Eschright这样的程序员来说，问题不在于他们是否愿意接受新技术。而在于，人工智能代理既缺乏经验丰富的开发人员所拥有的广博知识，也缺乏资深员工所具备的特定工作经验。

Housh表示，这可能会造成摩擦，并增加开发人员的疲劳。Eschright指出，此外，使用人工智能代替手动编写代码可能会使开发人员无法参与解决方案的整个流程。

埃施赖特说：“编写代码就是解决问题的方式。作为一名软件工程师，弄清楚一个系统的功能就是改变它。”

Stack Overflow 的调查结果也反映了这种观点。超过 74% 的受访者认为人工智能对代码编写很有用，约 72% 的受访者认为它有助于调试，但只有 34.4% 的受访者认为人工智能对沟通和设计等其他工作职能有用。

调查显示，只有 6.6% 的人信任人工智能，认为它无法做出重要决策。

约 78% 的人表示，在检查 AI 生成的代码时，来源归属非常重要。

Chandrasekar指出，如今开发者对人工智能的信任度格外谨慎，因为随着越来越多的公司将人工智能项目从试点阶段推进到主流应用阶段，风险也随之增加。德勤1月份发布的一份报告显示，在其调查的3200多名企业和IT领导者中，54%的人预计在未来三到六个月内，会将40%或更多的人工智能实验投入生产环境。

“我不喜欢失去控制感；我喜欢批判性思维和好奇心，”胡什说。“保持好奇心比以往任何时候都更加重要。”