Building AI Capability: Part 3
Understanding which skills and expertise matter as AI transforms software development.
AI can generate code, identify bugs, write tests and assist with increasingly complex development tasks. What once required hours of manual programming can sometimes be completed in minutes.
So, where does that leave software developers?
According to the 2025 Stack Overflow Developer Survey, 84% of respondents were using or planning to use AI tools in their development process. Yet, despite widespread adoption, trust remains a challenge: 46% actively distrust the accuracy of AI-generated outputs.
For businesses investing in software development, this raises an important question: If AI can write the code, what should we be hiring developers to do?
In the first two articles of our Building AI Capability series, we explored how Australian and New Zealand companies can access specialist AI talent in Vietnam and build integrated, long-term development teams.
Now, we turn to the changing role of the developers themselves, and why experienced technical talent remains important in an AI-assisted development environment. How is AI changing the value of software developers and the skills companies should prioritise when hiring?
AI-assisted development has moved well beyond simple code suggestions.
Developers increasingly use AI tools to generate functions, troubleshoot problems, create documentation, write automated tests and accelerate routine programming tasks.
According to Google's 2025 DORA State of AI-assisted Software Development Report, AI acts primarily as an amplifier. It can strengthen already effective development organisations, but it can also magnify existing weaknesses in processes, coordination and delivery.
In other words, generating more code does not automatically mean delivering better software.
The 2025 Stack Overflow Developer Survey illustrates this challenge:
These findings suggest that although AI can accelerate individual tasks, organisations still need the technical expertise to evaluate outputs, understand dependencies and ensure that systems work reliably.
FWH's CEO, Øystein, has experienced this shift firsthand. Rather than reviewing AI-generated code line by line, he increasingly focuses on the bigger picture: Does the solution work? Do the components interact correctly? Does the overall system make sense?
He sees AI as another member of the development team, with developers increasingly responsible for directing its work, evaluating solutions and making architectural decisions.
As AI takes on more of the coding, the developer's value increasingly lies in understanding how everything fits together, not just writing the code itself.
Traditionally, software developers have been evaluated partly on their programming knowledge, technical proficiency and ability to implement features.
Those skills remain relevant. However, as AI takes on more routine coding activities, the value of developers increasingly extends beyond producing code.
Consider the difference between generating a working function and building a production-ready system.
AI may generate a technically correct piece of code, but that does not guarantee it fits the wider architecture, handles unexpected scenarios, protects sensitive data or solves the intended business problem.
This is where experienced developers and technology experts contribute.
System architecture and integration
Understanding how applications, infrastructure, databases, APIs and services interact, and ensuring that changes work across the entire system.
Problem-solving and technical judgement
Translating business requirements into practical technical solutions, evaluating alternatives and recognising when an AI-generated approach is unsuitable.
Quality assurance and reliability
Testing functionality, identifying potential vulnerabilities, validating outputs and ensuring software performs as intended under real-world conditions.
Business and product understanding
Recognising why a feature is being developed, how customers will use it and whether the proposed solution creates meaningful value.
Security, governance and accountability
Ensuring that software and AI-enabled systems meet organisational requirements for security, privacy, maintainability, ethical and responsible use.
These capabilities are not exclusive to senior developers. However, experience often becomes particularly valuable when decisions involve complex systems, competing priorities and long-term consequences.
If AI can assist with many routine programming tasks, companies may need to reconsider how they evaluate technical talent.
The ability to write code quickly is no longer sufficient on its own.
When recruiting experienced software developers, CTOs and technology leaders should increasingly consider:
This does not mean every company needs to recruit AI researchers or machine learning specialists.
A business integrating AI into an existing application may need experienced backend developers, API specialists, data engineers or software architects rather than a dedicated AI research team.
The right combination depends on what the organisation is actually trying to build.
There is another important distinction between making individual developers more productive and making an entire development organisation more effective.
AI can help someone complete a task faster. But software delivery still depends on collaboration, shared technical understanding, coordination and effective decision-making.
In a 2026 DORA analysis, Google researchers observed that time saved through AI-assisted code generation can be redirected towards auditing and verification. Their research also found that greater AI adoption was associated with increased software delivery throughput alongside increased delivery instability.
This highlights why productivity cannot be measured simply by the volume of code produced or the number of tasks completed.
For companies building distributed development teams, including those working across Australia, New Zealand and Vietnam, effective integration becomes particularly important.
Developers need access to the same product context, architectural decisions, documentation and organisational priorities.
Otherwise, faster individual output may create additional complexity for the wider team.
At FWH, we believe that building effective development capability requires more than access to individual technical skills.
Our approach is to recruit experienced technology experts in Vietnam who work exclusively with each client as part of their existing organisation.
This model becomes especially relevant as AI changes development workflows.
Developers who understand a company's products, systems, domain, customers and technical history are better positioned to evaluate how new AI capabilities fit into the wider organisation.
That understanding develops over time.
A developer who has worked on the same product for several years may recognise architectural dependencies, recurring challenges and business requirements that are not immediately apparent from a codebase or an AI-generated suggestion.
At FWH, our clients stay with us for an average of eight years, reflecting the long-term nature of these partnerships.
The objective is not simply to provide additional programming capacity, but to help organisations build technical teams that can adapt as technologies and business requirements change.
For ANZ companies considering international recruitment, this is an important distinction between accessing developers for individual projects and investing in an integrated development team.
AI is unlikely to change every development role in the same way.
Routine programming tasks may become increasingly automated, while responsibilities involving architecture, system integration, technical evaluation and product understanding may take on greater importance.
For some organisations, this could change the balance between junior and senior roles. For others, AI may allow smaller teams to deliver more, provided they have the expertise and processes to manage the resulting systems.
What remains important is the distinction between producing software and producing software that works reliably, securely and effectively for its intended users.
As AI continues to evolve, businesses should therefore look beyond how much code their developers can write.
The more useful questions are:
Can our developers understand the system as a whole? Can they evaluate what AI produces? Can they make sound technical decisions? And can they connect technology to the needs of our business?
Because in an environment where generating code is becoming easier, knowing what to build, how it fits together and whether it actually works may become more valuable than ever.
Part 1: How to Hire AI Developers in Vietnam in 2026 / Explore Vietnam's AI talent market, expert roles and hiring considerations for ANZ companies.
Part 2: How to Build an AI Development Team in Vietnam in 2026 / Learn how team structure, integration, retention and the right development partner contribute to long-term AI capability.
This article draws on the 2025 Stack Overflow Developer Survey, covering AI tool adoption, developer trust, productivity and challenges with AI-generated code; Google's 2025 DORA State of AI-assisted Software Development Report, examining how AI influences software development performance and organisational effectiveness; and DORA's 2026 analysis, Balancing AI Tensions: Moving from AI Adoption to Effective SDLC Use, exploring the relationship between AI-assisted development, verification, productivity and delivery stability.
*FWH-specific details and figures are based on internal company data.