AI can help people explore ideas, learn skills and plan businesses. But without reliable power, internet access, trusted teachers, equipment and local knowledge, its answers may lead nowhere.
In November 2025, 20 students and their teachers gathered at Botoku Technical Institute in rural Ghana to explore artificial intelligence. Most of the students were young women learning dressmaking or hairdressing. Some had children and other responsibilities. Their questions were immediate: how do I find customers? What should I charge? How can I save enough to start a business?
Four students had iPhones, but only two had heard of AI. They worked in small groups, sharing those four phones and their teachers’ phones. Hesitation soon gave way to curiosity as they used AI to explore advertising, costs, savings and ways to turn a skill into an income. We ended the first day with more questions than we had begun with. The students were ready to continue the next morning.
There was no second day. The electricity failed, and the workshop had to be cancelled.
In a recent P&I piece, Sami Mahroum argues that AI is moving the bottleneck rather than removing it. Courts, research funders and government agencies can use AI to process far more material, but responsible human judgment cannot expand at the same pace. Automating the work around a decision does not necessarily improve the decision.
Seen from a rural vocational classroom, the bottleneck appears earlier. A student needs electricity, affordable internet and a device. She needs a teacher who can guide experimentation and the confidence to question an answer that sounds convincing but may be wrong. She also needs local knowledge. AI can suggest a price for a dress, but it cannot know what nearby families can afford, whether fabric is available or how many other dressmakers serve the area.
Even useful advice is not opportunity. A student may identify a promising business but still lack a sewing machine, start-up funds, transport or somewhere to work. The machine can answer her question without removing what prevents her from acting on the answer.
The ‘digital divide’ describes only part of this problem. Digital access allows someone to reach AI; capability allows them to use it for purposes they value.
The Botoku workshop does not prove that AI will transform rural vocational education. Twenty students cannot carry that claim. It does show what matters when introducing it. Twelve students wanted help starting a business, 10 finding customers, eight setting prices, six saving money and four finding employment. AI helped them explore those questions. The power failure showed how quickly possibility meets material reality.
One teacher offered perhaps the most valuable lesson. After students received an answer, he asked: “What question didn’t you ask it?”
We teach people to write better prompts. We should also teach them to notice what an answer leaves out. Whose knowledge is missing? What assumptions has the machine made? What would a customer or an experienced craftsperson ask? What can the machine simply not know?
The same questions apply to institutions introducing AI. It is easy to ask whether students can use the technology. It is harder to ask whether schools and governments are building the conditions for them to benefit from it.
Australia and Ghana are different societies. Rural Ghanaian communities should not be treated as interchangeable with remote Indigenous communities. Yet both settings can reveal the same policy mistake: providing technology while underinvesting in the people and local institutions that make it useful.
Australia could support locally governed AI learning spaces in remote schools, TAFEs, libraries and Aboriginal community-controlled organisations. These could begin with a few reliable devices, connectivity and trusted local facilitators. But equipment needs to be funded alongside teacher development, critical literacy, privacy safeguards and community control over purposes and data. Modest funds to test promising ideas would help learners move beyond plans they cannot put into practice.
For policymakers, the test is practical. Before funding an AI initiative, ask:
- Can people use it reliably? Are power, connectivity and devices available when needed?
- Can they question it? Is someone equipped to help learners test answers and recognise what is missing?
- Does it fit local life? Are community knowledge, language, priorities and control built into its use?
- Can people act on what they learn? Is there a pathway to the equipment, finance, markets or services they need?
- Who decides whether it worked? Will the people meant to benefit help judge the results and shape what happens next?
These questions bring the risks of error, commercial dependency and exposure of sensitive information into a guided, locally accountable decision. Without that care, well-resourced institutions will build their AI capability while disadvantaged communities receive occasional devices, brief demonstrations and another promise of transformation.
The blackout at Botoku does not mean communities must wait for perfect infrastructure. It reminds us that the machine cannot provide the conditions that make its answers useful. When the power returns, the question is whether the student can do something with what she has learned.
Komla Tsey
Komla Tsey is a Ghanaian–Australian writer, retired Professor of Education for Social Sustainability, and part-time Professorial Research Fellow at the Jawun Research Institute, Central Queensland University. His writing explores education, democratic culture, identity, Indigenous empowerment, colonial legacies and the moral certainties that shape public life. His forthcoming memoir, Botoku Child, traces a journey from a small village in Ghana to universities and communities across the world, examining how inherited beliefs, curiosity and lived experience shape both personal lives and democratic societies.