You will be able to measure how long tasks take now so that any time saved later is real, not a feeling.
Priya, whose task list you followed in lesson 1.2, "Which tasks pay back and which cost you time", is about to start using AI on some of her work. Before she does, she times three of her tasks and notes how well they turn out, doing them the way she always has.
That record is called a baseline. It is a written record of how long your chosen tasks take and how well they turn out, done the usual way, before any change. The rule behind it is to measure how you do a task now before you change how you do it.
People who have used AI at work for about a month usually give a confident answer when asked how much time it saves them, such as "Two hours a week" or "Half a day" (example claims). Ask how they know, and most can't say clearly. They remember the fast moments, like an email draft that appeared in ten seconds. They forget the time spent fixing what the AI produced, like a summary that got a client's name wrong. The gap between the time people feel they save and the time they really save is the reason to measure.
Memory keeps vivid moments and drops routine ones. An instant draft is vivid, but ten minutes of small corrections spread over an afternoon is not. When you look back, the fast moments stand out and the slow fixes blur into the rest of the day.
The second problem is that once you start using AI on a task, you can no longer recall clearly how long it took before. Say you remember your weekly report taking "about two hours". The real figure could have been ninety minutes or two and a half hours, so a claimed saving of forty minutes measured against that memory can't be relied on. A baseline fixes both problems because it is written down before the change, so later enthusiasm or disappointment can't rewrite it.
You don't need to baseline every task. Pick three to five from the ones you sorted in lesson 1.2, choosing those you are most likely to work on during this course, such as email replies, a weekly report or meeting notes. Then do them the usual way for one to two weeks. One week is enough for a daily task. A weekly task needs two weeks at least, so you have more than one data point.
Timing can be simple. Write start and end times in a notebook or spreadsheet, or use a phone timer. Time the task every time it comes up, including the times you would otherwise forget.
Note interruptions too. Say one report takes 150 minutes with the phone ringing four times and another takes 150 minutes of steady work. The minutes match, but they are different data points, and you need to be able to tell them apart later.
For small, frequent tasks, timing each item is fiddly. You can time a batch instead, count the items and work out an average per item.
Time is only half the record. A task done in half the time but sent back twice by your manager hasn't saved much, and you will only notice this if you recorded how often it was sent back before. Useful quality notes include:
rounds of edits a manager or colleague asked for errors caught later by you or someone else, such as a wrong figure or a missing attachment whether the output was used as it was or reworked by someone else optionally, how you felt about the result, in a word or two
With these notes, at the end of the course you can say whether a task got faster without getting worse. You may instead find that a task got faster and worse. That finding is useful as well, because it tells you the saving isn't real.
Keep everything in a spreadsheet with five columns: task, date, minutes taken, quality notes, and AI used (yes/no). Each entry should take about a minute to fill in. If it takes longer, people stop using the log.
During the baseline, the AI used column says "no" on every row. Keep the column anyway. From module 3 you will use AI on the same tasks and log them in the same sheet. The AI column separates the two periods, which makes the comparison in module 9 straightforward.
Don't change how you work during the baseline weeks. If you try a prompt on one of your tasks during that time, mark the row or leave it out of the average, because a baseline mixed with experiments measures neither the old way nor the new one.
Priya chose three tasks: routine email replies, her Monday report, and the notes she writes after the weekly operations meeting. She doesn't time each email. She times a batch each morning, counts the emails she sent and works out an average per email, which is a perfectly good method for small, frequent tasks.
Here are her first week's results, as example figures. The Monday report took 125 minutes with one interruption, and her manager asked for one correction to a figure. Her email batches averaged just over six minutes per reply. The operations meeting notes took 25 minutes and went out with one action item missing, which a colleague pointed out the next day.
The results weren't flattering, but all of them were useful. When Priya later tests an AI assistant on her meeting notes, the bar to beat is one missing action item per meeting as well as twenty-five minutes. Your own baseline sets the bar for your tasks in the same way, on both time and quality.
Your log is the next thing to build. Set it up now with the five columns, so it is ready the next time one of your chosen tasks comes round.
Set up a simple baseline log with columns for task, date, minutes, quality notes and whether AI was used.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).