Author: Luan Khanh – Master of Computer Science, CEO of SpeedUP Technologies Vietnam
Luan Khanh is the Founder and CEO of SpeedUP Technologies Vietnam, a technology company established in 2007 specializing in point-of-sale Speed POS solutions, digital payments, enterprise automation, and artificial intelligence applications.
He holds a Bachelor’s degree in Computer Science, majoring in Artificial Intelligence, and a Master of Computer Science from the University of Science, Vietnam National University Ho Chi Minh City, where his research focused on Rough Set Theory and Data Mining.
With more than 25 years of experience in enterprise technology, he leads SpeedUP in developing next-generation solutions for Enterprise AI, Retail Automation, Digital Payment, and Immersion Cooling infrastructure for AI computing. His current research and business interests focus on Knowledge Transformation, Agentic AI, and building intelligent enterprises that transform data into knowledge, knowledge into decisions, and decisions into sustainable business value.
Over the past two decades, digital transformation has fundamentally altered how businesses operate. Paper records have been replaced by electronic data; cash has gradually given way to digital payments; independent stores have been connected into chain networks; and sales, logistics, finance, and customer service operations have been migrated onto centralized management platforms.
The rapid advancement of artificial intelligence is opening the next phase. This is no longer merely about digitizing another process or adding a new feature to a software suite. AI is step-by-step changing how humans utilize data, access knowledge, perform work, and make decisions.
This shift can be described as a transition from digital transformation to knowledge-based transformation (or knowledge-base transformation). While “knowledge-based transformation” is not currently a standardized legal or academic term, it is used in this article to describe the phase where enterprises transition from digitizing data and processes to leveraging AI to analyze, forecast, recommend, and support execution.
If traditional software primarily helps record what has already happened, AI can support humans in recognizing what is currently taking place, forecasting what is likely to happen next, and proposing options worth considering.
Therefore, the AI revolution is not just a technology story. It has the potential to exert far-reaching impacts on employment, education, commerce, corporate governance, the distribution of economic opportunities, and the structure of societal trust.
Computers once helped humans calculate faster. The Internet allowed information to flow more broadly. Smartphones brought digital services into nearly every moment of daily life.
AI represents another developmental leap: assisting humans in processing vast amounts of information, recognizing data patterns, generating content, planning, and executing specific task sequences within a delegated scope.
An employee can use AI to:
According to the OECD, among countries with available statistical data, the proportion of enterprises reporting AI usage increased from 8.7% in 2023 to 14.2% in 2024, and reached 20.2% in 2025. These figures demonstrate that adoption is accelerating rapidly, yet they also indicate that AI has not yet been deployed evenly across the broader economy.
AI can thus be viewed as a new form of intellectual infrastructure. Similar to data, cloud computing, or payment infrastructure, AI will increasingly become a foundational layer of capability embedded into the daily operations of enterprises.
However, AI does not automatically create value simply because an organization owns a model or a tool. Value emerges only when AI is connected to reliable data, clear processes, domain expertise, and human accountability mechanisms.
One of the greatest concerns today is whether AI will replace humans.
The reality is likely more nuanced than the binary framing of “replace or not replace.” An occupation typically consists of various distinct tasks. AI can automate certain tasks, augment others, and simultaneously generate entirely new responsibilities.
The International Labour Organization (ILO) notes that approximately one-quarter of global jobs have a degree of exposure to generative AI. However, the ILO emphasizes that job transformation is a more probable impact than the total automation of entire occupations. Clerical and administrative roles continue to exhibit the highest exposure, while the degree of impact is also rising in professional occupations with highly digitized workflows.
The impact of AI on the labor market can be conceptualized across three categories:
The World Economic Forum’s Future of Jobs Report 2025, compiled from the insights of over 1,000 businesses representing more than 14 million workers across 55 economies, identifies AI and Big Data, networks and cybersecurity, along with technological literacy, as the three fastest-growing skill categories. Concurrently, analytical thinking, creativity, adaptability, leadership, and collaboration remain core human competencies.
