The Future of AI: Opportunities, Challenges, and Impact on the Labor Market

Introduction: Where Are We on the AI Journey?

If someone asked me whether 2023 was the year of Generative AI, the answer would be yes. But if they asked whether 2025–2026 marks the moment AI begins to “come of age” — then the answer is also yes, and far more emphatically so.

We are living in an era where the pace of technological change is no longer measured in years, but in months. Since the launch of ChatGPT in late 2022, the world has witnessed an unprecedented shockwave: GPT-4, Claude, Gemini, Llama, Qwen, and hundreds of open-source models. What matters isn’t the sheer quantity, but the nature of the transformation itself.

AI is no longer just a keyboard tool that answers questions. It is becoming an entity capable of planning, decision-making, and acting on behalf of humans in specific tasks. The biggest question I — and perhaps you — keep asking ourselves is: Where do we go from here?

The difference between 2023 and today is stark. Back then, AI was a novelty — something to experiment with, to marvel at. Today, it is embedded in workflows, integrated into enterprise software, and increasingly trusted with mission-critical decisions. We have moved from “wow” to “what’s next?”

2. Agentic AI — The Next Wave

When most of us think of AI, we still picture a chat box: you type a question, it replies. But that’s only the surface layer. Beneath it, a quiet revolution is underway — the revolution of Agentic AI.

From Passive Chatbots to Proactive AI

Imagine an AI that doesn’t just tell you how to book a flight — it actually opens your browser, visits the airline’s website, compares prices, books the ticket, and adds the itinerary to your Google Calendar. That is an AI Agent.

Unlike traditional chatbots — which only respond when asked — an AI Agent possesses three core capabilities: (1) Planning — autonomously breaking down a large goal into smaller steps and executing them in sequence; (2) Using tools — calling APIs, browsing the web, running code, and interacting with software interfaces; (3) Decision-making — analyzing intermediate results, evaluating multiple options, and adjusting the approach when something goes wrong.

Platforms like LangChain, AutoGPT, Claude Computer Use, OpenAI Operator, and Manus are turning this concept into reality. The 2025–2026 period is widely predicted to be the era of “Agentic Workflows” — processes that no longer require a human in the middle at every step.

Real-World Applications

Consider a small-to-medium e-commerce company in Vietnam. Instead of hiring three customer service representatives, a data entry clerk, and a marketing assistant, they could deploy an AI Agent system: one agent handles order processing (checking inventory, reconciling payments, sending confirmation emails); another analyzes data (reading sales reports, recommending restock priorities and identifying slow-moving inventory); and a third manages marketing (creating ad content, running A/B tests on Facebook and Google, and optimizing campaign budgets in real time). This isn’t science fiction — it’s all feasible today.

In logistics, AI agents can coordinate supply chains — tracking shipments, predicting delays, rerouting goods, and communicating with suppliers — with minimal human oversight. In healthcare, agents can schedule appointments, send reminders, triage symptoms, and follow up with patients automatically.

Creative AI and Content Creation

Beyond automation, AI is revolutionizing how content is produced. Text generation, image synthesis, music composition, and video creation are now accessible to anyone with an internet connection. Voice synthesis tools like ElevenLabs are making AI narration indistinguishable from human voiceovers, enabling creators to produce audiobooks, podcasts, and multilingual content at a fraction of the traditional cost. This democratization of creative tools is reshaping entire industries — from marketing agencies producing personalized ad campaigns to indie game developers generating full voice casts for their characters.

The Multi-Agent Paradigm

Perhaps the most exciting development is the emergence of multi-agent systems. Instead of a single AI tackling a complex problem, teams of specialized agents collaborate — each handling a domain it excels at. One agent researches, another drafts, a third verifies facts, and a fourth polishes the final output. This division of labor mirrors how human teams work, and early results suggest it dramatically improves output quality and reliability.

3. On-Device AI and Edge Computing

While Agentic AI captures the tech world’s attention, another wave is quietly transforming how everyday users interact with AI: AI that runs directly on your device.

Apple Intelligence and the On-Device Race

In late 2024, Apple officially announced Apple Intelligence — an AI system integrated directly into iOS, iPadOS, and macOS, running primarily on the device’s Neural Engine. This means: no need to send data to the cloud, no network latency worries, and most importantly — personal data stays on the user’s machine. Google and Samsung are not standing by either. Gemini Nano runs on Pixel and Galaxy S series devices, while Qualcomm continues to push AI capabilities on the Snapdragon 8 Gen 3 chip. The competition is fierce, and consumers are the ultimate beneficiaries.

