List the decisions a personal AI policy should settle before you need them.
A personal AI policy is a short set of rules you write once, while nothing is urgent, and then follow. It means you no longer work out every AI decision from the beginning each time one comes up. You have already done most of the thinking for it in earlier modules.
Jia Hui, the marketing executive from module 1, had no such rules on a busy Tuesday. A supplier emailed her a spreadsheet of customer delivery addresses and asked her to "get AI to sort out the duplicates". An assistant could have done the job in thirty seconds. Instead she had to stop and think it through: which tool to use, which account, whether the spreadsheet held personal data, and whether that made a difference. After ten minutes (also invented) she still had no answer. A written list would have told her straight away that customer addresses are personal data and belong in the approved account only. It would also have reminded her that they are safer to process once the names are taken out, which lesson 2.2 explained.
Your policy has rules in four areas. The first is tools, accounts and settings. The second is what never goes in. The third is what you always check, together with your safe word. The fourth is what you keep doing yourself. Each area draws on something you produced in the course, mainly your audit table, your task map and your safe word plan.
Rule 1 says which assistants you use, on which accounts, with which settings. Take the audit table you built in lesson 1.4, "Audit the AI tools you used this month", and copy over the tools you chose to keep. Beside each one, note the account type, what you use it for and the settings you picked. Lesson 1.3, "History, memory, temporary chats and deletion", went through the settings that matter most. An app update can change settings without telling you, and a written record gives you something to check against after each update. Jia Hui's version sends work tasks to the company's Copilot and personal tasks to her own ChatGPT, with training switched off and memory reviewed once a month.
Rule 2 uses the three data groups from lesson 1.1, "What happens to the words you type into an assistant": fine anywhere, approved account only, and never. Put each kind of data you handle into one of the three. Then write your never list, the data that must not go into any AI assistant, in everyday words. Most people's list includes:
passwords and one-time codes NRIC numbers bank and card details health or financial information about other people anything a client or employer has told you in confidence
After that, write the approved-account-only list. It covers data that may go only into an approved account, such as your company's tool. Internal documents, work plans and customer data with the identifiers stripped out usually sit here. Customer addresses belong on this list too, because they are personal data, and taking the names off before processing makes them safer.
Rule 3 covers the AI output you always verify before you use it, and how you verify it. In your task map from lesson 7.4, "Map your tasks: delegate, assist or keep", you wrote checks for each "assist" task. Pick the most important ones and fold them into one line. Jia Hui checks every price, date and product name against the source in any AI-drafted text before it goes to a customer.
Two habits from modules 5 and 6 also go under Rule 3. One is the safe word routine: when anyone makes an urgent request for money or codes, you ask for the family safe word and then call back on a number you already know. Put the routine in your policy and keep the word itself out of it. The other is a sharing rule about what you pass on to others. For instance, you don't forward a claim, image or statistic until you have traced it to its original source.
Rule 4 lists the skills you still do by hand. People leave this rule out more often than any other, since it doesn't protect you from an obvious danger. It is also the rule that keeps you able to judge the rest of your AI use. Go back to lesson 7.2, "Skills you lose when you stop practising them". For each skill you keep, write down the skill and the habit that keeps it in use, such as estimating any number yourself before you check it with a tool. Jia Hui writes the first draft of every campaign brief herself and only then asks AI to critique it.
You will only follow rules you can recall when the moment comes. Aim for five to eight of them, each a single sentence you could say aloud. Nobody else reads this, so write in your own words and leave out formal policy wording.
Jia Hui's first draft runs to five lines. Her first line puts work in Copilot and personal tasks in her own account with training off. The second says never to paste passwords, codes, NRIC numbers, bank details or anyone else's private information. The third keeps customer data in Copilot only, with names removed first. The fourth has her check prices, dates and names against the source for anything going to a customer. The fifth gives urgent money requests from family the safe word and a callback.
When she reads it back, she sees that her by-hand skill and her sharing rule are missing, and she plans to add both in lesson 8.4. Your own first draft will have gaps as well. A first draft is meant to be incomplete, and lesson 8.4 is where you finish it.
To start yours, gather what you made in this course, especially your audit table, your task map and your safe word plan. Your first draft will come mostly from those.
Draft your personal policy as five short rules. You should be able to read all five in under a minute.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).