AI literacy training includes five things: how the tools work, prompting, judgement, safe use, and practice on real work. Anything less produces people who have seen AI but cannot yet use it responsibly.
The first part builds a mental picture. Learners find out what a model is, where its answers come from, and why it can be confidently wrong, and none of it requires coding. The second part is prompting, which means writing an instruction that gets useful output and then improving it when the first attempt misses. It is a learnable craft rather than a talent. The third part is judgement: knowing when AI output is good enough to use, when to verify it, and when to leave AI out entirely, such as in decisions about people. The fourth covers safe and legal use, including what information may go into a tool, client confidentiality, and Kenya's data protection rules, and this section protects both staff and employer. The fifth is practice on real work, because every module should end with a task from the learner's actual job.
As a concrete example, our Thinking With AI course at EddyLabs covers AI basics, a five step prompting method, and an introduction to agentic AI, which is AI that plans and completes a series of steps with limited supervision. It runs about six and a half hours and is designed for beginners.
Teaching this to staff uses the same five parts, with your company's own policies replacing the generic ones.