Why On-Device AI Matters

Three key reasons make this shift significant:

• Privacy — nobody wants to record a private meeting and send it to a third-party server for processing. On-device AI keeps sensitive conversations and personal data exactly where they belong: on the user’s device. This is especially critical in regulated industries like healthcare and finance.
• Speed — voice processing on-device takes milliseconds instead of seconds over the network. Real-time language translation, live captioning, and instant photo editing become genuinely seamless when there is no round-trip to the cloud.
• Offline capability — rural Vietnam, remote areas with weak or intermittent internet — on-device AI brings technology to everyone, regardless of connectivity. This closes the digital divide in a way cloud-dependent solutions never could.

Smart Wearables

Meta Ray-Ban smart glasses, Nothing Ear earbuds with ChatGPT integration, and devices like the Humane AI Pin and Rabbit R1 have opened a new frontier: AI as a constant companion — always present, always listening, always ready to assist. These form factors represent a fundamental shift in how we interact with AI: from typing queries on a screen to natural, hands-free conversation. The implications for accessibility are profound — AI becomes useful not just for tech-savvy professionals, but for elderly users, people with disabilities, and anyone who finds screens cumbersome.

4. Major Challenges

If AI were all sunshine and rainbows, this article would have ended at Section 2. The truth is, the challenges AI presents are every bit as formidable as the opportunities — and ignoring them would be irresponsible.

4.1. Energy — How Much Power Does AI Really Consume?

A single ChatGPT query consumes approximately 10 times the energy of a standard Google search. Multiply that by billions of queries per day, and the energy picture becomes alarming.

According to Goldman Sachs (2024), electricity demand from AI data centers is projected to increase by 160% by 2030. Microsoft, Google, and Amazon are scrambling to secure renewable energy sources — Microsoft even signed a deal to restart the Three Mile Island nuclear plant. Training a single large model can emit as much carbon as five cars over their entire lifetimes. Can the planet sustain AI at a global scale? That question remains unanswered, and the answer will shape the industry’s future direction.

Promising solutions are emerging: more efficient chip architectures, model quantization, distillation techniques, and the growing adoption of renewable energy. But these solutions need to scale far faster than they currently are.

4.2. Regulation — The Race Between Speed and Safety

The EU leads the way with the AI Act — the first comprehensive legal framework classifying AI risk from low to high, imposing strict requirements on high-risk applications. The US has issued an Executive Order on AI safety and is debating federal legislation. China is tightening controls over algorithms and AI-generated content, requiring labeling and approval for certain AI services.

The problem: regulation always lags behind technology. By the time laws are drafted, debated, and enacted, the technology has already moved three steps forward. How do we encourage innovation while protecting citizens from harm? There are no easy answers. Overregulation risks stifling progress and ceding leadership to less scrupulous players; underregulation risks real damage to individuals and society.

4.3. Ethics — Bias, Deepfakes, and AI Safety

AI is trained on human data — and that data is full of biases, prejudices, and blind spots. A hiring model favoring male candidates over equally qualified women? It’s already happened — multiple times. Your voice and face deepfaked without consent? It happens every day, fueled by increasingly accessible tools. An AI “jailbroken” into bypassing its safety guardrails? This is an endless arms race between attackers and defenders.

The most concerning dimension is the erosion of trust. When AI-generated content is indistinguishable from human-created material, how do we know what is real? The responsibility lies not just with developers, but with regulators, educators, and end-users to build a culture of critical consumption.

4.4. Cost — Is This a Game Only Big Tech Can Play?

Training GPT-4 reportedly cost an estimated $100–200 million. Running it daily costs millions more in compute resources. Is AI destined to be a playground for giant corporations only? Fortunately, the tide is shifting. Open-source models like Llama, Qwen, Mistral, and DeepSeek are steadily closing the performance gap with proprietary systems. Startups can fine-tune these models for just a few thousand dollars, bringing cutting-edge AI capabilities to organizations of all sizes.

The real competitive advantage is shifting from model size to data quality, application design, and user experience — areas where innovators can compete regardless of their compute budget.

5. Impact on the Labor Market

This is perhaps the most anxiety-inducing topic — and also the most misunderstood. Headlines scream about job apocalypse, but the reality is far more nuanced.

WEF Forecast: 97 Million New Jobs, 85 Million Lost

The World Economic Forum’s “Future of Jobs 2025” report projects that by 2030, AI and automation will eliminate 85 million jobs while simultaneously creating 97 million new ones. The key message from this data: AI won’t just take your job — AI will transform your job. And those who fail to adapt will be left behind.

This is not the first time technology has reshaped the labor market. The industrial revolution displaced agricultural workers but created entire categories of factory jobs. The internet age decimated travel agencies and brick-and-mortar retail while birthing e-commerce, digital marketing, and social media management. AI is following the same historical pattern — destruction and creation happening side by side.

