Business leaders are being told AI will replace their development teams. Make everyone 10x more productive. Eliminate the need for senior engineers.

Some of this is true. Most of it is dangerously misleading.

Here’s what you actually need to know — and what it means for how you structure your teams.

The demo isn’t the product

You’ve seen the demos. “I built this app in one prompt.” Impressive. Misleading.

One-shot treats AI like a magic genie. Describe what you want perfectly, and get exactly what you need. But AI can generate code that works without grasping the What and Why behind a complex business problem.

That understanding isn’t discovered in prompts. It’s discovered in iteration — by building something, watching it fail, and building again. One-shot assumes perfect knowledge upfront. It’s waterfall wearing a hoodie.

Real AI Augmented Development is something different. It’s AI as a teammate — working alongside experienced practitioners through the full development cycle. Tools like Claude Code let developers stay in flow, iterating rapidly without context switches that break concentration. Not a tool you reach for occasionally, but a collaborator integrated into how you think, plan, and build.

What AI actually replaces

To understand what’s changing, you need to understand how software engineers contribute.

Most knowledge work — including software development — follows a progression that goes back to the guild system:

• Directed contribution. You’re given a specific, well-defined task in an unambiguous context. “Implement this spec.” “Build this API endpoint exactly as described.” You don’t need to understand the What and Why. You execute the How, under someone’s guidance.
• Independent contribution. You’re trusted to tackle problems independently — first in well-defined situations, then in ambiguous ones. You figure out both what to do and how to do it. You understand enough about the business to make judgment calls when the spec is incomplete or wrong.
• Working through others. You set vision and direction. You guide others. You’re accountable for outcomes, not just outputs.

Here’s what AI Coding Assistants like Claude Code do well: Directed contribution. Give them a specific, well-defined task in an unambiguous context, and they execute. Often better than a human, because they don’t get tired, don’t make typos, and don’t need coffee breaks.

This is precisely the work that large offshore development teams were built to do. And not just offshore — too many developers everywhere, even with years and even decades of work experience, including Singapore, operate in directed contribution mode.

Also Read: AI fluency or disaster: Decide before it decides for you

The model that’s dying

For decades, the dominant model looked like this: Product managers sit near the business. They write detailed PRDs and specs. Those specs get shipped over the wall to a large development team — often offshore in the Philippines, Vietnam, Indonesia, and India. Each developer gets a well-defined slice. They implement exactly what’s described. Ship it back.

The economics seemed compelling. Senior engineers in the US cost US$150,000 or more. Offshore developers cost a fraction of that. Scale up the team, ship the specs, get the code back.

But this model carried a hidden cost: the collaboration tax.

Communication gaps. Lost context. Misalignment with business needs. Revision cycles. Specs that are outdated before they’re implemented. The PM who wrote the requirements isn’t sitting with the developers who have questions.

Research on proximity and collaboration is unambiguous. A University of Michigan study found that researchers on the same floor are 57 per cent more likely to collaborate than those in different buildings. For every 100 feet of shared walking path, collaborations increased by 20 per cent. MIT research on the “Allen Curve” — named after MIT professor Thomas Allen — shows that even basic conversations become much less likely when workers are more than 10 meters apart.

Look at startups when they’re hitting things solidly and delivering customer value. They’re sitting in one another’s laps. The communication bandwidth is massive. Questions get answered in seconds, not days.

The collaboration tax was always there. Companies accepted it because the labour cost arbitrage seemed worth it.

AI augmented development changes the equation.

The math has changed

When AI handles directed contribution, you don’t need a 20-person team executing specs. You need a small team of experienced practitioners who can work with AI to iterate rapidly on complex problems.

Consider the economics:

• Traditional model: 1 PM + 15-20 developers. Lower cost per person, but high headcount, high collaboration tax, slower cycles, and lower alignment with business needs.
• Emerging model: 1 technically-fluent AI PM + 3-4 senior co-located engineers, all working with AI tools. Higher cost per person, but dramatically fewer people. Lower total cost. Faster cycles. Higher quality. Better business alignment.

The smaller team isn’t just cheaper. It’s better.

AI augmented development compresses the cycle from weeks to hours. A working prototype can replace a 50-page PRD. Instead of describing what you want the software to do, you can show it — then iterate based on reality rather than imagination.

At Apple, we had a saying: Demo beats deck. A working demonstration trumps a polished presentation every time. AI augmented development is that principle writ large. When you can produce a working prototype in hours, why spend days writing a document describing what it should do?

