The Shadow of the Human Advantage

I have spent much of this year writing about judgment, empathy, contextual reading and relational intelligence that allow a person to understand what another person needs, build trust and solve problems in ways that AI still struggles to replicate.

I still believe this matters.

But a report released recently made me confront the darker side of that argument.

INTERPOL’s African Cyberthreat Assessment 2026, drawing on data from 36 African countries, found that artificial intelligence now drives 55 percent of reported cybercrime across the continent. Reported losses have more than doubled since 2024, from $192 million to $484 million. Seventy-two percent of the countries surveyed reported organised scam centres operating within their borders. The patterns vary by region: mobile-money fraud and infrastructure ransomware are prominent in East Africa, while romance scams and business email compromise feature heavily in Central and West Africa.

Read that pattern again.

Romance scams. Business email compromise.

These are not simply attacks on computer systems. They are attacks on relationships.

And relationships are precisely where I have argued that human capability remains unusually valuable.

That is the shadow of the argument.

What the criminals actually learned

The report does not make this connection. But it made me see one.

The skills I have written about for months — reading a person accurately, understanding context, knowing what someone needs to hear before they say it, holding a conversation with enough warmth and credibility that a stranger begins to trust you — are not inherently good or bad.

They are capabilities.

A person can use them to sell a product, help a customer, manage a team or solve a difficult problem. The same capabilities can also be used to manipulate someone.

AI makes the problem more serious because it changes the economics of deception.

A scammer no longer needs to personally possess exceptional writing ability, cultural fluency or conversational skill. AI can supply much of it.

A badly written scam email used to be its own warning sign. That warning sign is disappearing.

A criminal can now generate convincing messages, adapt them to different audiences, respond in real time, translate across languages and personalise interactions at a scale that would previously have required an army of people.

What once required genuine social skill can increasingly be simulated.

That matters for Africa in particular because of the argument I have been making about the value of contextual and relational capability in an AI economy.

If an AI-fluent African operator can use contextual intelligence to understand a customer, navigate a complex business environment and build trust, an AI-assisted criminal can use the same underlying capability to manufacture trust that does not deserve to exist.

The advantage is real.

So is the shadow.

This is not a reason to abandon the thesis

It does not mean the human-layer argument is wrong.

It means the human layer is not automatically good.

It is a capacity, and like every meaningful capacity, it can be directed toward very different ends.

More importantly, AI is beginning to change what we should think of as the scarce resource.

For years, one of the strongest defences against fraud was the difficulty of producing convincing human interaction at scale. Now AI is steadily reducing that constraint.

The result is not that human judgment becomes irrelevant.

It is that the ability to distinguish genuine judgment from synthetic social intelligence becomes more valuable.

That is a very different problem.

And it takes us beyond the question of what individuals can do with AI and toward a more important question: what institutions are capable of doing with the people who can use it?

The institutional gap

The problem the INTERPOL’s report exposes is not simply a gap in African talent. It is also a gap in institutional capacity.

Cybercrime legislation across the continent remains fragmented. AI readiness within law enforcement agencies is low. Banks, telecom operators and police do not yet share information in real time at the scale required to respond effectively. And many of the legal and investigative frameworks being used to confront today’s cybercrime were designed before generative AI made sophisticated social engineering cheap and scalable.

This is a familiar diagnosis.

It is the same problem we see in education, universities and labour markets: technology changes faster than the institutions responsible for helping people understand, use and govern it.

Africa does not have a shortage of capable people.

But capability without institutions can become either underutilised or misdirected.

Training AI-fluent operators without simultaneously strengthening the investigators, regulators, legal systems and cross-border networks capable of detecting AI-enabled crime is building only half the house.

The opportunity hiding inside the problem

There is, however, an opportunity here.

Every one of the capabilities that makes an effective social engineer also has a legitimate defensive application.

Reading a pattern that does not fit.

Sensing that something is wrong before the data fully confirms it.

Understanding how a particular community communicates.

Recognising when a message sounds superficially right but contextually wrong.

These are forms of human judgment.

The question is where we point them.

Africa does not only need more operators who can build with AI. It needs operators who can investigate with AI, regulate with AI, defend with AI and recognise when AI is being used to deceive.

That is not a peripheral requirement for the AI economy.

It is part of its infrastructure.

The same gift

I still believe that the judgment a young African develops while solving real problems under real constraints can be one of the continent’s deepest advantages in an AI-enabled economy.

INTERPOL’s report has not changed that conviction.

It has sharpened it.

An advantage this real was always going to attract people who wanted to misuse it.

And that means we need to stop talking about human capability as though its value is self-evident. It isn’t.

The same contextual fluency that helps someone solve a problem can help someone exploit one.

The same AI that gives a young African worker unprecedented productive capacity can give a criminal unprecedented capacity to deceive.

The question, then, is not whether AI will amplify human capability.

It is which capabilities we choose to amplify, which institutions we build around them, and whether those institutions are strong enough to tell the difference between trust that has been earned and trust that has been engineered.

That is the shadow of the human advantage.

And what happens next will depend less on the technology than on what we build around the people who use it.