What the global data actually says, and why Nigeria's story looks different from the headlines

In March 2026, MTN Nigeria quietly replaced more than 200 workers who had been handling SIM registration with thirteen AI agents built on computer vision. Around the same time, Kuda Technologies cut nineteen of its forty marketing staff (CTR0 Tech, "Nigeria's AI Boom Is Real. The Ownership Is Not."). Neither event made the kind of global headlines that accompany a Silicon Valley layoff announcement, but together they represent something worth paying close attention to: the automation story usually told about wealthy, desk-job economies is now visibly playing out inside Nigerian companies too, just not always in the roles or the proportions most people assume.

This is the genuinely confusing part of any conversation about AI and jobs right now. Global research and local Nigerian data tell two related but importantly different stories, and conflating them leads to bad individual decisions, a software developer in Lagos worrying about the wrong thing, or a trader at a market stall assuming a risk that barely applies to them at all. Understanding both pictures, the global pattern and the specific Nigerian variation on it, is the only way to actually assess where you personally stand.

The Global Pattern: What Keeps Showing Up Across Every Major Study

Despite differing in methodology, scope, and headline numbers, the major studies on AI exposure (Microsoft Research's analysis of 200,000 real Copilot conversations across 785 occupations, the Anthropic Economic Index, a joint GovAI and Brookings Institution study covering over 350 occupations, and IMF cross-country labour exposure research) converge on a remarkably consistent underlying pattern.

Most exposed: routine information work. Clerical and administrative roles consistently rank as the single most exposed category across nearly every study. The GovAI and Brookings research found these roles combine the highest AI exposure with the lowest estimated ability for affected workers to transition into other roles, a combination the researchers flagged as a genuine inequality risk, especially given that 86 percent of workers in the most exposed clerical roles in their dataset were women (Advisory Board). Microsoft's Copilot-usage study found a similar clustering at the top: computer, mathematical, and administrative occupational groups dominate the most-exposed list, since generative AI tools are disproportionately being used, in actual practice, for exactly the writing, analysis, and information-retrieval tasks that sit at the core of these jobs (apidots).

Least exposed: physical presence, hands-on care, and licensed judgment. The opposite end of nearly every ranking is occupied by the same recurring categories: construction trades, installation and repair technicians, agriculture workers, personal care and healthcare roles, and jobs requiring physical dexterity in unpredictable environments. Anthropic's own research places construction workers, installation and repair technicians, agricultural workers, transportation workers, and personal care workers among the least exposed categories overall (Forbes, citing Anthropic data), and a September 2026 Hamilton Group analysis found several healthcare and emergency roles, including EMTs and paramedics, with automation risk scores close to zero (American Bazaar Online).

The pattern inside healthcare is more nuanced than "safe" or "not safe." This is worth naming specifically, because healthcare often gets lumped together as uniformly AI-resistant, and it isn't. Roles requiring hands-on physical care or complex clinical judgement, surgical technologists, registered nurses, physicians, home health aides, remain genuinely low-exposure. But healthcare administrative roles sitting closer to clerical work, medical secretaries, pharmacy technicians, medical and health services managers, show meaningfully higher exposure than their clinical counterparts, because the underlying task, not the industry label, is what actually determines exposure.

Software development sits in a genuinely contested middle, not a safe zone. Several rankings place software engineers and programmers in a medium-risk band rather than either extreme, reflecting a real, ongoing debate: AI tools are demonstrably reshaping how code gets written, but the judgment, system design, and debugging work surrounding that code has proven harder to fully automate than early predictions suggested.

What's Actually Driving the Pattern

The mechanism underneath all of this is consistent, even when the specific numbers differ between studies: current-generation AI is strongest at tasks that are routine, codifiable, and primarily digital, the kind of work that can be fully described in text and completed on a screen. It remains comparatively weak at tasks requiring physical manipulation of the real world, situational judgment in unpredictable environments, or the kind of interpersonal trust that a patient, a client, or a grieving family member extends specifically to another human being. Jobs built almost entirely around the first category face real, measurable exposure. Jobs built around the second remain comparatively insulated, not because AI can't touch them at all, but because the core function resists full automation with current technology.

There's also a documented generational and demographic dimension worth taking seriously. Research following the rise of generative AI since late 2022 has found early-career workers, particularly those aged 22 to 25 in the most AI-exposed occupations, experiencing a measurable relative decline in employment even after controlling for broader firm-level shocks, while overall employment in less exposed categories has continued growing (Wikipedia, summarising the underlying labour research). This suggests the disruption isn't abstract or future-tense. It's already visibly reshaping who gets hired into entry-level knowledge work roles right now.

The Nigerian Picture: Genuinely Different, Not Just a Smaller Version of the Global Story

This is where the conversation needs real local grounding rather than simply importing Western headlines wholesale, and the data here is both specific and, in important respects, reassuring.

An independent mapping project scoring 280 Nigerian occupations on a 0 to 10 AI exposure scale found Nigeria's workforce averaging 3.6 out of 10, compared to an average of 5.3 out of 10 in the United States, with 53 percent of scored Nigerian jobs falling into the low-risk 0 to 3 range (aiexposure.com.ng, via DEV Community). The underlying reason is structural rather than a matter of luck: Nigeria's economy runs heavily on physical presence, cash transactions, and interpersonal trust networks that current AI systems simply cannot replicate. An okada rider navigating Lagos traffic scored 1 out of 10. A suya seller scored 0 out of 10. A POS agent operating on cash-based trust relationships scored 3 out of 10. A software developer, in sharp contrast, scored 9 out of 10, on par with the most exposed roles anywhere in the world, underscoring that for screen-based digital work specifically, geography provides little protection at all.

