Across Southeast Asia and India, a quiet revolution is unfolding not in the replacement of human minds, but in their augmentation. Agoda's 2025 survey of over 600 developers reveals that artificial intelligence has become nearly universal in the region's software workflows, yet its role remains that of a capable apprentice rather than a master craftsman. The deeper story is one of pragmatic wisdom: engineers are embracing speed while guarding quality, teaching themselves in the absence of institutional support, and navigating a technology whose promise still outpaces its reliability.
AI Adoption Soars Among Southeast Asian Developers, But Maturity Lags Behind Usage
Developers are using AI pragmatically, not replacing skill with automation.
Mark
So 95% of developers are using AI weekly. That sounds like the technology has already won—it's just part of how work gets done now.
Mimi
It has become routine, yes. But routine doesn't mean transformative. Most of that use is concentrated in code generation, which is the easiest task to automate. When you move downstream—testing, deployment, documentation—usage drops significantly. Developers are selective.
Luke
Do we know why usage drops? Is it because AI is worse at those tasks, or because developers don't trust it yet, or because those tasks require different skills?
Mimi
The report points to inconsistency and unreliability as the main barrier. Seventy-nine percent cite that. And the fact that 67% review all AI code before merging suggests developers don't fully trust the output.
Mark
But they're still saving four to six hours a week. That's real.
Mimi
Absolutely. And 72% report better code outcomes overall. The human review process isn't slowing them down; it's actually strengthening the work.
Luke
That's interesting—so the constraint is creating better results. But I want to flag something: we don't know if those four to six hours are coming from AI doing better work or from developers spending less time on tasks they used to do manually. The causation isn't entirely clear.
Mark
Fair point. What about the training gap? Singapore developers getting twice as much formal training as Vietnam—that seems like it could create real inequality.
Mimi
It's already creating it. But here's the counterweight: 87% of developers are self-directing their learning anyway. They're not waiting for their employers to train them. They're teaching themselves through tutorials and side projects.
Luke
Which is admirable, but it's also a sign that organizations aren't stepping up. If developers have to teach themselves, that's a gap in institutional responsibility, even if individual initiative is filling some of it.
Mark
So the story is: adoption is universal, but maturity is uneven, and it depends partly on where you sit geographically and how much your employer invests in you.
Mimi
That's it exactly. The technology is mainstream. The question now is whether the region's organizations will build the structures to support it, or whether individual developers will keep carrying that load.
The Pulse
- Ninety-five percent of developers across seven markets now use AI weekly, with more than half running an assistant continuously throughout their workday — adoption has effectively reached saturation.
- Yet confidence lags far behind usage: only 22% trust AI when facing genuinely unfamiliar problems, and fewer than half believe it matches the judgment of a mid-career engineer.
- Inconsistent outputs are the central friction point, with 79% citing unpredictability as the main barrier to deeper reliance — a constraint developers are managing through rigorous human review rather than waiting for AI to improve.
- Productivity gains are real and measurable — 37% of developers reclaim four to six hours weekly — but they are won through discipline, with 67% reviewing all AI-generated code and 70% reworking outputs before use.
- A structural inequality is emerging beneath the surface: Singapore developers are nearly twice as likely as those in Vietnam to access formal AI training, threatening to widen capability gaps even as adoption rates across the region converge.
Across Southeast Asia and India, a quiet revolution is unfolding not in the replacement of human minds, but in their augmentation. Agoda's 2025 survey of over 600 developers reveals that artificial intelligence has become nearly universal in the region's software workflows, yet its role remains that of a capable apprentice rather than a master craftsman. The deeper story is one of pragmatic wisdom: engineers are embracing speed while guarding quality, teaching themselves in the absence of institutional support, and navigating a technology whose promise still outpaces its reliability.
Artificial intelligence has settled into the daily rhythms of software development across Southeast Asia and India, but it has arrived as an accelerant rather than a revolution. Agoda's study, surveying more than 600 developers across Indonesia, Malaysia, Singapore, Thailand, the Philippines, Vietnam, and India, finds 95% using AI weekly — with speed and automation cited by 80% as the primary motivation. The payoff is tangible: more than a third of developers save between four and six hours each week.
The technology's foothold is strongest where verification is easiest. Ninety-four percent of developers rely on AI for code generation, but adoption drops sharply for documentation, testing, and deployment — tasks where errors carry heavier consequences. Rather than replacing judgment, developers are treating AI as a first draft, with 67% reviewing every line before merging and 70% reworking outputs for correctness. Only 22% turn to AI for genuinely unfamiliar problems, and fewer than half believe it performs at the level of an experienced engineer.
