AI is transforming how software is built, happening faster than most of us expected. But what does this shift mean for engineering teams and leaders?
At DeveloperWeek 2025, I had the opportunity to moderate a panel with some of the biggest players in AI — Amazon, Microsoft, Google, and Augment Code — to tackle this question. We had an insightful discussion on how LLMs are changing the software development process, the new skills engineers need, and what leaders should do to stay ahead.
Joining me on stage were:
- Anoop Deoras – Director, AI/ML at Amazon Q
- Paulo Zacchello – Lead TPM, HITL AI Platforms at Google
- Nilo Dutta Roy – Sr. Director, Product Management at Microsoft AI
- Vinay Perneti – Engineering Director at Augment Code
Here are my biggest takeaways from the discussion:
1. AI Coding Assistants Are Moving Beyond Code Generation
We’ve all seen AI tools that can generate code, but today’s most significant advancements go far beyond that. Our panelists shared how AI is now helping engineers with:
- Automated testing: AI can automatically generate unit tests and improve code coverage.
- Code migrations: Amazon shared how AI helped migrate 30,000 applications to a new Java version, saving 4,500 hours of manual effort.
- Incident response: AI is reducing the stress of on-call engineering by surfacing relevant logs, recent commits, and troubleshooting suggestions in real-time.
“These AI systems are accelerating a variety of software development tasks by 80%. They’re not just generating code — they’re transforming entire workflows, from documentation to complex engineering tasks,” said Anoop Deoras at Amazon.
Vinay Perneti, Engineering Director at Augment Code, highlighted the AI shift:
“If an AI assistant understands your codebase, documentation, and past conversations, it doesn’t need to get the perfect answer — it just needs to get you started. That’s a huge difference. It reduces stress, improves productivity, and gives developers more time to focus on product-level thinking.”
The key shift is that AI is no longer just a coding assistant — it’s becoming a real-time engineering companion.
2. Human Expertise Is Still Critical
While AI can speed up many tasks, it still needs human oversight — especially when it comes to:
- Code quality and security — AI-generated code can introduce vulnerabilities if not adequately reviewed.
- Best practices — AI can write code but doesn’t always follow optimal design patterns unless trained with the correct data.
- Avoiding hallucinations — Microsoft and Google emphasized that human-in-the-loop reinforcement learning is essential to prevent AI from generating misleading or incorrect code.
Nilo Dutta Roy, Sr. Director of Product Management at Microsoft AI, emphasized the role of human feedback:
“There’s a difference between models becoming intelligent and useful in real-world applications. Human labels are making that difference. Reinforcement learning with human feedback ensures that AI-generated code follows best practices, security standards, and performance optimizations.”
The consensus? AI is an amplifier, not a replacement. Engineers who learn how to guide and validate AI’s output will have the most significant advantage.
3. The Skills Gap Is Shifting — Engineers Must Learn to Direct AI
AI isn’t taking away engineering jobs — it’s changing the skills engineers need to be effective.
The panelists agreed that the best engineers in an AI-driven world will be the ones who:
- Understand where AI excels and where it fails — Engineers must learn how to work with AI effectively, using it to automate tedious tasks while still applying human judgment.
- Think beyond syntax — As AI handles more of the “grunt work,” engineers must focus on architecture, system design, and strategic problem-solving.
- Embrace experimentation. Teams that adopt and refine AI tools will outpace those that resist change. Leaders must encourage a culture of curiosity and continuous learning.
“Your developers need to be in the driver’s seat, directing the AI agent and defining the goals,” said Anoop Deoras from Amazon.
This shift requires a mindset change — instead of simply writing code, engineers will guide AI agents to execute complex technical workflows, ensuring that AI solutions align with business objectives and engineering best practices.
4. Measuring AI’s Impact: Adoption Over ROI Calculations
A common challenge for engineering leaders is measuring AI’s impact. Should we track lines of code written by AI? Are bugs prevented? Is productivity increased?
The panel agreed that developer adoption and satisfaction are the best proxies for AI’s success. If engineers keep using an AI tool because it makes their lives easier, it’s delivering value. If it’s gathering dust, it’s not.
At Augment Code, for example, they track:
- Repeated usage — Are developers using the AI tool daily?
- Accepted tab completions — Are developers satisfied with the suggestions from the AI assistant?
- Developer sentiment — How likely are they to recommend it to their peers?
“The best proxy is adoption,” said Vinay Perneti at Augment Code. “Are people repeatedly using the tool? Not just on day one but months later?”
At Revelo, we’ve observed that teams that embrace AI early gain a competitive edge, while those that delay adoption risk falling behind in speed and efficiency.
5. What Engineering Leaders Must Do To Stay Ahead
The role of engineering leadership is evolving. AI is breaking down traditional team structures, and the boundaries between engineering, product, and data science are blurring.
Key actions for leaders:
- Encourage experimentation — Give engineers space to try new AI tools and integrate what works.
- Redefine team structures — Consider cross-functional teams where engineers, product managers, and designers collaborate with AI to iterate faster.
- Lead from the front — Leaders must use AI to understand its strengths and weaknesses. You can’t delegate AI adoption — you need to be hands-on.
Paulo Zacchello at Google emphasized the shift in team structures and talent requirements:
“We’re moving away from traditional team setups of frontend and backend engineers. Instead, we’re seeing the rise of holistic, multidisciplinary teams working together, including product, UX, and engineering. There’s also a growing need for talent that can blend engineering with data science. Leaders need to recognize and nurture this hybrid skill set.”
Final Thoughts
AI is already transforming software development, and we’re only at the beginning.
If there’s one thing I took away from this panel, it’s this: The future belongs to engineering leaders who embrace AI — not just as a tool, but as a fundamental shift in how teams work and build software.
A huge thank you to our panelists Anoop, Nilo, Paulo, and Vinay for sharing their expertise and to the DeveloperWeek team for putting together such a fantastic event.
The post How AI Is Reshaping Software Engineering: Key Takeaways From DeveloperWeek 2025 appeared first on The New Stack.
