From Developers to Strategists: How AI Is Redefining Roles, Teams, and Career Paths in Technology

At DevDay 2026, the biggest tech event in Central Vietnam, technology leaders from Enosta, CMC Global, MADLAB, Mynavi TechTus, and Xbrain explored how AI is reshaping software engineering, organizational workflows, and the future of technical careers.

Here’s what the industry is learning about AI-native development, engineering transformation, and the next generation of software teams.

What’re New Roles of a Software Engineer in AI?

Not long ago, being a good software engineer mostly meant being good at writing code. You knew your framework, you could go deep into a tech stack, and you could ship things.

That’s still part of the job. But it’s no longer the whole picture.

Nguyen Duc Nguyen, CEO of MADLAB Vietnam, described a shift that’s already happening inside many engineering teams. A year ago, AI was mainly used to support development work – smarter autocomplete, faster boilerplate, less repetitive work. He described this as the AI-assisted phase.

Now, AI is becoming a central part of the development process itself. Engineers are no longer only responsible for implementation. They also need to shape the input, provide the right context, and review whether the outputs are actually usable.

“The role of engineers is now shifting toward creating inputs and context that are good enough for Gen AI to produce useful outputs. Then having the capability to review those outputs. Engineers are moving from ”doers” to reviewers and managers who guide AI to work effectively”, Nguyen explained.

And that requires a skill many engineers are still learning: thinking beyond technical problems and understanding the business context behind them.

Software engineer in AI era are now spending less time manually writing every line of code and more time reviewing outputs, validating logic, and making decisions.

Read more: AI x Human: Productivity 10x Without Losing Human Judgment

How Are Companies Adopting AI, And Where Is Everyone Actually Stuck?

Here’s something worth being honest about: most companies think they’re doing AI well. Most of them are only at the beginning.

Tung Nguyen, Head of Delivery at Enosta, laid out a picture of how AI adoption actually progresses, and why so many teams are stuck at level one without realizing it. He described AI adoption as happening on 5 levels.

The gap between “everyone is using AI individually” and “the whole company works in an AI-native way” is bigger than it looks. Some engineers have figured out great personal workflows. Others barely use AI. The result is wild inconsistency in speed, in quality, in how work gets reviewed and handed off.

Nguyen Duc Nguyen at MADLAB shared the solution to address this, which is moving from individual prompting and individual context-building toward spec-driven development, where the entire project context is managed centrally and maintained by the whole team.

When that foundation is in place, individual engineers can prompt with simple instructions and still get results that follow the team’s agreed standards.

Read more: AI for Business: Pragmatic Strategies for Real Impact

What Does A Great Engineer Look Like in the AI Era?

If you ask the panelists what they’re looking for when they hire or develop engineers now, the answers converge around three things.

• First: Technical depth is still essential

A lot of engineers have started to think: if AI writes the code, why do I need to go deep technically?

Because you need to review what AI writes. And reviewing requires real knowledge. If you don’t know how a system should behave, you can’t tell when AI has produced something subtly wrong. You can’t catch security issues. You can’t spot architectural decisions that look fine now but will create problems in six months.

• Second: Think across the whole system and the business.

Binh Duc Quan, CTO of Mynavi TechTus Vietnam, introduced a term that captures this well: the “intentional thinking full-stack developer”, means full-stack with purpose.

The idea isn’t just to be able to work across front-end and back-end. It’s to be someone who thinks deliberately about what they’re building and why, and who brings that intentionality to every layer of their AI usage.

“To be full-stack, you need to come back to system thinking – thinking more for the clients, not limiting yourself to mobile, front-end, or back-end. At the same time, you still need to go deep on the technical side and build a solid AI foundation – from knowing how to use AI to using it at a higher orchestration level.” Binh mentioned

Binh also pointed to something specific about outsourcing work that makes this even harder: when AI comes into a legacy system, engineers need to be able to reverse-engineer it, document the information, organize all the context, and even distinguish between what’s still in use, what should keep evolving, and what’s already outdated.

• Third: Understand how AI workflows actually work, not just how to use the tools.

Nguyen Duc Nguyen from MADLAB pointed out that one of the biggest mistakes engineers make today is treating AI as purely a technical tool.

According to him, if engineers only provide technical inputs to AI without understanding the business side of the problem, the outputs are rarely useful in the real world. That is why MADLAB is increasingly looking beyond technical knowledge during interviews. Instead of only asking about frameworks or coding skills, the team also evaluates how candidates think about business problems, workflows, and real-world use cases.

For students, that might mean understanding how a school management or library system actually works. For experienced engineers, it means being able to explain not just what they built, but how they understood the business needs behind the project.

The bigger point was simple: in the AI era, business context is becoming part of technical skill itself.

