AI Productivity Mastery

What this course covers and where it starts

This course is about using AI tools in everyday knowledge work. It covers reusable prompts, email, meeting notes and how you keep track of what you know. It starts somewhere less exciting. Before you add any tool, you need a clear record of where your working hours go today.

Chapter 1 builds that record. This lesson explains what to look for. The later lessons in the chapter go deeper into each pattern and walk you through a full audit. The chapter ends with a baseline. That is a measured picture of your week that you can compare against later.

A time audit records what happened, not what you planned

A time audit is a record of how you actually spent your working hours over a set period. It is not a plan or a to-do list. It shows what happened.

The simplest method uses fixed blocks. Every 30 minutes, write one line about what you did in the last 30 minutes. Write what happened, not what you meant to do. "Worked on the report" is too vague to help. "Looked for last quarter's figures, asked a colleague, waited, then started the summary" is useful.

Two rules keep the record honest:

• Write each line at the end of its block. Do not wait until the evening, when the details have blurred.
• Do not sort or judge entries while you collect them. Label them afterwards. If you label as you go, you are likely to change what you do, and the record stops showing a normal day.

A common belief is "I already know where my time goes." Memory is a weak tool for this. A day of meetings and messages blurs together. You remember the big tasks better than the small ones, and the small ones are where time hides. To test the belief, write down your guess before the day starts. Then compare it with the record.

Repeated manual tasks

A repeated manual task is one you do by hand, on a schedule, with the same steps each time. Here are three examples. You copy figures from one system into a slide template every week. You write the same status summary every Monday. You clean up duplicate records after each data import.

Each instance feels small. Twenty-five minutes on a Monday does not seem worth fixing. The cost shows up when you multiply. Twenty-five minutes a week adds up to 1,200 minutes over 48 working weeks, on one task.

To spot these tasks in your record, ask four questions of each one that comes back:

1. Have I done these same steps at least three times?
2. Could I write the steps down for someone else to follow?
3. Does any step need my judgment, or only my attention?
4. How often does it come back: daily, weekly or monthly?

A task that repeats, can be written down and needs little judgment is a strong candidate for automation or AI help later in this course. A task with one judgment step can still be split. You hand off the copying and keep the decision.

Context switching and its cost

Context switching means moving your attention from one task to another. Say a message arrives while you write a report. You read it, reply, and come back. The reply took four minutes, but the switch cost more than four minutes.

Part of your attention stays on the task you left. The researcher Sophie Leroy named this attention residue in a 2009 paper. In her studies, people did worse on a new task when the task they had just left was unfinished.

You can see it in the report example. You come back still thinking about the message. You reread your last paragraph to find your place. The first stretch back goes slower than it would have without the break. An unfinished task makes this worse. A question left open in a meeting, or a thread still waiting for a reply, keeps pulling at you.

"Just focus harder" is a weak fix, because it leaves the open threads in place. Fewer switches and fewer open threads help more. In your record, mark every block where you switched tasks more than twice. Note which switches you chose and which came from outside, such as a message, a call or a request. Those marks show where your focused time breaks up.

Work that only looks productive

Some work feels useful and looks busy, but adds little to what you deliver. It takes forms like these:

• A weekly report whose readers cannot name a decision it changed.
• A recurring meeting that ends without a decision or an action.
• Reformatting data by hand to move it between two systems.
• Replying at once to every message so that you look responsive.

This is sometimes called performative productivity. It is activity that others can see, as opposed to output that moves a goal forward. The test is one question per task: what changed because I did this? If you cannot name a decision, a deliverable or a problem solved, mark the task.

Use this label with care. Some quiet work has real value, such as helping a new colleague or tidying a shared file that others depend on. The question is not whether the work is visible. It is whether the work has an outcome someone needs.

Small gaps between tasks

Look at the edges of your calendar. Between a meeting that ends at 1:30 and one that starts at 1:45, there are 15 minutes. On paper, that is free time. In practice, it fills with a quick check of messages, a half-read document and a glance at the next agenda.

These gaps are hard to see because no calendar labels them. Each one is too short for real work. Together they add up. A day with eight gaps of ten minutes each loses more than an hour this way.

Breaks are different. A planned break, such as a walk or lunch, is part of a sound working day. The gaps worth noting are the ones you did not choose and did not use.

A worked example: one analyst's audit day

Illustrative example: an operations analyst at a 300-person analytics firm runs a one-day audit on a normal Wednesday. Her plan included two hours on a forecasting model. Here is what her labelled record showed for a working day of eight and a half hours.

What the record showed	Time
Copying figures from the CRM into slides for a colleague	40 min
Weekly metrics pull into the same deck, same steps as every week	25 min
Removing duplicate CRM records, a clean-up done after every import	15 min
A status meeting that ended without a decision	30 min
Nine gaps between meetings, not used for planned work	90 min
Work on the forecasting model	50 min, in three pieces

The model got 50 minutes, not two hours. It came in three pieces between interruptions. Two of the pieces started right after a meeting, so their first minutes went to finding her place again.

Now look at what the record lets her do. Two entries are tasks she knows come back on a schedule: the weekly metrics pull, and the duplicate clean-up after each import. Both follow the same steps each time and need no judgment. Those are her first automation candidates. Her focused time breaks up after meetings, which points to a change in her schedule, not a new tool. And she now has numbers to compare against in a month.

Measure a baseline before you judge an AI tool

A baseline is a measurement of how long something takes now, taken before you change it. Without one, any claim that a tool made you faster is a guess.

Here is how that goes wrong. Someone starts using an AI assistant to draft weekly summaries. It feels faster, so they report that it feels like a big gain, with no number to compare against. But they did not time the old process. They also did not count the new work: writing the prompt, checking the draft and fixing its errors. The real saving could be large, small, zero or negative. Without the old number, the claim cannot be checked.

A useful baseline for a task has three parts:

• How long one instance takes, from start to the point it is truly finished.
• How many times it happens in a week.
• What extra work it causes, such as follow-up questions or corrections.

A baseline also helps you plan. The psychologists Daniel Kahneman and Amos Tversky described the planning fallacy in 1979. It is the pattern of estimating your own tasks as quicker than they turn out to be. A timed record of past instances gives you a better estimate than memory.

To build a baseline, time two or three real instances and average them. After you change the task, time it again the same way. Include the time you spend checking the AI's output. Then compare.

There is a second reason to measure first. AI can speed up a task that should not exist. If a report has no reader, drafting it in half the time still wastes that half. The audit tells you which tasks to remove, which to change and which to hand to a tool.

What to do this week

Pick one normal workday. Avoid a day of travel or an all-day planning session. Then:

1. Before the day starts, write your guess: how many hours will go to focused work?
2. Every 30 minutes, write one specific line about what you did.
3. At the end of the day, label each line: repeated manual task, task switch, looks productive, gap, focused work or other.
4. Total each label. Compare the focused-work total with your guess.
5. Circle the repeated task that cost you the most time. Write down its steps.

Keep the record. The next lessons in this chapter use it to build a full baseline.