This demonstrates that the future of employment is not simply a race between humans and machines. It will most likely be a re-allocation of tasks between humans and AI systems.
The future of work is not simply human versus machine, but human working with machines.
In such an environment, workers need more than just the ability to operate tools. They must know how to ask the right questions, verify outputs, identify anomalies, combine AI capabilities with domain expertise, and take responsibility for the final decisions.
In traditional organizations, an employee’s capacity relies primarily on their personal knowledge, experience, and the tools provided to them.
In an AI-enabled enterprise, an employee can be supported by multiple specialized assistants:
Under this paradigm, the actual performance capability of a position is formed by three components:
Human Capability + AI Capability + Enterprise Data
A sales representative can instruct AI to synthesize a customer’s history, analyze their needs, prepare meeting briefs, and recommend optimal product configurations.
An HR specialist can leverage AI to standardize job descriptions, summarize applicant profiles, verify timekeeping data, or flag payroll discrepancies. However, AI should not independently make hiring, disciplinary, or termination decisions.
A manager can pose natural-language queries:
“Which branch experienced an unusual drop in revenue over the past three months, what underlying causes need investigation, and which action items should be prioritized?”
AI can drastically compress synthesis and analysis time. Yet, the accuracy of the answer remains dependent on input data quality, model selection, business logic rules, and human verification.
Therefore, AI should be viewed as a decision support system, not a default substitute for managerial responsibility.
Traditional software operates in a reactive chain:
Human enters data → Software processes → Human reads report → Human decides
With AI agents, this operational flow transforms:
AI monitors data → Identifies issues → Analyzes causes → Proposes actions → Requests approval → Executes within scope → Monitors outcomes
An AI agent does not merely generate text answers. When properly designed, it can leverage authorized tools, access permitted data sources, and execute a sequence of tasks to achieve a designated objective.
For instance, in the retail sector, an AI agent can:
Crucially, AI agents must function within clearly defined boundary conditions. Data access permissions, financial thresholds, execution scopes, approval criteria, and audit logs must be engineered from the outset.
Consequently, transitioning from traditional software to agentic systems is not merely a technical challenge—it is fundamentally a governance challenge regarding authority and accountability.
One of the sectors poised for visible transformation due to AI is e-commerce.
For over two decades, e-commerce has digitized storefronts and trade activities onto the Internet. However, consumers still manually navigate the majority of the buyer journey: searching for products, reading descriptions, comparing prices, reviewing ratings, selecting shipping methods, executing payments, tracking orders, and managing returns.
An emerging paradigm referred to by the technology and payments industries is agentic commerce. In this model, AI agents act on behalf of users to search, evaluate, select, and—when authorized—complete transactions.
Mastercard defines agentic commerce as a form of online or mobile shopping where AI agents can autonomously execute select end-to-end tasks for users—from discovery and comparison to purchase—with minimal to no manual steps required at each stage.
In this context, the term A-Commerce may serve as a concise shorthand for agentic commerce. However, it should be regarded as a proposed communication shorthand rather than a formal industry replacement for “e-commerce.”
Traditional e-commerce is unlikely to disappear entirely. A more realistic trajectory is that e-commerce will evolve to incorporate an additional interaction layer specifically designed for AI agents.
In an agentic commerce environment, individuals could maintain a personal AI assistant tailored to their specific parameters:
Instead of opening multiple browser tabs and querying keywords individually, a user might simply state:
“I need a lightweight laptop for business trips, capable of running local AI models, priced under 40 million VND, delivered before Friday, with official manufacturer warranty coverage in Vietnam.”
The personal AI assistant can then:
The shopping journey shifts from:
Search → Click → Compare → Pay
to:
Express Intent → AI-assisted Goal Fulfillment
In other words, commercial activities will no longer necessarily begin with a keyword; they can originate from an intent.
