Artificial intelligence is no longer a distant future for the workplace. It is already entering recruitment, hiring, performance management, learning and development, redundancy planning, safety management and daily business operations.
That is why the conversation must move beyond the headlines.
At the UN Responsible Business and Human Rights Forum in Bangkok, ICARUS AI joined Ius Laboris for the session “Beyond the headlines: What AI really means for the world of work.” The session brought together workplace law, employer practice, regional risk, education, workforce readiness and human capability perspectives.
The central question was not only whether organizations are ready to adopt AI.
The deeper question was whether people are ready to govern it.
Why Responsible AI at Work Matters
AI is already changing the employment lifecycle. It is being used to support job description creation, candidate scoring, resume parsing, AI interviews, performance management, learning and development, redundancy management and safety management. It is also being used externally in business operations, including research, drafting, document review, due diligence and financial modelling.
These changes create real opportunities. AI can expand access to knowledge, accelerate work, improve productivity and help organizations redesign how people learn and contribute.
However, they also create risks.
AI can affect fairness, privacy, discrimination, due process, transparency, worker dignity, accountability and the right to challenge decisions that affect livelihoods. The OECD notes that trustworthy AI in the workplace is difficult because AI can create risks for human rights, privacy, discrimination, labour rights, job quality, transparency, explainability and accountability.
This is why responsible AI at work cannot be treated as a software issue alone.
It must be treated as a human rights and workforce capability issue.
Moderator & Panel
- Moderator – Ken Chamuva Shawa, Senior Economist and Head of the Economic and Regional Analysis Unit, ILO
- Speaker – Desmond Wee, Partner, Ius Laboris Singapore
- Speaker – Catherine Leung, Partner, Head of Employment, Ius Laboris Hong Kong
- Speaker – Dr. hc. Nektar Baziotis, President – CEO, ICARUS AI Inc.
- Speaker – Dr. Christiane Haberl – CLO, ICARUS AI Inc.
The Moderator’s Opening Question: What Is Happening on the Ground?
The first round of the session asked a practical question: what is really happening on the ground? The moderator invited the panel to move away from abstract AI debate and focus on what employers, workers and institutions are already experiencing across the Asia-Pacific region.
This framing mattered because AI at work is no longer theoretical. It is already present in management decisions, workplace processes and the relationship between organizations and their people.
Desmond Wee: AI Is Changing Both Employment Management and Business Operations
The moderator asked Desmond Wee, Partner at Ius Laboris Singapore, what issues he is seeing in South-East Asia in relation to the use of AI in the workplace.
Desmond framed workplace AI through two lenses.
First, AI is being used internally for employment-management functions: recruitment, hiring, performance management, learning and development, redundancy management and safety management. Second, AI is being used externally to support business operations, including legal research, drafting and due diligence.
His first observation was that, at least in the current phase, AI is not always the primary cause of workforce restructuring. In his experience, many restructurings in Asia have been driven by wider macroeconomic pressures, the aftermath of Covid-19 and the technology downturn. However, he also warned that this may change as organizations increasingly rationalize the business use case for AI.
His second observation focused on regulation. Because AI adoption across ASEAN countries is uneven, the regulatory landscape is also uneven. In many jurisdictions, the governance of workplace AI still relies heavily on existing employment, privacy, discrimination and dismissal laws rather than dedicated AI legislation.
His third observation was that organizations are placing more emphasis on governance, ethics and responsible AI deployment. He highlighted transparency toward employees, intellectual property, confidentiality, data privacy, AI policies and internal governance structures as urgent priorities.
Desmond’s contribution made one point clear: responsible AI at work requires policies that are not decorative. They must be understood, used and lived inside the organization.
Nektar Baziotis: The Real Risk Is the Gap Between AI Adoption and Human Readiness
The moderator then asked Dr. h.c. Nektar Baziotis, President–CEO of ICARUS AI Inc., whether organizations are preparing the people responsible for using, questioning and overseeing AI systems quickly enough.
