Skillnaija Start free

AI & EMERGING TECH

Agentic AI: Build and Supervise AI Agents That Take Action, Safely

Hand real business chores (chasing invoices, sorting enquiries, drafting reports) to small AI agents on free tiers, with a human approving every send and every naira. Honest about what agents cannot do.

  • 12 lessons
  • ~3 hours
  • Intermediate
  • Project + certificate

Free account required to open the lessons. Any price is shown before you pay.

What you'll learn

  • Telling a chatbot, a workflow and an agent apart, and choosing the right one
  • The agent loop: goals, plans, actions, observations and stop conditions
  • Designing narrowly scoped tools, and what MCP is
  • Building a no-code agent on a free tier with a human approval step
  • Running a first agent on a laptop with a free model tier and made-up data
  • Memory, knowledge packs and retrieval
  • Permissions, approvals, sandboxes and stop buttons
  • Prompt injection and tool-use risks, and designing limits
  • Testing, logging and costing an agent
  • Nigerian data protection and consumer rules that apply to agents
  • Scoping and pricing a small, honest automation service

About this course

A practical course that follows Efe Omoregie, the operations officer at a school and office supplies distributor in Benin City, as she hands three real chores (chasing overdue invoices, qualifying enquiries and drafting the weekly report) to small AI agents and plain workflows without letting them send, pay, delete or decide on their own. Her cautious finance manager asks the hard questions, and a developer friend guides the laptop work. You will learn what an agent is and is not (chatbot, workflow or agent), how the loop of goal, plan, act and observe works, how tools and MCP let an agent do things, and when a plain workflow beats an agent. You will build a first no-code agent on a free hosted tier with a human approval step, which works in a phone browser, and, on an optional laptop track, run a first agent on a free model tier with made-up data. Then you make it safe and honest: memory, knowledge and retrieval; permissions, approvals and sandboxes; prompt injection, where there is no fool-proof prevention, only careful design; testing, logs and what an agent really costs; and the rules in plain words (the Nigeria Data Protection Act, impact assessments, data leaving Nigeria, consumer law, and what is still only a draft). The last lesson covers where the work is and what a beginner can credibly offer a business in three months. Product names and prices sit in dated boxes, checked on 4 October 2026, because the tools change and the ideas do not. Lessons include narrated animated explainers, credited photos and short clips; videos play only when you tap them. Free to learn.

Monday in August, 8 am, at Moatside Supplies Ltd, a school and office supplies distributor in Benin City. Efe Omoregie, the operations officer, opens her laptop to the same three chores as every week: chasing overdue invoices, sorting enquiries from the website form and drafting the weekly report. Together they take 14 of her hours. Her phone buzzes. Mr Jude Osazee, the managing director, has forwarded a video: "AI agents will run your whole business." Under it he has typed, "Efe, sharp sharp, can we get this?" Mrs Roseline Odia, the finance manager, has seen it too. She does not look up. "Which part of my ledger is this thing going to touch? If it touches money or a customer, a human presses the button." This course is Efe's answer to both of them.

Who it's for

Intermediate. Some grounding in the subject will help you move faster.

The lessons 12 lessons, ending in a graded project

  1. 1Chatbot, workflow or agent: what agentic AI is, and is not17 min
  2. 2The loop: goal, plan, act, observe16 min
  3. 3Tools and MCP: how an agent actually does things16 min
  4. 4When a plain workflow beats an agent16 min
  5. 5Your first no-code agent, with a human approval step17 min
  6. 6Laptop track: a first command-line agent on a free plan18 min
  7. 7Memory, knowledge and retrieval16 min
  8. 8Permissions, approvals and sandboxes: what an agent never does alone17 min
  9. 9Prompt injection and other tool-use risks16 min
  10. 10Testing, logs and cost: knowing that it works18 min
  11. 11The rules in plain words: data protection, consumer law and what is still a draft18 min
  12. 12Where the work is, what you can sell in three months, and your project17 min
  13. Project: An agent pack for one real choreGraded

THE PROJECT AND YOUR CERTIFICATE

An agent pack for one real chore

One shared Google Doc link with the chore brief and decision, the loop trace and tool list, evidence of a working build with its knowledge pack and approval sheet, the permission ladder and stop plan, the risk check, the test results with log rows and a cost estimate, and the rules note.

Certificates on Skillnaija are earned, not issued for attendance: you submit this project, an examiner grades the real work against a published rubric (you need 65% to pass), and passing issues a certificate anyone can verify. If you don't pass yet, you get written feedback and can try again.

  • The chore is real and specific, each piece has a justified workflow-or-agent decision, and the goal and traced loop show a task an agent can actually finish and stop20%
  • The tools are scoped narrowly, the build works on made-up data, and every outbound, money or irreversible action sits behind a human approval that the evidence shows was really used25%
  • The permission ladder, the five never-alone lines, the risk table and the harmless injection test show that limits were designed in, not assumed20%
  • The test set has clear pass rules and repeat runs, the log is readable and free of personal data, and the cost estimate shows its working (in tokens or in the builder's own unit) in the provider's currency with a date20%
  • The rules note keeps personal data out or minimised, names the human who decides, separates law from provider terms and drafts, and the submission contains no real personal data, logins or keys15%

See the full project brief, rubric and free workbook →

Frequently asked questions

Is Agentic AI: Build and Supervise AI Agents That Take Action, Safely free?

Most Skillnaija courses can be started free with a free account: your first two courses are free to start, then one new course each month. Some courses are paid; the course page in the app shows whether Agentic AI: Build and Supervise AI Agents That Take Action, Safely is one, and its price, before you pay. You can also buy any course on its own for twelve months, or open every course with the All-Access Pass. Certificates are never sold: you earn one by passing the course's project.

How long does Agentic AI: Build and Supervise AI Agents That Take Action, Safely take to complete?

6 weeks · ~3 hrs/week. There are 12 lessons, about 3 hours of material in total. You learn at your own pace and your progress is saved.

Do I need any experience to start Agentic AI: Build and Supervise AI Agents That Take Action, Safely?

Intermediate; a phone is enough, a laptop only for optional Lesson 6. Some prior grounding in the subject will help you move faster.

What will I learn in Agentic AI: Build and Supervise AI Agents That Take Action, Safely?

You will build practical skills in Telling a chatbot, a workflow and an agent apart, and choosing the right one, The agent loop: goals, plans, actions, observations and stop conditions, Designing narrowly scoped tools, and what MCP is, Building a no-code agent on a free tier with a human approval step, Running a first agent on a laptop with a free model tier and made-up data, Memory, knowledge packs and retrieval, Permissions, approvals, sandboxes and stop buttons, Prompt injection and tool-use risks, and designing limits, Testing, logging and costing an agent, Nigerian data protection and consumer rules that apply to agents, Scoping and pricing a small, honest automation service. Every lesson is written in plain English, with examples drawn from how the work is actually done and paid for, in Nigeria and beyond.

Do I get a certificate for Agentic AI: Build and Supervise AI Agents That Take Action, Safely?

Yes, but it has to be earned. You complete the course project — An agent pack for one real chore and submit it for grading. Passing the project issues a verifiable certificate; finishing the lessons alone does not.

Ready to start Agentic AI: Build and Supervise AI Agents That Take Action, Safely?

It is self-paced, and it ends with something you can show an employer.

Start learning free