You will be able to estimate whether AI help on a task will save time once checking and fixing are counted.
When you look at your own work and wonder where AI could help, the obvious place to start is the task that takes the longest. But the biggest task on a list is not always the one where AI gives you time back. A rough estimate takes about five minutes per task. It tells you which tasks are worth trying first, and it can stop you spending a month on a task that was never going to pay.
When people say AI saved them time, they usually mean the draft appeared quickly. That is only part of the picture, because every AI-assisted task adds three costs that the old way of working did not have.
Prompting time is writing the request, pasting in the material and answering any questions the assistant asks back. Checking is reading the output closely enough that you would put your name on it. Fixing is correcting what was wrong, adding what was missing and getting it into the format your manager expects.
Once you have those three numbers, the sum is simple. Time saved is the time the task used to take minus the time spent prompting, checking and fixing. If the result is positive, AI helps on that task. If it is close to zero or negative, AI does not help, however impressive the first draft looked.
Priya runs operations for a mid-sized logistics firm in Tai Seng. After lesson 1.1, they wrote down their ten biggest tasks. All of Priya's figures here are example figures. The weekly report sat at the top at two hours, followed by email, then meeting notes, then a monthly reconciliation of supplier invoices against delivery records. Priya's first instinct was to point an assistant at the biggest number on the list, which was the report.
By hand, the report takes 120 minutes. With AI, Priya estimated 10 minutes prompting, 40 minutes checking figures against the source spreadsheets and 20 minutes fixing, which comes to 70 minutes. The saving is 120 minus 70, or 50 minutes a week. That is worth testing, though it is less than the headline number suggested.
Routine email replies looked smaller. Each one takes about 6 minutes by hand. With AI, Priya estimated 1 minute to prompt, 1 minute to read and 1 minute to fix. That makes 3 minutes, so each email saves 3 minutes. Priya sends around 40 of these replies a week, and 3 times 40 is 120 minutes a week saved, which is more than the report.
Frequency is where the savings come from. A task you do many times a week multiplies every minute saved. A task you do twice a year barely changes your week, however much faster it gets. So when you go through your list, ask how many times a week or month each task comes up as well as how long it takes.
Frequency also decides whether the setup cost pays back. Setup means writing and testing a good prompt. Priya considered spending 30 minutes writing and testing an email prompt, including a short note on their tone and the points they always cover. At 3 minutes saved per email, that pays back after 10 emails, which is two or three days of normal work. The same 30 minutes spent on a prompt for an annual budget memo might never pay back, because by next year the format, the numbers and possibly the tool will have changed.
Watch for one trap. When checking AI output means redoing the whole task, AI help rarely pays back. Detailed financial work often falls into this trap, especially when an error is costly and the only way to catch one is to redo the work.
Priya most wanted to hand over the reconciliation, because they find it tedious. Then they realised that trusting an AI reconciliation would mean comparing every line against the invoices and delivery records, and that comparison is the reconciliation itself. By hand it takes 90 minutes a month. With AI, the estimate was 10 minutes prompting and 85 minutes checking, 95 minutes in all. That is a loss of 5 minutes before any fixing.
You can still get help with a task like this. A formula that matches records automatically, covered in module 5, might help a great deal. Asking an assistant to do the matching and then checking its work line by line usually does not.
A quick test helps you spot this trap. For each task, ask yourself "how would I know if this output were wrong?" If the answer is "I would read it in a minute and see", the task is a good candidate. If the answer is "I would have to do the whole thing again", it usually is not.
Research points the same way. Lesson 1.1 described a 2023 field experiment run by a research team that included Harvard Business School academics, working with consultants at Boston Consulting Group. On tasks that suited the model, people using it did better and worked faster. On a task chosen to sit outside what the model did well, people using it were more likely to get the answer wrong than people working without it.
Priya ended up with a short list they could defend: email replies first, the weekly report second, and the reconciliation moved to module 5 so it can be handled with a formula instead of an assistant.
To build your own list, start with the tasks from lesson 1.1. For each one, note how long it takes by hand, then estimate the prompting, checking and fixing time with AI. Subtract those three from the original time to get the time saved, and multiply that by how often the task comes up each week or month. If a task needs a prompt written and tested first, compare that setup time with the saving per use to see when it pays back. Then apply the quick test. Rank what is left into a short list, and move any task where checking means redoing it to another approach, such as a formula.
Treat every estimate as a guess until you have tested it. Each one is a hypothesis that the baseline in lesson 1.3 and the exercises later in the course will confirm or overturn. A task that looks ideal can turn out to need heavy fixing, and a task that looks risky can turn out to be easy to check.
Your next step is to run the same sum on your own tasks, starting with five of the ten you wrote down in lesson 1.1.
For five tasks on your list, estimate current time, likely prompting time and likely checking time, and mark which ones should save time.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).