You will be able to build a one-page list of the AI in your working life, sorted by layer and by risk.
Nurul is an HR executive at a logistics company in Jurong (an invented example). Ask most people how much AI they use at work and they name one chatbot and stop, and Nurul would likely have done the same.
When the same people walk through a normal working day, the count rises quickly. Take a Tuesday (an illustrative day): the email filter sorts the inbox, Outlook autocomplete finishes a sentence, a meeting produces a transcript, someone taps the WhatsApp translation button, and a dashboard flags overdue invoices. The exercise here turns a typical day into an inventory, a one-page table showing where AI already touches your work and where its mistakes would hurt. Allow about twenty minutes. You need the inventory table template that comes with the lesson, and last week's calendar or sent folder to jog your memory.
Go through a normal working week day by day and think about each tool you opened. Ask one question of each: does it make a prediction, suggest something, or produce content for you? If yes, it goes on the list.
Be generous and include small things. Suggested-reply chips in email count, as do a spellchecker that rewrites a whole phrase, an HR system that ranks job applicants, a calendar that proposes meeting times and a bank app that sorts your spending into categories. Add the tools your company runs on your behalf that you never see, such as a filter that quarantines suspicious attachments. Aim for at least eight rows; most people find twelve or more once they look.
Beside each tool, write which of the three layers from Lesson 1.1, "AI is a family of techniques, not one thing", it belongs to.
A rule-based tool runs on logic a person wrote in advance, such as a form that rejects a claim over a fixed limit. A machine learning tool learned a pattern from past examples and gives you a label, score or ranking, such as a spam filter or a recommendation list. A generative tool produces new text, images or audio, such as a chatbot, a meeting summariser or an image tool.
If you can't tell which layer a tool belongs to, put a question mark rather than guess. Vendors rarely explain how a feature works, so a question mark is usually the honest entry, and it also gives you a list of tools to ask the IT team about.
For each generative tool, write what you check before you use its output and what you leave unchecked. Base this on what you actually do on a busy day, even if you would do more with time to spare. An honest note might say: "I paste the meeting summary into an email to my boss without checking it against my own notes." You are recording real habits so you can see them.
Generative tools usually fail by producing text that reads smoothly and is wrong. Their errors reach other people through whatever you skip, so the unchecked column shows you where that can happen.
Go down the table and ask of each row what would happen if the tool got it wrong and nobody noticed. A spellchecker changing "their" to "there" causes minor embarrassment. A summary that drops a client's deadline, or a screening tool that filters out a strong candidate, costs far more. Weigh each one in money, reputation or someone else's welfare, then circle the single tool whose unnoticed mistake would cost most. Pick the one you really think is riskiest, even if it is a tool you like. Modules 6 and 8 keep coming back to the tool you circle.
Nurul's table is here to show the method; her tools and habits are her own, so build your list from your week. Her rows, in order:
An applicant tracking system that ranks CVs for each opening: machine learning, with a question mark because the vendor only describes it as "smart matching". She reads the top twenty CVs and never opens the rest. Outlook suggested replies: machine learning, low stakes. Teams meeting transcription: the speech-to-text part is machine learning and the summary is generative. She checks names and dates in the summary but not the action items. A leave system that blocks requests beyond the leave balance: rule-based. ChatGPT for drafting job descriptions: generative. She checks the salary range and job title but not the list of duties, which she has noticed it pads out. Payroll software that calculates CPF contributions: rule-based. Google Translate for messages from a foreign worker's family: generative, because it writes new text in another language. She doesn't check it at all, since she cannot read the language. LinkedIn suggested candidates: machine learning.
She circled the applicant tracking system, because a ranking she cannot see inside decides which CVs a person ever reads, and she never looks below the top twenty. Translation came a close second, since an error would reach a family who could not tell her it was wrong.
Once the table is full, read it again from top to bottom. Most people notice two things: they use more AI than they thought, and the tool they trust most is often one they never check. Keep the page somewhere you can find it, as you will compare it with your answers at the end of the course.
Open the template now and start with Monday.
Fill in the tool inventory table with at least eight rows, then write two sentences on which tool you trust most and which least, and why.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).