AI and Labour: Preparing for a Future We Cannot Stop

There is a question about artificial intelligence being asked almost everywhere today: Will AI take away our jobs?
It is an understandable question, but perhaps not the best one with which to begin. Every major technological transformation has changed the nature of work. The steam engine, electricity, computers and the internet each eliminated some occupations, created others and transformed many more. AI is different in one important respect: it is entering the domain of human cognition—analysing, writing, translating, designing, diagnosing, predicting and making decisions. And it is doing so at extraordinary speed.
The anxiety is reflected in a recent paper, The AI Layoff Trap, by researchers associated with Boston University and the Wharton School. Using a mathematical model, they describe a possible dystopian outcome in which competition drives firms towards excessive automation: each gains by replacing workers, but widespread layoffs eventually weaken the purchasing power on which the economy itself depends. It is a theoretical model, not a prediction, but it raises a disturbing question. At the other extreme, Elon Musk has envisaged a future in which AI and robots could make employment largely optional, creating a world of abundance.
Between these extremes lies the future governments must actually confront: a transition in which some jobs disappear, many are transformed and new ones emerge. The question is not whether this transformation can be stopped. It is how we prepare for it, protect people from its undesirable consequences and use its possibilities for their benefit.
The Job May Survive—But Change
The most visible concern about AI is job loss. Some jobs will undoubtedly disappear. But an equally important possibility is that many jobs will survive while the work within them changes dramatically.
A lawyer may spend less time researching; a doctor may work alongside intelligent diagnostic systems; a factory worker may supervise increasingly autonomous machinery. The International Labour Organization has similarly suggested that, in many occupations, AI is more likely to transform or augment work than eliminate entire occupations.
The worker’s greatest vulnerability, therefore, may not always be unemployment but the obsolescence of skills.
Skill development must consequently become a continuing process. The idea that education is acquired in youth and then used throughout a career is becoming outdated. Workers may have to learn, unlearn and relearn several times during their working lives.
The State has to prepare for that reality.
When the Boss Becomes an Algorithm
There is another, less visible change. Algorithms can increasingly participate in recruitment, allocation of work, performance assessment and scheduling. A worker may receive an assignment from an app, have his performance continuously assessed by software or find that an algorithm has influenced whether he is retained.
This raises a simple but uncomfortable question: if a machine makes a decision about me, who do I appeal to?
Workers may need new rights in the age of AI—the right to know when an important decision has been influenced by an algorithm, the right to an explanation and human review, and protection against discriminatory or intrusive systems.
Technology can assist management. It should not make management unaccountable.
Social Security Must Follow the Worker
AI will also expose weaknesses in traditional social-security arrangements. The worker of tomorrow may move between regular employment, gig work, self-employment, unemployment and retraining. If benefits remain tied to a particular employer, workers may repeatedly fall through the cracks.
Social security must therefore increasingly follow the worker, not the job.
India has begun laying foundations through e-Shram and the recognition of gig and platform workers in the social-security framework. The next step should be more portable, life-cycle protection covering pension, provident fund, health, disability, maternity, accident protection and assistance during periods of transition.
AI could make such a system more responsive. Instead of expecting a worker to know which benefits he is entitled to and how to claim them, technology could help identify his eligibility and guide him through the process.
The principle is simple:
The worker should not have to discover the welfare system; the welfare system should discover the worker.
Prepare. Protect. Leverage.
The State’s response to AI and labour can, I believe, be reduced to three words.
Prepare. Anticipate which occupations and skills are likely to change and provide opportunities for continuous learning and retraining. A worker whose occupation is disrupted by technology should not simply be told to “adapt” and left to fend for himself.
Protect. AI-era labour rights will need to address algorithmic decision-making, workplace surveillance, discrimination and the use of personal data. We should also consider an AI Labour Impact Assessment when major technological changes are likely to affect large numbers of workers. It could examine employment effects, skill requirements, retraining needs, occupational safety and social-security implications. Its purpose would not be to stop automation, but to make technological change socially responsible.
Leverage. Perhaps the most overlooked opportunity is to use AI itself for the benefit of labour. It could help labour inspectors identify establishments where violations are most likely, detect patterns preceding workplace accidents, identify workers entitled to social-security benefits and make grievance redressal easier.
Imagine a worker being able to ask, in his own language: What are my rights? What benefits can I claim? Where do I complain?
That is not AI replacing labour. It is AI working for labour.
A Future We Can Shape
We cannot—and should not—stop technological progress. But neither should workers be left alone to absorb its costs. The gains from higher productivity must not accrue only to those who own the technology while the risks are borne by those who depend upon their labour.
The real challenge is not to decide whether we are for or against AI. It is to ensure that, as machines acquire greater capacity to make decisions, human beings do not surrender their capacity to make the important decisions about machines.
We may not be able to determine how intelligent machines will become. We can still determine what role we allow them to play in the lives of working people.

(Uday Kumar Varma is an IAS officer. Retired as Secretary, Ministry of Information & Broadcasting)