Welcome: what this course will change in your work

Every finance professional now works next to a machine that can read a hundred-page annual report in seconds, draft a cash flow statement from a trial balance, and write variance commentary that sounds like a seasoned financial controller. It can also invent a number, misapply a standard with total confidence, or leak a client's confidential data into a tool your firm never approved. Both things are true at the same time, and that is exactly why this course exists.

The professionals who thrive over the next decade will not be the ones who use AI the most. They will be the ones who use it well: who know which tasks to hand over, how to instruct it precisely, how to check what comes back, and how to sign their name to the result with a clear conscience. That is the skill we are going to build together, one lesson at a time.

What AI is already good at in finance

Think of today's AI tools as a very fast, very well-read junior colleague who never gets tired and has never worked at your company. That colleague is excellent at some things straight away:

• Reading and summarising. Contracts, board papers, annual reports, results call transcripts, regulatory notices. AI can pull out the lease terms, the revenue recognition clauses or the covenant ratios in minutes.
• First drafts. Variance commentary, management report narrative, accounting policy notes, audit planning memos, emails to the business asking for missing information.
• Reshaping data. Turning a messy bank export into a clean table, mapping a trial balance to statement lines, converting a schedule into journal entries.
• Explaining and teaching. Walking you through how a standard applies to a fact pattern, or how a formula works, at whatever level of detail you ask for.
• Writing code and formulas. Excel formulas, Power Query steps and short Python scripts that automate the dull parts of your week.

And there are things that colleague is not ready to do alone:

• Final numbers. AI does not "know" your ledger. Anything it calculates must tie back to source data that you control.
• Judgements. Whether a contract contains a lease, whether a receivable is credit-impaired, whether a going concern disclosure is needed: AI can lay out the arguments, but the judgement and the accountability stay with you.
• Anything confidential, in the wrong tool. Client data, unpublished results and personal data belong only in tools your organisation has approved for that purpose.

The two lenses of this course

This course looks at every topic through two lenses at once.

The AI practitioner's lens asks: how does the tool actually work, what is it likely to get wrong, and how do I make the output reliable? You will learn enough about how large language models and machine learning behave to predict their failure modes, and you will build habits such as structured prompts, grounding in source documents and independent checks that turn a clever demo into a dependable process.

The investment professional's lens asks: what do these numbers mean, who relies on them, and what do I owe those people? This is the discipline of the analyst and the portfolio manager: valuation, risk, the quality of earnings, and the ethical duty to have a reasonable basis for every conclusion you communicate. A beautifully formatted AI-generated report that misleads a reader is not a productivity gain. It is a professional failure.

When the two lenses agree, you have something valuable: work that is faster and better than what you could produce alone.

How the course is organised

There are twelve modules, and they build on each other:

Each lesson follows the same rhythm: the idea in plain English, a worked example with real-looking numbers, prompt templates you can use the same day, a "Check it" list for verifying the output, a lecturer's note on the judgement call, and a short practice task. Each module ends with a quiz.

Meet the companies we will work with

To keep the examples concrete, we will return to three fictional organisations throughout the course:

• Ardena Foods Plc, a fictional Nigerian food manufacturer with a head office in Abuja and factories in Kaduna and Ogun. It reports under IFRS in Naira, has revenue of about ₦48 billion, and buys some raw materials in US dollars. We will close its books, build its financial statements and, in the capstone, produce its reporting pack.
• Velra Capital, a fictional Abuja asset manager running Nigerian equity and fixed income funds. We will use it for research, valuation and portfolio work.
• Tamari Microfinance Bank, a fictional microfinance bank. We will use it for credit scoring, expected credit losses, anti-money-laundering monitoring and fraud detection.

A few facts about Ardena hold throughout: it has 4,000 million ordinary shares of 50 kobo each, and Ardena Logistics Ltd is its wholly owned haulage subsidiary. Beyond those, each module builds the figures it needs for what it teaches, so the numbers agree within a module but not always between modules. The one thread that runs all the way through is the set of IFRS financial statements you build in module 5: the capstone in module 12 picks up from exactly where they end.

Worked example

Here is the kind of task you will do routinely by the end of the course. Ardena's financial controller has the monthly management accounts and wants a first draft of the commentary for the finance director.

A weak prompt is "Write commentary on these numbers." The AI will produce fluent text, and it may invent reasons such as "strong demand in the North" that nobody has told it.

A professional prompt gives the role, the context, the data, the facts behind the variances, the format, and the rules. The controller adds what the business has confirmed: a price increase of about 4% from August, and maize and imported packaging costs pushed up by the weaker Naira. The AI is told to use only those reasons, to show its arithmetic, and to flag anything it cannot explain.

The result is a short, accurate paragraph: revenue was ₦170 million ahead of budget, mainly because of the August price increase; cost of sales was ₦240 million over budget, because input costs rose faster than prices; and gross margin therefore fell three points, from 31.9% to 28.9%. It ends with a flagged question: is the remaining volume effect in revenue confirmed by the sales team? The controller checks the figures against the ledger, answers the question, and has a finished draft in ten minutes instead of an hour.

Prompt templates

You are an experienced financial controller writing for a finance director.
Context: [company, period, reporting basis].
Data: [paste the table of actual, budget and variance].
Confirmed explanations from the business: [list them].
Task: write variance commentary of no more than [150] words.
Rules: use only the explanations given; show each figure you quote exactly as in the table;
list any variance you cannot explain under "Questions for the business".

Use this whenever you need management commentary that must not invent reasons.

Before you answer, restate the figures you will rely on in a table, then check that each
subtotal adds up. If anything does not add up, stop and tell me instead of continuing.

Add this to any prompt that works with numbers. It makes arithmetic errors visible early.

Check it

• Every figure in the commentary appears in, or is calculated correctly from, the source table.
• Every reason given was supplied by the business, not invented by the AI.
• Percentages and margins are recalculated by you, not taken on trust.
• The "Questions for the business" list is answered before the draft goes anywhere.
• No confidential data went into a tool that is not approved for it.

Lecturer's note: The most common mistake new users make is judging AI output by how professional it sounds. Fluency is not accuracy. Train yourself to read AI-generated finance text the way an auditor reads a draft set of accounts: assume nothing, tie everything back to source, and ask "how do we know that?" of every explanation.

Key points

• AI is a fast, well-read junior colleague: excellent at reading, drafting, reshaping and explaining, but never the final authority on numbers or judgements.
• This course uses two lenses together: the AI practitioner's (reliability) and the investment professional's (meaning, risk and duty to users).
• Good prompts state the role, context, data, facts, format and rules, and ask the AI to flag what it cannot support.
• Verification is part of the work, not an extra: every figure ties to source, and every reason comes from the business.
• Confidential data goes only into tools your organisation has approved for it.

Practice task

Start your AI log: a simple document or spreadsheet where you will record, for every task you try with AI during this course, the task, the prompt you used, what went wrong, how you checked it, and the time saved. Then take one real (or invented) monthly variance table from your own work, remove anything confidential, and use the first prompt template above to draft commentary. A good answer is a commentary in which every number ties to your table, no reason appears that you did not supply, and at least one question for the business has been flagged.