On the seller’s side, enterprises can deploy merchant AI agents configured with controlled access to:
When a buyer’s agent submits an inquiry, the enterprise’s agent can automatically respond with:
In business-to-business (B2B) transactions, both agent systems can facilitate technical specification exchanges, draft quotations, compare bill-of-materials, verify budgets, prepare contracts, and initiate internal approval workflows.
However, full agent-to-agent negotiations or fully automated contract executions remain subject to legal, technical, and regulatory prerequisites. In the near term, the viable model remains one where agents prepare and recommend options, while humans or authorized representatives retain final approval authority.
Future commercial transactions may operate on a multi-tier structure:
Buyer → Personal AI Assistant → Agent Transaction Gateway → Merchant Agent → Inventory, Payment, Loyalty & Logistics Systems
Payment networks have already begun establishing the infrastructure for this scenario. Mastercard, for instance, introduced Agent Pay as a framework to facilitate payments executed by AI agents, prioritizing transaction verification, control, and security.
While fully autonomous commerce is not yet ubiquitous, these steps show that foundational infrastructure layers are actively being assembled to accommodate AI-driven transactions.
Current e-commerce websites are primarily engineered for human navigation—featuring banners, drop-down menus, search bars, product detail pages, shopping carts, and checkout interfaces.
In an agentic commerce paradigm, enterprises will require machine-readable communication interfaces, such as:
Agents operate far more efficiently when accessing structured data feeds and governed APIs rather than web-scraping unstructured visual elements from human-facing interfaces. Technical documentations on agentic commerce similarly point toward direct API interactions as a core characteristic of higher-level agentic capabilities.
Future enterprises will therefore need more than just great user interfaces for humans; they will need accurate, structured, and securely accessible data architectures for AI systems.
Today’s e-commerce predominantly starts with keywords. However, human requirements are multidimensional—encompassing context, budgets, timelines, preferences, and risk tolerance.
A user may type “AI laptop,” but their actual intent might be:
“A device light enough for business travel, capable of running local AI software, with all-day battery life, and an accessible local service center.”
AI agents must interpret the intent behind queries rather than merely matching keywords against product catalogs.
Accordingly, commercial interaction is gradually evolving:
This represents an evolving transition rather than an instant replacement. Search engines and websites will continue to coexist alongside, and be augmented by, conversational and intent-driven interaction layers.
In traditional e-commerce, businesses compete for human attention via search ads, social media campaigns, influencer partnerships, promotional banners, and remarketing loops.
When AI agents act as intermediaries in the evaluation process, marketing must convince more than just human perception. Information must pass structured validation criteria:
This dynamics pushes marketing from an attention economy toward a trust and relevance economy.
This projection should also be interpreted prudently. Emotion, branding, aesthetics, and visual experiences will always hold substantial influence over human choices. Nevertheless, accurate data and transparent terms will become far more critical as AI functions as a primary filtering layer. A strong brand in the AI era is not merely one with high mindshare; it must be a brand with verifiable data, transparent policies, consistent quality, and secure transaction infrastructure.
Agentic commerce can only scale sustainably if authority management is resolved.
Users must establish granular parameters:
These can be mapped into four operational levels:
In the initial adoption phases, the most appropriate operating principle remains:
AI recommends, human confirms.
Autonomous execution rights should expand only as systems prove reliability, auditability, and robust user safety mechanisms.
When AI can generate comprehensive answers in seconds, educational frameworks can no longer center primarily on rote memorization and information recall.
Essential competencies increasingly prioritize:
When AI can supply answers, human value resides in asking the right questions, evaluating outputs, and converting knowledge into impactful action.
While AI offers immense potential to personalize education to individual learning paces, it also carries risks of over-reliance, diminished self-guided research habits, or widening the divide between those with technology access and those lacking infrastructure.
Thus, AI literacy must extend beyond knowing how to prompt tools. Learners must understand verification protocols, system limitations, data security risks, and the accountability associated with utilizing AI-generated content.
During the expansion of the Internet, societies confronted the digital divide.
