AI in HR and the Law: Build the Guardrails Before Regulation Forces You To

When organisations begin exploring Artificial Intelligence in Human Resources, one question often comes up: why do we need to create so many systems, controls and guardrails? Why can an organisation not simply purchase an AI product and start using it?

The answer lies in the nature of AI itself.

Artificial Intelligence is still an evolving domain. There have been different levels of development and progress, but AI continues to evolve. More importantly, the laws and guardrails governing the use of AI are also evolving.

An organisation may feel that it can purchase an AI product today, use it within the Human Resources function, and deal with regulation whenever it arrives. But that approach carries a serious risk. Unlike a conventional product that can simply be purchased, used and replaced, an AI system is trained over time.

The organisation purchases the AI and then continues to train it. It adds information, develops use cases and learns how to use the system more effectively. The more the organisation uses the AI, the more useful its outputs may become. Over time, the AI may begin to add considerable value by reducing work, making processes faster or supporting the organisation in managing the Human Resources function more effectively.

Now imagine that the organisation invests significant time, resources and effort in training the AI. The system becomes deeply connected to HR processes and starts delivering meaningful value. Then a law or regulatory requirement arrives and the organisation discovers that the way it has trained or used the AI is not acceptable.

The organisation may then be required to discontinue or shut down the system after investing heavily in it.

That is precisely why an organisation should not wait for every AI law to arrive before introducing responsible AI governance. The better approach is to understand the guidelines and laws that already exist globally and ensure that the organisation’s AI in HR systems are not operating in contradiction to them.

Why Global AI Guidelines Matter to Human Resources

Since AI-related laws and guardrails are still developing, organisations need to take guidance from the frameworks that are already available globally.

One important reference is the guidance relating to AI issued by the International Labour Organization, or ILO. An HR practitioner may ask why an organisation should be concerned about ILO guidelines if those guidelines are not the organisation’s local employment law.

The reason is that labour laws in many countries have been influenced by the principles and standards associated with the ILO. Therefore, in my view, it is reasonable to believe that AI guidelines issued by the ILO may eventually influence employment-related AI laws and regulations that countries introduce.

Organisations should therefore be clear about one point: employment-related AI regulation will eventually become a reality. AI is increasingly being used in areas connected with employment, employees and careers. It is reasonable to expect that laws will emerge to govern how AI is used in these areas.

The second important reference is the European Union Artificial Intelligence Act. The European Union is generally seen as progressive in its approach to legislation, and it already has an AI law explaining how and where AI may be used, including in employment-related situations.

When new employment-related AI regulations are framed elsewhere, those responsible for drafting such laws are likely to examine existing references. Since the EU AI Act is one of the early laws in this field and is already in place in the European Union, it can become an important reference point for future lawmakers.

The third important reference is the AI guidance associated with UNESCO.

An organisation may again ask why UNESCO guidance should matter to an HR function. The reason is that UNESCO approaches these issues at a much broader level. It considers the implications from the perspective of human beings and civilisation.

Guidance framed from such a broad human perspective can become relevant when countries develop laws that affect employment and the use of AI in relation to people. These ideas may eventually find their way into the laws and regulatory expectations that organisations in India will need to follow.

My recommendation is not that every organisation must immediately treat every international guideline as a local legal obligation. Whether an organisation chooses to follow each guideline directly is a decision it can make.

However, an organisation should at least ensure that its AI systems and AI in HR practices are not operating in contradiction to the principles reflected in these significant global references.

The practical reason is straightforward. If the organisation designs its AI use in a manner that is fundamentally opposed to the direction in which global regulation is moving, it may eventually have to undo years of training, investment and implementation.

AI Is Not a Product You Simply Replace

A traditional product can often be purchased, used and then changed when a better product becomes available. Artificial Intelligence works differently because the organisation does not merely purchase it. The organisation also trains it.

The value obtained from an AI system can keep increasing as the organisation continues to use it, train it and develop its application. The AI may gradually become more useful because it begins to support more work, respond to more requirements and produce outputs that save time or reduce effort.

This creates both value and dependency.

The more resources the organisation invests in the system, the more damaging it can be if the system later has to be abandoned. That is why AI governance cannot be treated as something to be considered only after the AI has been deployed.

Guardrails need to be understood before the organisation becomes heavily dependent on the system.

An organisation should consider the direction of ILO AI guidance, the EU AI Act and UNESCO AI guidance while developing its own AI in HR systems. The objective is to reduce the possibility that the organisation builds an AI capability that later becomes difficult or impossible to continue because of emerging legislation.