How Is AI Changing Business Model & Organizations?

The effects of AI aren’t just showing up in how individual engineers work. They’re reshaping how companies price their services, structure their teams, and what clients demand from them.

The productivity paradox: faster delivery, more pricing pressure.

The business impact is becoming impossible to ignore. Most tech companies charge by time and effort. If AI makes engineers 30-40% more productive – and some companies are already measuring exactly that – the math becomes awkward. You can do the same work with fewer people, or deliver faster for less money. Clients know this. They’re using AI themselves to estimate project timelines and costs, and they’re pushing prices down.

This puts companies in a bind. They need to adopt AI to stay competitive on speed. But the faster they go, the more pricing pressure they face.

The only real exit is to compete on something beyond speed – architecture, strategy, governance, judgment. The companies that figure out how to sell that are the ones that will survive the transition.

Enterprise clients are getting more demanding.

Ha Manh Ngo, CEO of Xbrain and Vietnam’s sole Anthropic partner, described what the banks and large financial institutions he works with are now asking for. They’re not asking for AI tools or AI-assisted features anymore. They want full AI platform architectures – a centralized layer that governs how AI is used across the entire organization.

“The large enterprises now need: if you’re a service provider going into AI, you need to advise me on how to build an AI Platform or ML Platform so that all the AI applications, use cases, and centralized data can be used efficiently – and it covers the security problem, the exploitation problem, the centralized governance and permissions problem, the control gate and quality gate problem.” – Ha Manh Ngo, General Manager / CEO, Xbrain

This is a significant shift. Building an AI platform architecture requires a different kind of engineer – someone who understands data pipelines, governance, security, and organizational processes, not just code.

Large organizations face real inertia.

Duy Khanh Bui, Delivery Unit Director at CMC Global, walked through what makes AI transformation hard at scale. He compared a 3,000-person company to an airplane: hard to turn around quickly.

The specific pain points: legacy systems with deeply tangled processes that need input from many departments before anyone can decide where to apply AI. Culture and morale – easy to roll out for 10 or 30 people, much harder for thousands.

Privacy and security constraints from Japanese and Korean clients whose data is government or banking data – some of them ban AI usage entirely and require signed NDAs. And the fundamental ROI question: before committing to company-wide AI, leadership needs to be able to measure whether it actually works.

“One of the biggest challenges is still the ROI question. Leaders need to know what AI investment actually returns – how productivity improves, how much time is reduced, and whether those results can be measured and verified across projects”, Duy shared

He stated that transformation needs to happen layer by layer, from the CEO setting a strategic direction, to the CTO defining the path, to the delivery units testing in the real context, to a dedicated AI advisory team collecting know-how and data, all the way down to vendor cooperation for private, secure model deployment.

The Future Career Path: What Young Talents Need to Prepare?

A common fear among students and junior software engineers in AI is that AI will make their entry-level roles obsolete. The panelists offered a nuanced correction: technical depth matters more than ever, but its application has changed.

AI won’t eliminate the need for engineers. It will, however, raise the bar for what a good engineer looks like.

Technical intuition is still something AI can’t replicate.

AI can generate code, but it can’t develop the intuition that comes from real technical experience — knowing when something looks wrong, sensing an architectural risk before it becomes a problem, and understanding why a system behaves unexpectedly.

That intuition is exactly what’s needed to review AI output well. Technical skills are no longer the end product of an engineer’s value. They’re the foundation for judgment.

There are two kinds of AI opportunities – and engineers have access to the better one.

Everyone in every role will need to learn AI utilization and how to use AI tools to work more efficiently. That’s table stakes, and it’s not unique to engineers. But there’s a second layer that only engineers can access:

“AI development, that’s something management can’t do. Leadership can’t do it. Businesses can’t do it. Marketing can’t do it. Only you can do it. You’ll be able to understand how AI works, what it means when an agent appears on screen, and how multi-agent systems work.” – Ha Manh Ngo, General Manager / CEO, Xbrain

The playing field is more even than you think.

Concepts companies are urgently hiring for right now – AI platform architecture, multi-agent orchestration, spec-driven development, and AI governance – are things that even experienced practitioners only started learning one or two years ago. Nobody has a ten-year head start on this. The window is genuinely open.

Key takeaways for Stakeholders

Conclusions

AI is not really about replacing people, but it’s definitely raising expectations. The industry still needs engineers, probably more than ever, but it needs engineers who can think critically, understand business problems, guide AI effectively, and make smarter decisions.

The software engineer in AI era can thrive if they combine technology, systems thinking, business understanding, and human judgment together.

A recap from the DevDay 2026 panel: “From Developer to Strategist: How AI Is Redefining Roles, Teams, and Career Paths in Technology.”