World Bank warns AI could widen Kenya's worker divide without better internet and skills

TEXT START: Artificial intelligence could help Kenyan workers improve productivity and access new economic opportunities, but unequal access to internet connectivity, digital skills and technology risks leaving some workers behind, according to a new World Bank report.

The Dissection

The article converts a structural ownership crisis into an infrastructure checklist. It frames AI disruption as a problem of internet access, skills, electricity and local data—real constraints, but not the terminal mechanism.

Its central move is reassurance through “augmentation.” If Kenyan workers can access AI, the article implies, they can remain productive participants. That preserves the comforting fiction that the main question is who gets the tools, rather than who owns the systems, captures the surplus and decides which human tasks remain economically necessary.

The 88 per cent “low exposure” figure is treated as evidence of protection. Under the Discontinuity Thesis, it is more accurately evidence of Kenya’s concentration in agriculture, informal services and manual work—sectors with weaker bargaining power and less ability to capture AI gains. Low direct exposure is not immunity from indirect displacement through automated management, cheaper competitors, platform control and collapsing demand for human labor.

The Core Fallacy

The article conflates access to AI with economic security.

AI does not need to replace every Kenyan worker directly. It only needs to make firms, platforms and capital owners more productive and more selective. Once AI-enabled operators can deliver comparable output with fewer people, labor demand and wages can fall even in occupations where workers never touch an AI tool.

The article also assumes that productivity gains will flow to workers as income. That is not a technical consequence. Without ownership or control, workers receive the residual bargaining position after capital has taken the gain.

Skills and connectivity may widen the pool of people who can use AI. They do not solve the ownership problem, the coordination problem or the collapse of productive participation. Under P1, cognitive automation keeps advancing. Under P2, institutions cannot preserve stable human-only economic domains at scale. Under P3, the majority can lose economic necessity even while AI access becomes more widespread.

Hidden Assumptions

• Worker augmentation will create enough new work to absorb displaced labor.
• “New activities” such as data services and digital entrepreneurship can scale faster than automation destroys existing bargaining power.
• Training is the binding constraint, rather than ownership of compute, platforms, capital, data and distribution.
• Low-exposure agricultural, informal and manual work will remain economically protected.
• Employers will share productivity gains instead of using them to reduce headcount, wages or prices.
• Better internet access gives workers agency rather than making them more legible, replaceable and dependent on platforms.
• Local-language AI and locally relevant data will be controlled by Kenyan workers or firms rather than external infrastructure owners.
• Governments can distribute AI benefits without confronting the concentration of AI capital.
• Gendered exposure can be managed through reskilling, despite women being concentrated in clerical, retail and customer-service roles that are vulnerable to automation.
• A delayed transition is equivalent to averted collapse.

Social Function

Primary classification: transition management and ideological anesthetic, with a substantial partial-truth component.

The article correctly identifies unequal access to connectivity, skills and technology. It also correctly notes that adoption will be uneven. But it packages those facts as a solvable development gap, allowing institutions to promise that infrastructure and reskilling will preserve broad participation.

That is elite self-exoneration in policy language. The system can claim the future remains inclusive if only Kenya builds enough digital capacity, while avoiding the harder question: who owns the AI systems that determine whether Kenyan labor remains necessary at all?

The Verdict

This is a competent early-warning report wrapped in a developmental lullaby. Better internet and skills will create niches, improve selected workers’ output and produce some Sovereigns and Servitors. They will not preserve mass employment by themselves.

Kenya’s low AI exposure is not a fortress. It is a lag. Lag delays the impact; it does not repeal the mechanism. The article mistakes the entrance ramp to AI adoption for a solution to the collapse of the wage-to-consumption circuit. Under the Discontinuity Thesis, the decisive divide will not be merely between connected and disconnected workers. It will be between those who own or control AI capital and those whose labor is no longer required.