New York

October 15–17, 2025

Berlin

November 3–4, 2025

Why developers and their bosses disagree over generative AI

How to fix the disconnect over generative AI adoption and developer productivity.
May 08, 2025

You have 1 article left to read this month before you need to register a free LeadDev.com account.

Estimated reading time: 7 minutes

There’s a stubborn disconnect between what engineering leadership thinks will make developers productive and what they actually want.

In a survey of 2,150 IT managers and developers last year, leaders listed AI as the most important technical factor in improving developer productivity and satisfaction. All while only a third of developers reported experiencing any significant AI productivity gains.

Why the gap? It’s likely that generative AI is being adopted in the wrong places. Coding assistants are primarily applied to writing code, which is precisely what developers want to spend more time doing. This tunnel-vision definition of developer productivity ignores more pressing developer frustrations.

Instead, generative AI investments should focus on the whole software development lifecycle (SDLC). And since developer happiness drives productivity, it’s crucial to gain their buy-in for any AI adoption strategy. 

First, ask your devs

As a recent Hacker News thread highlights, so much AI hype is grounded in a misunderstanding of engineering work actually is. Top-down AI mandates often forget your most important stakeholders – your developers.

“Most leaders invest in AI, hoping it will boost productivity, but developers often see it as just another layer of noise,” said Andra Stefanescu, a neuro-mindfulness coach and trainer. “The brain is wired to resist solutions that feel imposed or misaligned with real pain points. When managers assume instead of asking, they miss the mark, and trust erodes.”

This feeds a real concern that an increase in AI adoption will leave developers feeling even more insecure in their jobs.

“Ask what’s actually slowing developers down. Create psychological safety so they’ll tell you the truth,” Stefanescu said. “Then use AI to solve those problems, not the ones leadership imagines exist.”

Think small, then scale

When a manager wants to put an AI strategy in motion, Andrew Zigler, senior developer advocate at LinearB said, they should look for atomic ways “instead of clobbering things with AI.” 

Despite the top-down pressure to add AI everywhere, let different development teams set their priorities and test AI where they see fit. Then, encourage them to share results with other teams, answering questions and advocating in favor of or against solutions.

The inaugural DORA report on the Impact of AI in Software Development found that if AI adoption is increased by just 25%, there’s a greater than 2% increase in flow, productivity, and job satisfaction. 

DORA outlined five strategies for ensuring AI amplifies developers’ value:

  1. Enable GenAI usage across the software delivery lifecycle – not just to create code.
  2. Emphasize that AI doesn’t replace developers, and acknowledge adoption of these new tools and workflows takes effort.
  3. Reward outcomes, not just time spent. 
  4. Frame GenAI as a way to learn new skills.
  5. Don’t make any GenAI mandatory.

For AI to be successful, Jamil Valliani, Atlassian head of product for AI, thinks it must be embedded within a team’s workflow, “almost without them having to actively think about it.” That includes both something like GitHub Copilot providing suggestions, and agentic AI weaved into automations across the entire SDLC.

GenAI as a partner, not a replacement

“Generative AI’s real value may lie not in how much code it helps write – but in how it helps developers do their best work more often,” said Lizzie Matusov, CEO of Quotient. 

The biggest shift from generative AI is not even in code output, but in how developers think, feel, and approach their work. Which you can only work out by asking them about it.

Like any technological adoption, engineering leadership needs a strategy to foster software developers’ trust in GenAI. DORA offered a five-step plan to achieve this:

  1. Create and communicate a generative AI usage policy.
  2. Emphasize fast feedback through code reviews and automated testing.
  3. Encourage developers to test out AI, especially in their preferred programming languages.
  4. Talk with developers about how they can envision future work partnering with AI integrated across the SDLC.
  5. Encourage GenAI usage, never mandate it.

Atlassian’s State of DevEx Report found that developers lose one day a week to inefficiencies, including more than 30 minutes a day spent looking for things. This ties directly to developers pointing to technical debt and insufficient documentation as their biggest impediments to productive time.

