Skills, jobs, and AI

A nurse, farmer, teacher, or designer does many kinds of work. AI may speed up a narrow task while judgment, relationships, and accountability remain human responsibilities.

Skills such as asking good questions, explaining clearly, collaborating, and learning new tools can transfer across changing workplaces. Build them alongside technical confidence.

Fears that machines will abolish work are older than electricity, and the historical record shows a consistent, more complicated pattern that economists describe with a task-based model of labour. A job is a bundle of tasks, and technology rarely automates the bundle; it automates specific tasks within it, which changes the job, shifts where the human effort goes, and alters how many workers are needed — sometimes upward. The classic worked example is the automated teller machine. When ATMs spread through banking from the 1970s onward, the obvious prediction was the end of the bank teller. Instead, in the United States the number of tellers grew for decades afterward. The mechanism: ATMs made each branch cheaper to run, so banks opened many more branches, and the tellers' job shifted from counting cash to customer relationships and selling services — the tasks machines could not do. The same pattern appears elsewhere. Spreadsheets automated arithmetic and eliminated armies of manual bookkeeping clerks, yet the number of accountants and analysts rose, because cheap calculation made financial analysis worth doing everywhere. Weaving machines in the nineteenth century destroyed the livelihoods of skilled handloom weavers — the human cost of transitions is real and fell hardest on those least able to retrain — while making cloth so cheap that total employment in the industry initially expanded. Three lessons carry into the AI era. First, automation hits tasks, so the right question about any role is which of its tasks are routine enough to codify, not whether the whole job survives. Second, when a task becomes cheap, demand for the surrounding work often grows, and entirely new roles appear around the technology, as social media managers appeared around platforms. Third, the transition is uneven: gains and losses land on different people, in different places, on different timelines, and countries like Zimbabwe experience these shifts through their own labour market structure, where a large informal sector changes how technology arrives. The misconception to correct is the lump of labour fallacy — the idea that an economy contains a fixed amount of work, so any task a machine takes is a task a person loses forever. Work is not a fixed lump; it expands and reshapes as costs change. That is not a promise that everyone wins, but it does mean the useful preparation is watching tasks, not mourning job titles.

There is a precise economic idea that explains how people remain valuable even when machines outperform them at many things, and it is worth learning properly because it converts vague anxiety into a practical strategy. The idea is comparative advantage, borrowed from trade theory. It says that the sensible division of labour depends not on who is absolutely better at a task, but on what each party gives up to do it. Even if an AI system were better than a person at both drafting documents and comforting an anxious client, the AI's time spent drafting costs almost nothing, while a human hour spent drafting is an hour not spent on the client work that only humans can supply. The efficient arrangement gives the machine the tasks where its advantage is overwhelming and cheap, and concentrates human effort where machines are weak — regardless of whether the human is second-best at everything else. Work through a concrete example: a small Harare accounting firm adopts AI tools that draft financial statements and flag anomalies in minutes, work that once filled most of a junior accountant's week. The junior accountant does not become useless; her scarce, expensive hours move to the tasks where the machine is hopeless — sitting with a shop owner to understand why the numbers look wrong, judging whether an unusual transaction is an error or a warning, taking professional responsibility for the filing, and building the trust that brings the client back. The firm now serves more clients, and the accountant's judgment is applied to more decisions than before. This lens identifies which human capabilities the machine era makes more valuable, not less: judgment under ambiguity, where the right answer depends on context no dataset captured; responsibility, because clients, courts, and regulators require an accountable person; relationships and trust, which are built between people; physical dexterity in unstructured environments, which remains hard for robots; and the ability to define what is worth doing at all — machines optimise toward goals, they do not choose them. The misconception to correct is that competing with AI means racing it at its own strengths, memorising more or calculating faster. That race is lost and was never the point. The strategy comparative advantage recommends is the opposite: learn to use the machines fluently for what they do cheaply, and deliberately deepen the human capabilities they make scarce and precious.

An agricultural extension worker considers using AI to draft farmer-friendly advice. A responsible response is: Check the draft against local weather, crops, and expert guidance before sharing. Correct. Local conditions and expert review are essential. Use this case to separate what the technology contributes from what people contribute. The team should compare the intended outcome with a baseline where applicable, record important assumptions, and keep a clear route to correct or stop the process.

Ask whether the examples, data, and assumptions fit your school, company, or community. Generic demos often miss local names, laws, connectivity, and languages.

Workshops and teamwork build checking, briefing, and domain skill — the abilities AI makes more valuable.