Nektar argued that many organizations still look at AI through a narrow lens. They see a tool to cut costs or trim headcount, while missing its wider potential to unlock human capability, expand access to knowledge and identify talent in places that were previously ignored.
He then challenged one of the most common phrases in AI governance: human in the loop.
His question was direct: which human?
If a manager cannot interrogate the algorithm, if a worker cannot challenge a flawed output, and if neither has the authority to say “stop,” then they are not truly in the loop. They are standing near the loop, carrying responsibility for a machine they do not control.
This is one of the central ideas for ICARUS AI.
The greatest risk is not only the technology itself. It is the growing gap between AI adoption and human readiness. Organizations are investing heavily in software, but not always in the people expected to govern that software responsibly.
Catherine Leung: You Cannot Blame the Algorithm
The moderator asked Catherine Leung, Partner and Head of Employment at Ius Laboris Hong Kong, how AI is already changing the way organizations make decisions about people, from hiring and performance management to workforce monitoring, and where the biggest legal and governance risks are emerging.
Catherine opened with a striking workplace reality from Hong Kong: a high proportion of organizations are already using AI in day-to-day operations, with many using multiple AI systems across their business. Her point was that the question is no longer whether AI is here, but whether it is being governed properly.
She then shared a practical example involving an AI recruitment tool that appeared to disadvantage older candidates and women for a senior technical role. Her conclusion was clear: employers cannot simply say they did not program the system to discriminate. Responsibility still sits with the organization.
Catherine identified four major risk areas.
The first was bias and discrimination. The second was employee monitoring and algorithmic management, especially when workers are evaluated through scores they do not understand or cannot challenge. The third was confidentiality and data security, as employees may use generative AI with sensitive information without understanding where that data goes. The fourth was AI-assisted decision-making in discipline, dismissal and redundancy, where context and human nuance remain essential.
Her conclusion aligned closely with the ICARUS position: human judgment is not optional. It is the safeguard. Meaningful oversight requires knowledge, authority, time to review evidence and the confidence to reject an AI recommendation when needed.
Chris Haberl: Workers Must Be Included Before the Decision Is Already Made
The moderator asked Dr. Chris Haberl, Chief Learning Officer at ICARUS AI Inc., whether workers are being meaningfully included in how AI tools are selected, introduced and evaluated, or whether they are only informed after decisions have already been made.
Chris argued that, in too many organizations, team members are handed a fait accompli. Management selects the software, changes the workflow and sends an announcement after the real decision has already been taken.
This is a serious governance weakness.
The people doing the work every day understand the operational friction, the human context and the precise places where an algorithm may misread reality. If those people are not included early, organizations lose essential knowledge.
Chris introduced the ICARUS model of three rotating roles: Creator, Checker and Approver. In this model, team members rotate through responsibility. One person may generate concepts, another may audit for bias and hallucinations, and another may take final responsibility before release.
The point is simple but powerful.
AI governance cannot be a top-down email. It has to become daily practice.
Round Two: From Policy to Responsible Use of AI
The second round shifted from the present state of workplace AI to the question of responsible implementation. The moderator framed the discussion around moving from policy to ethical or responsible use of AI.
This is where the session moved from identifying risk to asking what organizations should actually do.
Catherine Leung: AI Could Weaken the Entry-Level Talent Pipeline
The moderator asked Catherine how employers should rethink workforce strategies as AI squeezes entry-level roles and changes the skills organizations need.
Catherine used the image of a diamond-shaped organization: fewer entry-level roles at the base, a bulge of mid-level employees and a shrinking pool of future leaders because the traditional learning ground for early-career talent is being eroded.
She warned that employers who cut early-career hiring for short-term savings may later face expensive leadership gaps. If entry-level roles disappear too quickly, organizations may lose the very pipeline that develops future managers and leaders.