Responsible AI implementation is therefore not only about avoiding a present problem. It is also about protecting the organisation’s future investment.

If the organisation knows that regulation is evolving, it should design the system with sufficient foresight. It should avoid those practices that appear likely to conflict with the principles already being discussed internationally.

The organisation should not wait until a law forces it to change. It should put sensible governance systems in place while its AI capability is still being developed.

Three Critical Guardrails for Implementing AI in HR

If I were advising an organisation on the most critical steps it should take while implementing AI in Human Resources, I would begin with three areas: a strong steering committee, continuous testing by an HR Tech Expert, and transparent communication with employees.

The first requirement is a strong AI steering committee.

Every organisation implementing AI should create a steering committee that guides the implementation. This requirement is relevant even for a small organisation.

AI can provide an organisation with a valuable way of running parts of the Human Resources function. But this makes the quality of the steering committee even more important. If the organisation gets the committee right, it improves the possibility that the appropriate systems, controls and decisions will be put in place from the beginning.

The steering committee provides direction. It helps the organisation look beyond the immediate attractiveness of the AI product and consider how AI is being implemented, how it is being used and whether the right governance structure is supporting it.

The second requirement is a strong and continuous system of AI testing led by an HR Tech Expert.

Testing should happen when the organisation is considering the purchase of an AI product. However, testing should not stop after the product has been purchased and implemented.

The AI product needs to be tested continuously. The organisation should review it periodically, including every few months, to understand how it is performing and whether its behaviour continues to meet the organisation’s requirements.

This is a necessity, not an optional activity.

If an organisation implements AI in HR without continuous testing, it is likely to face problems. The AI is being trained, its use is expanding and its outputs are affecting HR activities. The organisation therefore needs an HR Tech Expert who can test the product and support an ongoing assessment of its performance.

The organisation should not assume that a successful purchase decision proves that the system will remain suitable indefinitely. Continuous use requires continuous testing.

The third and perhaps most important requirement is continuous and transparent communication with employees.

Employees should know where AI is being used. They should understand how AI is affecting their careers. They should know whether AI is enabling or not enabling career-related outcomes inside the organisation.

This communication should not happen once. It should be continuous. It should also be transparent.

Employees need confidence that decisions relating to their careers are ultimately being taken by human beings, even when those decisions are supported by AI.

This distinction is essential.

AI may provide information, analysis or support. But eventually, a human being should be able to answer the employee and explain why a particular career-related decision was taken.

The answer cannot be, “This decision was taken because the AI said so.”

If a human decision-maker cannot provide a reason and can only state that the decision was based on what the AI produced, then the AI system and the surrounding governance have failed.

An appropriate AI in HR system must preserve human accountability. Even where AI supports a decision, a human being must remain capable of explaining the decision and taking responsibility for it.

Human Accountability Must Remain at the Centre of AI in HR

AI in HR will increasingly affect employment-related processes and employee careers. This makes governance, testing and communication essential.

Organisations should begin by recognising that AI legislation is still developing. They should study the direction reflected in ILO AI guidance, the EU AI Act and UNESCO AI guidance, and ensure that their systems are not designed in contradiction to these emerging principles.

This is important because an AI system grows in value as the organisation uses and trains it. The organisation may invest considerable resources before the system starts producing substantial benefits. It would be unfortunate to reach that stage and then discover that the AI has to be discontinued because the implementation is inconsistent with the law that eventually develops.

Organisations should therefore build responsible AI in HR systems from the beginning.

A strong steering committee should guide implementation. An HR Tech Expert should continuously test the AI product, both before purchase and after implementation. Employees should receive regular and transparent communication about where AI is being used and how it affects their careers.

Most importantly, employees should always know that a human being remains responsible for career-related decisions.

AI can support the decision. AI can assist the organisation. AI can help the Human Resources function become faster and more effective. But AI cannot become the reason that nobody is willing or able to explain a decision.

If the only explanation available is that “the AI said so”, the organisation does not have a properly governed AI in HR system.

The objective is not to avoid AI. The objective is to implement AI in a manner that the organisation can continue using, developing and trusting as legal expectations evolve.

That requires foresight.

It requires strong governance.

It requires continuous AI testing by an HR Tech Expert.

It requires transparent communication with employees.

And it requires human accountability for every decision that affects an employee’s career.

Build these guardrails before regulation forces you to rebuild the entire system.