Yet, despite these persistent complaints, developers rarely report wanting to spend time paying down technical debt or writing documentation. Whether it’s unraveling your complex web of infrastructure and ownership, or filling in your docs gaps, AI makes for a near perfect explainer. 

Spotify recently announced AiKA, its AI knowledge assistant that runs across Backstage to battle this knowledge fragmentation, even reminding developers when they are responsible for creating or reviewing documentation. 

Encourage conversational prompting that treats AI like a pair programmer, not a one-way Google or StackOverflow search. GitHub Copilot’s code review agent has already reviewed more than 8 million pull requests, showing how developers are comfortable treating AI as a collaborative partner.


Common AI code assistant use cases

While no team is the same, looking at common developer challenges and mapping them to AI code assistant use cases is one way to get started. 

In a just released AI assisted engineering guide by DX, almost a third of respondents are already using AI to refactor existing code, helping to pay down that technical debt.

Because generative AI is particularly good at explaining complex ideas, it’s no surprise that popular AI use cases uncovered include:

  • Learning new techniques.
  • Complex query writing.
  • Brainstorming and planning.
  • Code explanation.

Overall, developers reported that their most time-saving AI use case was stack trace interpretation and explanation, enabling them to more quickly or accurately pinpoint the exact location of an error in code.

Using AI for code documentation was only the fifth most common dev response, even though last year’s DORA report found generative AI could help increase documentation quality by 7.5%.

Another quarter of respondents are using AI to generate test cases, which isn’t many developers’ idea of fun. 

When in doubt, regularly ask developers what parts of their jobs they find most frustrating. That’s what your AI strategy should focus on – not replacing the part they love.

How managers can aid impactful AI adoption 

“Leaders are under these mandates from nontechnical leaders to integrate AI into their workforces and workflows and to connect that to ROI and impact – because AI is not free,” Zigler said. This can lead to “a lot of friction with their engineering practitioners who maybe don’t see as much value or see it as something getting in the way.”

Part of the problem is that measuring the impact of engineering is already hard. Add in the unknowability of AI and it only gets harder.

Try picking a small thing where AI is just one of a number of potential solutions and keep a dialog open with developers as they test and learn. 

Stefanescu offered three questions that can help open up developers to discussing change:

  • Can you help me understand your perspective on this issue?
  • What would a positive outcome look like for you?
  • What’s one thing I could do to help you succeed right now?

“When devs feel heard, AI becomes a tool they want, not a strategy they resist,” Stefanescu said, “because when people feel heard, they stop resisting and start building with you.”

Manage top-level expectations

An experimental, iterative approach helps equip engineering teams for conversations with the rest of the business.

“The divide is real. It’s important to acknowledge that there are two different worlds in engineering in how teams and leaders are looking at using new tools,” Zigler said. “A VP of sales would know their pipeline, but a VP of engineering does not have those same shared terms with their leaders to have an honest conversation of what they’re building.”

Engineering leadership must be able to manage expectations from the C-suite. 

Execs are ‘AI everything’ – they don’t have the technical grasp,” Zigler said. “You have to meet them in their world. You have to talk on their terms. You have to be a clear negotiator of what your execs can expect from AI. Setting expectations is a critical part of a successful adoption.”

As the DORA GenAI impact report warned: any adoption of new technology can lead to some near-term decreases in productivity and impact. Instead, it recommends:

  1. Share and be transparent about how your organization plans to use AI.
  2. Address developer concerns about AI’s impact.
  3. Allow ample time for developers to learn how to use AI.
  4. Create policies that govern adoption of AI.

“Developers are embracing generative AI tools like GitHub Copilot and Cursor to speed up their workflows,” said Zohar Einy, CEO of Port. 

Leadership, he warned, may not be fully aware of the knock-on effects of generative AI at scale. “Faster deployments come with new complexity, more microservices and incidents that are harder to troubleshoot – especially when humans didn’t write the code themselves. Left unchecked, this chaos can slow teams down instead of speeding them up.”

Indeed, Harness’s State of Software Delivery 2025 found that 67% of developers spend more time debugging AI-generated code, while 68% spend more time resolving security vulnerabilities. 

The whole engineering division must be mindful that moving even faster, without intent, could just break more things.