This connects to a broader structural reality that separate research has emphasised explicitly: roughly 81 to 82 percent of Nigeria's workforce operates in the informal economy (CSIS, "AI Strategies and the Informal Economy: Africa's Job Creation Test"), and in the short term, this vast informal sector is substantially buffered from AI-driven job losses precisely because current productivity gains from AI remain concentrated in desk-based, formal, professional roles that the informal sector largely doesn't contain.

But this protection comes with two serious caveats that deserve equal weight, not an afterthought.

First, Nigeria's formal sector shows sharp, concentrated exposure even as the broader workforce average looks comparatively low. A 2026 industry analysis found Nigeria faces roughly 9 million jobs at risk of automation by 2030, concentrated specifically in banking, public administration, and clerical services, the same routine-task categories that dominate every global study (BusinessDay NG). Clerical workers specifically, making up roughly 3.5 million of Nigeria's formal workforce, were found to face the highest AI exposure of any occupational category in the country, scoring 8.5 out of 10, a number essentially indistinguishable from the most exposed roles in wealthy economies (CTR0 Tech). The sectors employing the largest numbers of Nigerians, retail, financial services, agriculture, and logistics, are all described as currently being actively restructured by AI, not facing some hypothetical future disruption.

Second, the long-term outlook for the informal majority is genuinely less certain than the short-term buffering suggests. Analysis specifically focused on Africa's informal economy warns that while informal workers are relatively insulated from AI-driven job losses today, in the medium and long term they risk capturing only limited gains, or missing out entirely, on whatever broader economic upside AI eventually generates, precisely because national AI strategies across the continent, including Nigeria's, have so far concentrated almost exclusively on formal-sector industries (CSIS). The disruption reaching informal workers, where it's arriving at all, tends to look different from direct job replacement: AI increasingly reaches this part of the workforce indirectly, through the mobile platforms, delivery apps, and fintech tools, Bolt, Jumia, Opay among them, that are already reshaping how informal work gets organised and priced, even without replacing the human doing it (CTR0 Tech).

There is a genuine opportunity side to this picture as well, and it's worth stating plainly rather than only dwelling on risk. The same 2026 analysis projecting 9 million routine jobs at risk also projects roughly 11 million new technology-enabled jobs becoming available, but explicitly conditions that outcome on Nigeria investing seriously in digital skills development and workforce reskilling, and on the country shifting from simply adopting foreign-built AI tools toward building genuine local AI capability, infrastructure, and data ownership (BusinessDay NG). Without that investment, researchers warn, Nigeria risks becoming a high-volume consumer of AI technology built and owned elsewhere, while exporting the data that trains it, rather than a genuine participant in the value AI eventually creates.

What This Means in Practice

For job seekers and workers, the clearest actionable signal across both the global and Nigerian data is this: the determining factor is the actual task content of your specific role, not your industry label or job title in the abstract. Two people with the same job title can face very different exposure depending on whether their daily work is dominated by routine information processing or by physical presence, judgment, and interpersonal trust. A medical secretary and a surgical nurse share an industry but not a risk profile. A clerical officer at a Lagos bank and a field engineer at the same bank share an employer but not an exposure score.

For employers, particularly those reading this as part of a broader hiring and workforce strategy, the practical lesson is less about which roles to eliminate and more about where retraining investment will matter most. The global research is explicit that most firms observing AI exposure in their workforce are currently retraining rather than firing outright, and wages in AI-exposed roles have in many markets climbed faster than average as the nature of the work shifts toward oversight and judgment rather than disappearing entirely. For a Nigerian employer specifically, this points toward a clear priority: clerical, administrative, and routine data-processing roles are where proactive reskilling investment will pay off soonest, not as crisis management after a layoff, but as a deliberate response to a trend that's already visibly underway at companies like MTN and Kuda.

The Bottom Line

No job is fully immune to AI, and no honest analysis of this topic should claim otherwise. But exposure is neither uniform nor random. Globally, the pattern is clear and consistent: routine, codifiable, screen-based work faces the highest exposure, while work requiring physical presence, hands-on care, or deep human trust remains comparatively protected, for now. Nigeria's story sits inside that global pattern rather than apart from it, with a structurally lower overall exposure thanks to the sheer scale of its informal, physically grounded economy, but with sharp, globally comparable exposure concentrated precisely where the global pattern predicts it: clerical work, banking, and public administration. The most useful question any individual worker or employer can ask right now isn't "is my industry safe," but "how much of what I actually do every day is routine and codifiable, versus physical, judgment-based, or built on trust that only a human can provide." That question, asked honestly, predicts exposure far better than any job title ever will.

This article draws on current global AI exposure research, including studies from Microsoft Research, the Anthropic Economic Index, GovAI and the Brookings Institution, Hamilton Group, and the IMF, alongside Nigeria-specific research and occupational exposure mapping from BusinessDay NG, CTR0 Tech, aiexposure.com.ng, and CSIS, current as of late September 2026. AI capability and labour market impact are evolving rapidly, and the specific figures cited here should be treated as a snapshot of current research rather than a permanent forecast.