Which Industries Are Safe? Which Are at High Risk?

• High Risk — roles involving repetitive, rule-based tasks: data entry, administrative text processing, basic translation, tier-1 customer support, manufacturing assembly, and basic accounting. These jobs are most susceptible to automation, and the displacement is already underway.
• Relatively Safe (AI assists but cannot fully replace human judgment): software developers, doctors, lawyers, and teachers. AI can write code, diagnose conditions, and search case law effectively — but architectural design, patient interaction, strategic counsel, and human inspiration still require human expertise. The professionals who thrive will be those who learn to collaborate with AI rather than compete against it.
• Very Safe — roles requiring deep human connection and complex physical manipulation: psychologists and social workers; original creative artists whose value lies in unique perspective; electricians and plumbers who work with unpredictable physical environments; and senior leaders and high-level strategists who navigate ambiguity and organizational dynamics.

New Skills to Learn

We are entering an era where “AI literacy” is becoming a foundational skill — as essential as reading, writing, and using a computer. Emerging competencies include:

• Prompt Engineering — the art of asking the right questions and providing the right context to get accurate, useful AI responses
• AI-Augmented Workflow Design — building processes that effectively combine human and AI strengths, knowing when to automate and when to involve people
• Critical Thinking — when AI can generate any content convincingly, evaluating authenticity, spotting hallucinations, and questioning sources becomes more important than ever
• AI Ethics — understanding bias, fairness, transparency, and responsible AI deployment, a skill increasingly demanded by employers and regulators alike

6. Opportunities for Vietnam and Startups

Vietnam stands at a historic crossroads — and the question is whether we have the courage and vision to seize the moment.

AI Outsourcing — A Natural Stepping Stone

Vietnam is already a top destination for software outsourcing, with a well-earned reputation for quality and cost-effectiveness. The AI story rewrites this narrative: AI talent costs in Vietnam are 40–60% lower than in the US and Europe; the engineering workforce is young, hungry, and eager to learn; and an AI startup ecosystem is gradually taking shape with names like VNG and FPT AI leading the charge. Global companies are increasingly looking to Vietnam not just for coding talent, but for AI research, data annotation, and model fine-tuning.

Applying AI to Traditional Industries

The strength of Vietnamese startups lies in their deep understanding of the local market and its pain points. Rather than trying to compete with Silicon Valley giants on foundational models, they should focus on applied AI in sectors where local knowledge is a moat:

• Agriculture — AI-powered weather forecasting and pest detection via satellite imagery and drone data, helping farmers reduce crop loss and optimize yields
• Healthcare — AI diagnostic imaging to support district-level doctors who may lack access to specialist radiologists
• Education — personalized AI tutors that adapt to each student’s learning pace, bringing quality education to rural and remote areas
• Finance — AI credit scoring for the unbanked population, using alternative data (mobile usage, payment history) to extend financial services to millions

Products for the Southeast Asian Market

Southeast Asia — with 680 million people, a young demographic, and rapidly growing digital adoption — is a market full of gaps waiting to be filled. Vietnamese startups have distinct advantages: deep cultural and linguistic understanding of the region; lower operating costs compared to Singapore or Thailand; and a growing pool of venture capital flowing into AI across the region.

A healthcare AI app available in Vietnamese, Thai, and Indonesian — that is a product with pan-regional potential. An AI-powered fintech solution designed for the unbanked populations of Cambodia, Myanmar, and the Philippines — that could reach hundreds of millions of users.

7. Conclusion: Informed Optimism

I am not a believer in a utopian AI future where robots do everything and humans simply enjoy the fruits of their silicon labor. But neither do I subscribe to a bleak vision where AI steals everyone’s livelihoods and leaves society in ruins. The truth, as always, lies somewhere in between — and it is a place of enormous possibility.

AI is a tool — perhaps the most powerful tool humanity has ever created. But it remains a tool. Like fire, the wheel, the steam engine, or the internet — it can bring both immense good and tremendous harm. The outcome depends entirely on how we choose to design, deploy, and govern it.

What makes me genuinely optimistic is this: never before in history have so many humans had access to so much opportunity to learn, adapt, and create. With AI, a child in rural Vietnam can learn programming guided by one of the world’s most capable AI tutors — for free. A farmer can know exactly when to water crops and apply fertilizer thanks to data-driven weather and soil analysis. A young doctor in a provincial clinic can consult millions of medical cases and research papers in an instant.

We cannot stop the AI wave. But we can learn to ride it. The question is not “What will AI do to us?” but “What will we do with AI?” And the answer, as always, lies in our own hands.

This article is part of the AI market analysis series on ai1102.vip