But this only works with high-bandwidth collaboration. Tight feedback loops. The ability to walk through a prototype together, ask questions, and make changes on the spot. You can’t do that across communication gaps — whether those gaps are time zones, organisational silos, or simply being on different floors.

Also Read: Building with intention: The ethical dilemma of AI innovation and responsible creation

The what/why/how blur

Historically, product management owned the What and Why. Developers owned the How.

Those lines are blurring.

When AI can generate working prototypes from descriptions, the distance between “what we want” and “how it works” collapses. Product managers get closer to the How. Developers get pulled into the What and Why.

This isn’t a threat. It’s an evolution.

The technically-fluent AI PM isn’t someone who writes PRDs and waits for engineering. They’re producing prototypes that aren’t always throwaway demos — they’re starting points that engineering extends. A technically-fluent PM can prototype a feature in an afternoon, walk engineering through it, and iterate together — rather than writing a 20-page spec and waiting two sprints to see if engineering understood it. They understand the How well enough to make informed tradeoffs.

And developers now need to understand the What and Why deeply enough to make judgment calls when iterating. “This requirement doesn’t make sense given what I understand about the user” — because they do understand the user.

Everyone needs more business context. Everyone needs more technical fluency. The boundaries are dissolving.

Seniority isn’t what you think

Here’s where business leaders get confused.

“Our team has senior developers. They have 10 years of experience.”

But years of experience aren’t the same as how someone contributes.

Someone with 10 years of experience doing directed contribution work isn’t a senior developer. They have one year’s experience, executed under someone’s guidance, ten times over.

This isn’t just an offshore problem. Too many developers in Singapore, in London, in San Francisco, have spent careers in directed contribution mode — not because they were incapable of more, but because the organisations they worked for didn’t ask more of them. The PRD comes over the wall. They implement their slice. Ship it back. Repeat. And everyone prays that it all comes together and works.

Ten years of this doesn’t develop the skills the new model demands.

What AI augmented development requires is Independent Contribution in Ambiguous Settings. People who understand the business problem — the What and Why, not just the How. People who can make judgment calls when the spec is incomplete or wrong. People who can collaborate at high bandwidth and low latency because they share context with the business.

What this means for Southeast Asia

This isn’t about reshoring jobs to the US. It’s about the death of the human wave model that much of Southeast Asia’s software outsourcing industry was built on.

Vietnam produces 50,000 IT graduates annually. Over 45 per cent of its developer workforce is at the junior level — trained to do directed contribution work. The Philippines has built a massive tech services industry on similar foundations.

The question for the region isn’t whether AI will disrupt the traditional outsourcing model. It already is. The question is whether Southeast Asia can compete on value, not low-cost volume.

Can the region produce engineers who operate in Independent Contribution mode? Engineers who understand the What and Why, not just the How? Engineers who can be part of elite co-located teams — whether those teams sit in Singapore, Jakarta, Ho Chi Minh City, or alongside clients in Tokyo, Sydney, or San Francisco?

The opportunity isn’t to fight the transformation. It’s to ride it.

Also Read: AI’s reality check: Why 95 per cent of pilots fail and how to measure what actually matters

What business leaders should do

Audit your teams — not for years of experience, but for mode of contribution. How many of your people are doing directed contribution work that AI can now handle? How many can operate independently in ambiguous situations? How many understand the What and Why of your business, not just the How of their technical domain?

Rethink your team structure. Smaller. Co-located. Located where the business sits. The two-pizza team that Amazon pioneered is finally becoming real — but it only works when the team has the proximity and bandwidth to collaborate intensively. Higher cost per person, lower total cost, better outcomes. And maybe you go for a one-pizza team!

Invest in AI fluency — not just tool access. Throwing AI tools at people without helping them understand what AI can and cannot do is setting them up to fail. The failures we’ve seen — like Deloitte’s fabricated citations in government reports — come from people who knew enough to be cautious but not enough to be fluent.

And think hard about your talent pipeline. Entry-level tech hiring has collapsed — junior developer roles are down 60 per cent since 2022. The apprenticeship ladder that produced your current senior engineers is disappearing. Where do your future senior engineers come from if you’ve eliminated the model that trained them?

The bottom line

AI augmented development favours smaller teams of independent contributors who understand the business — co-located where the business sits. Higher cost per person, dramatically fewer people, better outcomes.

The human wave is dying. The question is whether you’re building the team that replaces it or the team that gets replaced.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.

Enjoyed this read? Don’t miss out on the next insight. Join our WhatsApp channel for real-time drops.

Image generated using AI.

The post AI augmented development: Hype vs reality appeared first on e27.