Key point: Certificates can show that a learner completed digital-skills training, but employers need stronger evidence of practical ability. A five-part work-sample portfolio—covering an unfamiliar task, a tool change, failure recovery, documented handoff and a measurable outcome—can show what a graduate can actually do.
Editor’s note: This is a contributed expert article. The views expressed are the author’s own.
Nigeria has no shortage of ambition around digital skills. Applications opened in late August for the fully funded Tech4Youth Advanced ICT Skills Training Programme in Enugu, offering 12 weeks of practical training in artificial intelligence, data analytics, cybersecurity and software development. Participants are also promised mentorship, career guidance, portfolio development, workplace preparation and support for employment, freelancing or entrepreneurship.
That is the right direction. Yet training programmes can still fall into a familiar trap: measuring what is easiest to count rather than what employers and clients ultimately need.
A certificate proves that someone completed a course. It does not necessarily prove that the person can solve an unfamiliar problem, adapt when the tool changes, recover when an AI system fails, explain the reasoning behind an output or hand the work to another person. Those capabilities matter even more as AI makes it easier to produce competent-looking first drafts, analyses, designs and code.
Nigeria’s larger 3 Million Technical Talent initiative makes this distinction increasingly important. The federal programme is designed to build Nigeria’s technical-talent pipeline at national scale, with AI and machine learning among its priority skills. As more Nigerians gain access to training, the challenge shifts from access alone to proving which learners can turn training into reliable work.
Training providers should therefore give every graduate a work-sample portfolio, not merely a certificate. That portfolio should test five forms of practical competence.
1. Test Learners With an Unfamiliar Task
Students usually improve rapidly on exercises that resemble what they practised during training. Employers rarely have that luxury. A junior analyst may be asked to clean a messy dataset that does not resemble the classroom example. A marketer may face a client with poor records and conflicting goals. A software developer may inherit undocumented code. This emphasis on practical learning can begin much earlier: parents choosing a school can assess whether students have opportunities to apply knowledge, use technology and solve unfamiliar problems, as discussed in our guide to choosing a secondary school.
Before graduation, learners should complete at least one consequential task whose structure they have not seen before. The point is not to surprise them for the sake of difficulty. It is to see whether they can identify the problem, choose an appropriate approach and recognise what they do not know.
2. Require a Change of Tool
AI training can accidentally turn into product training. A learner becomes proficient with one chatbot, coding assistant, analytics package or automation platform and mistakes familiarity with that interface for durable competence.
Programmes should require students to repeat a bounded task with a different tool or model. If performance collapses, the learner may have memorised a workflow rather than understood it. If performance transfers, the portfolio now contains evidence of a more valuable capability: understanding what the task requires independently of the software used to perform it.
3. Assess Failure Recovery
AI systems fail in ordinary ways. They invent facts, misunderstand context, produce brittle code, apply the wrong assumptions or return a polished answer that solves the wrong problem. Entry-level workers need practice detecting those failures before they are trusted with consequential work.
Give the learner a deliberately flawed output and ask for a recovery. The portfolio should show what went wrong, how the person detected it, what evidence they checked, how they corrected the work and what guardrail would reduce the chance of the same failure recurring.
This matters psychologically as well as technically. When people are trained to treat the system as an authority, they may hesitate to challenge it. When training normalises verification and correction, questioning the tool becomes part of competent performance.
4. Include a Documented Handoff
Work rarely ends with the person who started it. A useful digital workflow must survive vacation, turnover, promotion or simple collaboration. The learner should document one meaningful workflow well enough that another qualified person can reproduce it without a private briefing.
That reveals hidden knowledge. If the second person cannot understand which inputs matter, where the data came from, what the AI was allowed to do or when a human should intervene, the workflow remains dependent on its original builder.
AfricanBase recently highlighted the ECOWAS Young Graduates Professional Immersion Programme, which is built around hands-on professional experience rather than another classroom credential. That logic should carry into digital-skills programmes. Employers learn more from evidence that a graduate can perform and transfer real work than from another line on a CV.
5. Connect the Portfolio to a Real Outcome
Every portfolio should connect at least one project to a result that somebody outside the classroom values. That might mean reducing the time required to reconcile records, improving the accuracy of a report, helping a small business respond to customers faster, finding a software defect, producing a usable dashboard or completing a client deliverable that meets agreed requirements.
The outcome does not need to be spectacular. It needs to be inspectable. Students should state the baseline, the work performed, the human checking required, the result and the limitations.
How Employers Can Use the Five-Part Portfolio
This approach would also help employers. Instead of asking applicants whether they “know AI,” a vague question that invites vague answers, hiring managers could ask candidates to walk through five pieces of evidence:
- What unfamiliar task did you solve?
- What happened when the tool changed?
- Show me a failure you caught and corrected.
- Could another person reproduce your workflow?
- What measurable result did the work create?
The answers would give hiring managers more useful evidence than a broad claim of AI proficiency.
What Training Providers Can Learn
Training providers would benefit too. If graduates repeatedly struggle with the same stage, that is useful feedback about curriculum design. A cohort that performs well in guided exercises but poorly on unfamiliar tasks needs more independent problem-solving. Weak handoffs point to documentation gaps. Frequent failure-recovery problems suggest that verification deserves more practice.
Nigeria is right to expand access to technical training. The next step is to make the evidence of competence as scalable as the training itself.
A certificate should mark the completion of learning. A work-sample portfolio should show what the learner can now do when the tutorial ends.
About the Author
Gleb Tsipursky, PhD, is a behavioural scientist, CEO of Disaster Avoidance Experts and author of The Psychology of AI Adoption at Work: From Resistance to Results, published by Georgetown University Press in 2026.