In the AI era, a new gap threatens to emerge between:
AI can democratize enterprise-grade analytical capabilities for small businesses. However, realizing that potential requires societal investment in data infrastructure, workforce reskilling, technology access, and localized AI models tailored to the Vietnamese language and cultural context. OECD statistics show that AI adoption rates vary significantly across enterprise sizes and industries, underscoring that the benefits of AI will not distribute equitably by default.
AI systems can produce fluent yet inaccurate outputs (hallucinations). Systems may be influenced by biased training sets, unverified data sources, or prompt designs intended to manipulate results.
In agentic commerce, operational risks include:
Consequently, AI deployments demand robust protective layers:
Users must always be able to verify where AI sourced its information, why a specific recommendation was made, what parameters it operated under, and who bears legal responsibility for the outcome.
Trust is not an afterthought in AI design—it is a foundational prerequisite.
As computational, analytical, and generative capabilities become commoditized, uniquely human qualities become increasingly valuable:
AI can generate a decision proposal, but it cannot bear social or moral accountability for that proposal.
Technology expands human capabilities, but the intent behind those capabilities, the boundaries enforced, and the ultimate responsibility for outcomes remain fundamentally human and organizational commitments.
Over nearly two decades, SpeedUP has actively participated in digitizing business operations through solutions across point-of-sale (POS), restaurants, retail, payments, self-service kiosks, ticketing, aviation, hospitality, and enterprise systems integration.
From this practical experience, we observe three evolutionary phases:
SpeedUP does not view AI as a standalone add-on feature bolted onto software. We view AI as a foundational capability layer connecting data, people, hardware devices, and workflows.
In commerce, this leads toward an ecosystem where every individual has an AI assistant, every enterprise operates AI agents, and increasing steps along the transaction journey are coordinated seamlessly between intelligent systems.
In enterprise management, AI helps organizations transition from historical reporting to predictive forecasting, from reactive troubleshooting to proactive detection, and from manual execution to governed human-machine collaboration.
Yet, long-term value can only be sustained on a foundation of trustworthy data, transparent governance, and ultimate human control.
The AI revolution will not be defined solely by processor speeds, parameter counts, or the sheer number of automated tasks.
What matters more is how society chooses to deploy this technology.
AI can be used primarily to monitor, replace labor, and consolidate power. Alternatively, AI can be harnessed to augment human capability, democratize knowledge, eliminate repetitive toil, and deliver more intuitive public and private services.
A truly intelligent society is not necessarily the one with the most AI models.
It is a society that knows how to deploy AI effectively, responsibly, and for human well-being.
The AI revolution should not be measured only by how much work machines can replace, but by how effectively technology can expand human capability and create meaningful value for society.
References
International Labour Organization. (2025). Generative AI and jobs: A refined global index of occupational exposure. International Labour Organization. https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure
International Monetary Fund. (2024). Gen-AI: Artificial intelligence and the future of work. International Monetary Fund.
Mastercard. (2025). Agentic commerce explainer. Mastercard.
Mastercard. (2025). Agentic commerce momentum. Mastercard.
McKinsey & Company. (2025). The State of AI. McKinsey & Company.
McKinsey & Company. (2025). The agentic commerce opportunity: How AI agents are ushering in a new era for consumers and merchants. McKinsey & Company.
Microsoft. (2025). 2025 Work Trend Index Annual Report. Microsoft.
National Institute of Standards and Technology. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). U.S. Department of Commerce.
OECD. (2019, updated 2024). OECD AI Principles.
OECD. (2026). The OECD.AI Index. OECD Publishing.
OECD. (2026). The agentic AI landscape and its conceptual foundations. OECD Artificial Intelligence Papers, No. 56. OECD Publishing.
Stanford Institute for Human-Centered Artificial Intelligence. (2025). AI Index Report 2025. Stanford University.
UNESCO. (2021). Recommendation on the Ethics of Artificial Intelligence. UNESCO.
World Economic Forum. (2025). The Future of Jobs Report 2025. World Economic Forum.