The chief obstacle to deeper adoption is not distrust but unreliability — 79% point to inconsistent outputs as their main barrier. Yet this has not dampened results: 72% report clear productivity gains and improved code quality, suggesting that the habit of human oversight is itself a source of strength. Formal governance remains thin, with only one in four teams operating under official AI policies, but developers are building accountability into their workflows through peer review and validation.
Most of this learning has been self-directed. Seventy-one percent taught themselves through tutorials, side projects, and online communities, while only 28% received formal organizational training. The consequences of this gap are unevenly distributed: developers in Singapore are nearly twice as likely as those in Vietnam to access structured AI programs, a disparity that risks fracturing the region's capabilities even as adoption rates converge. Agoda's chief technology officer Idan Zalzberg captured the moment plainly — the challenge is no longer persuading developers to use AI, but building the governance, training, and organizational frameworks to sustain what they have already built for themselves.
Artificial intelligence has become a fixture in the daily work of software developers across Southeast Asia and India, but the technology remains a tool for incremental speed rather than fundamental transformation. A study released this week by Agoda, the digital travel platform, surveyed more than 600 developers across seven markets—Indonesia, Malaysia, Singapore, Thailand, the Philippines, Vietnam, and India—between August and September to understand how engineers are actually using AI in their workflows. The findings paint a picture of near-universal adoption paired with persistent uncertainty about what AI can reliably do.
Ninety-five percent of developers in the region use AI on a weekly basis, and more than half keep an AI assistant running constantly throughout their workday. The primary driver is straightforward: speed. Eighty percent of developers cite acceleration and automation as their main reason for turning to AI, and the payoff is tangible. Thirty-seven percent report saving between four and six hours each week through AI-assisted work. The technology has become most embedded in code generation—94% of developers rely on it for that task—but its utility drops sharply for downstream work like documentation, testing, and deployment. This uneven adoption pattern suggests developers trust AI most where they can verify its output quickly and least where mistakes carry downstream consequences.
The gap between adoption and confidence reveals itself in how developers actually use the technology. Only 22% turn to AI when facing genuinely unfamiliar problems, and fewer than half believe AI can perform at the level of a mid-career engineer. Instead of replacing human judgment, developers are using AI as a sparring partner—something to accelerate routine work while they maintain quality control. Sixty-seven percent review every line of AI-generated code before merging it into their projects. Seventy percent rework AI outputs to ensure correctness. This is not the story of automation replacing engineers; it is the story of engineers using automation as a first draft, then applying their own expertise to finish the work.
The barrier to broader AI use is not skepticism but unreliability. Seventy-nine percent of developers cite inconsistent or unpredictable outputs as the primary obstacle to expanding their use of AI tools. Yet this constraint has not stalled progress. Seventy-two percent report clear productivity gains and better code quality overall, suggesting that the discipline of human review actually strengthens outcomes rather than slowing them down. Formal policies to govern AI use remain rare—only one in four teams operate under official guidelines—but developers are building accountability into their workflows organically, through peer review and validation practices.
The region's developers are teaching themselves. Seventy-one percent learned about AI through tutorials, side projects, and online communities rather than through employer-sponsored training. Only 28% received formal instruction from their organizations. This self-directed learning has produced results: 87% of developers have adjusted their career plans or learning strategies to incorporate AI, and 62% expect the technology to expand their long-term opportunities. Yet access to structured training varies sharply by geography. Developers in Singapore are nearly twice as likely as their counterparts in Vietnam to have access to formal AI training programs, a disparity that risks widening capability gaps across the region even as adoption rates converge.
Idan Zalzberg, Agoda's chief technology officer, framed the moment as one of pragmatism rather than disruption. "What began as a way to speed up tasks like writing, testing, or debugging code has grown into a broader shift in how software is built," he said. "Developers are approaching AI with pragmatism—accelerating work, maintaining quality, and experimenting thoughtfully rather than replacing skill or judgment." The real challenge ahead, he suggested, is not convincing developers to use AI—that has already happened—but building the organizational structures, policies, and training frameworks to support what comes next. High adoption without mature governance and uneven access to training could leave some developers and markets better positioned than others to extract lasting value from the technology. The study was conducted in partnership with Macrame Consulting and drew on case studies from regional companies including Carousell, MoMo, Omise, and SCB 10x.
Notable Quotes
Developers are approaching AI with pragmatism—accelerating work, maintaining quality, and experimenting thoughtfully rather than replacing skill or judgment.— Idan Zalzberg, Chief Technology Officer at Agoda
Want the full story?
Read the original at Lokmat Times ↗
Technology