Smarter machines, weaker humans? AI and the future of the human worker

There is a peculiar irony in asking a machine to help us write, only to discover that it has helped us sound like everybody else. The sentences arrive polished. The paragraphs behave themselves. Nothing is obviously wrong, yet something is missing: the stubborn grain of a particular mind. The writer has acquired fluency and misplaced a voice.

Behind this irritation lies a larger question confronting the human worker. As artificial intelligence becomes more capable, will its users become more capable too? Or will the machine grow in competence while human beings lose the habits of thought that once justified the job?

In October 2025, Deloitte agreed to partially refund Australia’s government over a report on welfare administration costing A$440,000 that contained nonexistent academic references and a fabricated judicial quotation. University of Sydney researcher Christopher Rudge helped expose the errors. Deloitte subsequently acknowledged in correspondence released under freedom of information that generative AI had produced inaccurate citations which its review had failed to catch. The department maintained the report’s recommendations. But the episode exposed a failure of professional verification. A prestigious name on the cover could not make invented evidence real.

For Nigerian graduates navigating an unforgiving labour market, learning to use AI may improve their prospects. But acquiring an instrument differs from becoming dependent on answers one lacks the competence to question. The first can open a career. The second can hollow it out from within.

In an earlier column, “Who will teach AI about Africa?”, I argued that Africans must help shape the knowledge through which these systems interpret the world. The question now comes closer to home. When AI enters the workplace, who remains the author of the work?

In my ongoing research, I explore this question through the Hollywood film How to Train Your Dragon. Needless to say that its setting is distant from Nigeria, but its central encounter offers a useful metaphor.

Hiccup, a young Viking expected to kill dragons, instead learns to collaborate with one. He repairs its damaged tail with a prosthetic fin and builds a rig that allows them to fly together. The dragon supplies power; Hiccup supplies the missing control. Neither can fly alone.

AI, of course, is no creature awaiting friendship. The useful parallel lies in observation, experimentation and deliberate design. Hiccup must acquire skills and construct the means of collaboration. He cannot simply mount the dragon and announce a transformation.

Neither can a university buy subscriptions and declare its academics augmented.

Much of the anxiety about AI concerns replacement. That fear deserves serious attention. Yet a worker can retain a job and still lose something essential. The title remains, the salary arrives, but the discretion, craftsmanship and opportunity to exercise judgement gradually disappear. Employment survives while professional capability withers.

The alternative begins with how the tool is used. A generic request tends to invite a generic response. A professional can instead bring disciplinary knowledge, examples, contextual constraints and exacting standards into the interaction, then test and revise what comes back. “Training” here need not mean rebuilding the underlying model. It can mean carefully attuning its use to a particular practice.

Nigeria already offers a promising example. In June and July 2024, public secondary school students in Edo State participated in a six-week programme combining AI tutoring with teacher guidance. A World Bank evaluation reported improved English, digital and AI skills compared with a control group. Teachers introduced curriculum topics, guided interactions and helped students identify machine errors. The researchers cautioned that the results reflected the whole programme, not the chatbot alone.

That qualification is important. The teacher remained an educator, shaping how a new capability served learning. The machine’s usefulness depended on the quality of that relationship.

But personalisation alone is not enough. A system that imitates a professional’s voice while flattering every assumption is a sophisticated echo. Genuine augmentation should expose weaknesses, suggest alternatives and invite examination. The professional must be willing to challenge the answer and to reconsider a position when the evidence warrants it.

That is what I mean by agentic authorship: retaining a consequential hand in why work is undertaken, how it is performed and what counts as a satisfactory outcome. Authorship has never required doing everything unaided. Assistance need not diminish authorship; surrendering judgement does.

For young professionals, however, there is an additional danger. The tasks dismissed as routine are often where expertise begins. Drafting, calculating, checking and revising teach the novice to recognise patterns and detect mistakes. A junior accountant who delegates every calculation without learning its logic may produce impressive reports while acquiring little capacity to judge them. The ladder to expertise cannot be removed merely because the roof is now easier to reach.

Nigerian institutions must therefore think beyond courses in prompting. People need to use AI while developing independent competence. Sometimes that requires attempting a task before consulting the machine; sometimes, defending an answer or demonstrating how a conclusion was reached. The intellectual exercise through which capability grows must survive.

Individual effort, though, cannot settle the question. A worker may know that an automated recommendation is wrong and still lack permission to challenge it. “The system says so” can become the technological successor to “instructions from above.” Familiar hierarchies acquire a digital disguise, and obedience acquires the appearance of objectivity.

Consider a bank officer required to communicate a decision generated by a system she cannot inspect or override. If the decision causes harm, who answers for it? Responsibility should follow authority. The worker must not become the machine’s alibi.

There is another promise that deserves examination: AI will save time, freeing people for more meaningful work. Perhaps. But saving time is a technical achievement; deciding who benefits is an organisational choice. If a lecturer completes marking more quickly, the university may support deeper feedback, research or student contact. It may simply allocate more scripts.

An hour saved does not automatically become an hour returned. Without deliberate choices, the promised liberation becomes a faster treadmill.

ALSO READ: VIDEO: I respect Peter Obi, but he makes bogus claims – Shettima

For a Nigerian worker struggling to pay rent, talk of meaningful work may sound indulgent. Yet professional dignity includes the opportunity to develop expertise, the authority to exercise it and protection against carrying blame for decisions controlled elsewhere. These conditions also affect the quality of service received by customers, patients and citizens.

Employers should therefore ask more than how much faster AI makes a process. Who sets its purpose? Can the worker reject an output? Is human expertise being developed or quietly depleted? Where does the saved time go? These questions belong beside productivity figures whenever a system is introduced.

The choice is not between technological enthusiasm and a retreat into yesterday. Nigerian teachers, researchers, entrepreneurs and public servants have much to gain from tools that extend scarce expertise. But that gain must be built through practice, learning and institutions willing to give people a genuine voice.

In the film, the dragon expands Hiccup’s horizon without relieving him of the need to understand the flight.

That is the ambition worth pursuing. Let the machine carry more of the burden. Let the human being acquire more of the skill, judgement and freedom to decide where the work should lead. Smarter machines should help enlarge the people who use them. Otherwise, the future of work may arrive to find the worker present, productive and progressively absent from the work.