Her answer also emphasized that the skills landscape is changing. Critical thinking, judgment, communication, adaptability and learning agility are becoming more important as AI reshapes work.
This is a central human rights and workforce issue. AI should not create opportunity only for those already positioned to benefit. Responsible employers must ask how AI affects the next generation of workers, career-changers and those entering the labour market.
Chris Haberl: One-Off AI Training Does Not Build Capability
The moderator asked Chris how organizations can move from one-off AI training to continuous capability-building that helps people recognize risks, exercise judgment and protect rights in daily work.
Chris’ answer was direct: a one-hour video course does not build capability. It checks a corporate box.
Real capability comes from continuous, hands-on practice within structured roles. The ICARUS model uses the Creator, Checker and Approver structure to make responsible AI practical. The Creator uses AI while filtering restricted data. The Checker verifies facts, sources, hallucinations and bias. The Approver evaluates legal exposure, human rights impact and ultimate accountability before final release.
This model matters because responsible AI cannot depend on intention alone. It depends on what actions create in practice.
Chris also connected capability-building to a broader ethic of respect, responsibility and shared governance. Continuous capability means creating systems where people actively safeguard one another every day.
For ICARUS AI, this is the difference between awareness and capability.
Awareness means someone has heard the policy.
Capability means they know what to do when AI creates risk.
Nektar Baziotis: Education Is the Foundation of Human Rights Capability
The moderator asked Nektar what real human rights capability looks like inside an organization, beyond awareness training or compliance policies.
Nektar’s answer began with a fundamental principle: education is the foundation of every human right. Without education, people cannot defend their dignity, privacy or livelihood. Without education, rights risk becoming words without power.
He then defined human rights capability through three pillars.
The first is the sight to spot risk. People must be educated to see where an algorithm may compromise fairness, dignity or truth.
The second is the voice to challenge. When an AI system affects a person’s life or career, that person must be able to ask why the decision was made, what data was used and by whose authority.
The third is the power to act. A worker or manager who identifies a flaw must have the authority and governance pathway to pause the system, escalate the issue or reject the output.
This is the ICARUS contribution in its clearest form.
Responsible AI at work is not achieved when a policy is written.
It is achieved when people are capable of governing technology in daily practice.
Desmond Wee: Policies Must Be Understood, Not Just Written
The moderator then asked Desmond whether he agreed with Nektar’s point that organizations need a true understanding of how AI can and should be deployed, beyond having policies, trainings and processes in place.
Desmond agreed and strengthened the point from both a legal and organizational perspective.
He emphasized that comprehensive workplace AI policies are important because they create clarity of purpose, define the parameters of use and establish due process and governance. However, he also stressed that policies must be understood by the people responsible for using them. They must remain live documents that are relevant to users and business realities.
He shared a practical example of an organization with an impressive AI policy that restricted use to approved tools, while at the same time encouraging employees to use other non-approved AI tools outside work devices. The inconsistency between policy and practice exposed the company to risk.
This example captures the central challenge.
AI governance fails when the written policy and the lived practice point in different directions.
The Human in the Loop Must Have Authority
Across the panel, one theme returned again and again: the human in the loop must be real.
The OECD notes that AI in the workplace raises challenges around transparency, explainability and accountability, especially when AI affects employment-related decisions. The European Union’s AI Act also treats several employment-related AI systems as high-risk, including systems used for recruitment, selection, promotion, termination, task allocation, monitoring and performance evaluation.
These frameworks confirm what the panel discussed in practical terms.
When AI affects work, livelihood, dignity and opportunity, oversight cannot be symbolic.
A person must be able to understand the system, question the output, review the evidence, escalate concerns and stop a harmful decision.
Otherwise, the organization has not created governance.
It has created automated risk with a human signature at the end.
What Responsible AI at Work Requires
Responsible AI at work requires more than a technical rollout.
It requires clear policies, meaningful worker participation, human rights due diligence, practical training, continuous capability-building and governance structures that people can actually use.
