Lesotho is reaching a useful transition point in its AI journey. The country has spent months building policy, training people and expanding the pipeline of young innovators. Now it needs a way to distinguish AI familiarity from the ability to use AI well on real work.
That distinction matters because employers and clients do not hire a certificate. They hire evidence that someone can solve a problem, exercise judgment and deliver a useful result.
UNDP’s current Sebabatso and STEAM programme is designed to help young Basotho innovators and entrepreneurs strengthen businesses, reach markets and build capacity. At the same time, the Government of Lesotho has just concluded a five-day workshop focused on turning the National AI Strategy and Data Centre Blueprint into an implementation roadmap. The next layer should connect those two efforts: every serious AI-skills programme should produce proof of work that another person can evaluate.
Turn Training Into Evidence
A conventional training programme often ends with attendance, a quiz or a certificate. Those measures tell you that someone was present and may have learned concepts. They do not tell an employer whether that person can use AI to improve a customer process, research a market, draft a reliable analysis, reduce repetitive work or make a better decision.
A proof-of-work apprenticeship would end differently. Each participant would complete a real task drawn from a business, nonprofit, public service or community need. The task should have a clear starting point and a visible outcome. The participant would show what the work looked like before AI, how AI was used, what had to be checked or corrected, and what changed afterward.
That creates a portable artifact. A jobseeker can show it to an employer. An entrepreneur can show it to a customer or funder. A training provider can compare results across cohorts. Government can see whether public investment in AI skills is producing usable capacity.
Require the Human Handoff
The most important part of the exercise should be the handoff between AI and human judgment.
Lesotho’s own AI policy direction emphasizes capacity, openness, trusted data and human oversight. Those principles become practical when a learner has to answer four questions: What did I delegate to AI? What did I keep for myself? How did I check the output? What would have made me reject or escalate it?
Consider a young entrepreneur using AI to prepare a marketing campaign. Producing more social posts is easy to measure, but volume is not the business result. A stronger assignment would require the entrepreneur to identify a target customer, use AI to develop and test messages, verify claims, track qualified inquiries, and explain why the final campaign changed from the first AI draft.
The same logic applies to a jobseeker. If someone uses AI to analyze a spreadsheet, the assessment should include whether they can spot a suspicious result, explain the reasoning in plain language and reproduce the core conclusion when the AI tool is unavailable.
Make Proof Portable
Lesotho already has strong reasons to make these artifacts portable. The government’s AI Skills Development Programme under Sebabatso is explicitly aimed at helping entrepreneurs use practical AI skills to reach wider markets and improve competitiveness. A portfolio of verified work gives those skills a market signal.
Each completed project could contain five simple elements: the original problem, the AI-assisted workflow, the human checks, the final artifact, and a short evaluation from a manager, client, mentor or instructor. No complicated credentialing system is required at the start. A structured one-page record is enough.
The evaluator should answer practical questions. Did the work save time without lowering quality? Did it improve a decision? Did it create a usable product? Did the participant catch mistakes? Could the participant explain the result without hiding behind the tool?
Those questions measure capability rather than enthusiasm.
Use Employers as Validators
Training becomes much more valuable when employers help define the tasks. Lesotho can invite businesses, civil-society organizations and public agencies to contribute small, bounded problems that participants can solve during training.
A retailer might ask for a better inventory-analysis workflow. A tourism business might need a customer-response process. An agricultural cooperative might want a clearer way to organize market information. A nonprofit might need to reduce administrative work while protecting sensitive information.
The organization does not have to hire the participant. It only has to supply a realistic problem and judge whether the result would be useful. That small connection gives learners exposure to the standards of real work and gives employers a low-risk way to see emerging talent.
Scale What Produces Results
Lesotho does not need to slow its AI push. It needs better evidence about what deserves to scale.
The current Sebabatso and STEAM pipeline already seeks innovators who can address development challenges in sectors including agriculture, technology, health, transportation, energy and finance. Adding proof-of-work apprenticeships would make the pipeline more legible to employers, investors and customers.
It would also protect against a common problem in technology programmes: confusing access to tools with productive capability. Giving people AI accounts, workshops and certificates can be useful. The economic value appears when people can apply those tools to messy, consequential tasks and still know when human judgment must take over.
If Lesotho wants AI capacity that travels from the classroom into businesses and public services, every training programme should leave behind something more convincing than a certificate. It should leave behind work that another person can inspect, question and trust.
Gleb Tsipursky, PhD, a behavioral scientist, CEO of Disaster Avoidance Experts, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026). https://disasteravoidanceexperts.com/aibook