The ILO describes two broad workplace uses of AI: AI that automates tasks workers perform, and AI-based analytics or algorithms that automate managerial functions, often referred to as algorithmic management. This distinction matters because AI does not only change what work is done. It can also change how people are managed, monitored and evaluated.
That is why organizations should begin with practical questions:
- Which AI systems are being used?
- Which decisions do they support?
- Who is affected?
- Who understands the system?
- Who checks the output?
- Who has authority to challenge it?
- What remedy exists if harm occurs?
These questions move responsible AI from theory into practice.
The Next Workplace Divide
The next workplace divide will not only be digital.
It will be a capability divide.
Some organizations will use AI to strengthen workers, improve decisions and build more resilient workplaces. Others may use AI mainly to accelerate processes, reduce costs and transfer risk onto people who are not prepared to govern it.
The difference will not be the tool alone.
It will be the capability around the tool.
This is why ICARUS AI argues that digital training is not enough. Organizations need learning systems that build judgment, participation and decision rights. They need workers and managers who can recognize risks, ask better questions and act when human rights are affected.
Closing Reflections: One Sentence to Remember
The moderator closed by asking each panelist to share one sentence they would like participants to remember from the session.
Desmond’s closing message was that organizations must first determine AI’s fundamental value proposition and then chart a clear course toward that goal.
Nektar’s closing message was that democratizing knowledge gives every human being the voice and authority to shape their own destiny.
Catherine’s closing message was that AI is not conscious, and that companies will succeed by investing in people while keeping human judgment and human wellbeing at the centre of decision-making.
Chris’ closing message was that AI should elevate people through shared governance and equal accountability, creating organizations worthy of the future.
Beyond the Headlines
AI will continue to transform the world of work.
The question is whether organizations will treat that transformation as a race for efficiency or as an opportunity to build more responsible, inclusive and resilient workplaces.
The Bangkok session brought together two essential perspectives.
Ius Laboris brought the workplace law, employer practice and regional employment-risk lens.
ICARUS AI brought the human capability, education and workforce-readiness lens.
Together, the message was clear.
Responsible AI at work cannot be built by policies alone. It requires capable people, clear rights, meaningful oversight, worker participation and the courage to place human dignity at the centre of workplace transformation.
The future of work will not be shaped by AI alone.
It will be shaped by the people prepared to govern it.
Sources
- UN Responsible Business and Human Rights Forum — https://www.rbhrforum.com/
- UN Guiding Principles on Business and Human Rights — https://digitallibrary.un.org/record/720245
- OECD Employment Outlook 2023: Ensuring Trustworthy Artificial Intelligence in the Workplace — https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en/full-report/ensuring-trustworthy-artificial-intelligence-in-the-workplace-countries-policy-action_c01b9e49.html
- OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market — https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en.html
- ILO: Artificial Intelligence and the World of Work — https://www.ilo.org/topics-and-sectors/artificial-intelligence
- ILO: Algorithmic Management in the Workplace — https://www.ilo.org/algorithmic-management-workplace
- ILO: The Algorithmic Management of Work and Its Implications in Different Contexts — https://www.ilo.org/publications/algorithmic-management-work-and-its-implications-different-contexts
- ILO: Artificial Intelligence Adoption and Its Impact on Jobs — https://www.ilo.org/publications/artificial-intelligence-adoption-and-its-impact-jobs
- OECD Employment Outlook 2023: Social Dialogue and Collective Bargaining in the Age of Artificial Intelligence — https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en/full-report/social-dialogue-and-collective-bargaining-in-the-age-of-artificial-intelligence_5828691a.html
- EU AI Act: Annex III — High-Risk AI Systems — https://ai-act-service-desk.ec.europa.eu/en/ai-act/annex-3
- EU AI Act: Recital 57 — Employment and Workers’ Rights — https://ai-act-service-desk.ec.europa.eu/en/ai-act/recital-57