You will use your voice guide and editing passes on real work and measure the time against your baseline.
Run the full method (voice guide, drafting from bullets, one editing pass) on at least five real pieces of writing over one normal working week, and time each one from start to send. Then compare those times with your baseline and judge each draft strictly on whether it sounded like you.
Earlier lessons gave you a voice guide, a way of drafting from bullets and a way of editing that keeps your voice. You tried each of those on only one or two practice pieces. You don't know yet whether they hold up across a normal week of real work, with deadlines and interruptions, or whether they save time once everything is counted.
Pick at least five emails or short documents you genuinely need to produce this week. Where you can, take them from the tasks you targeted in the lesson 1.4 audit, because those already have baseline times to compare against.
Make it a mix. Include at least one routine reply, at least one message that asks for something, and at least one with some difficulty, such as a delay, a refusal or a complaint. If every piece is easy, your results will look better than real life.
If anything you pick was flagged as sensitive in module 2, follow the rules you wrote down in lesson 2.3. Redact sensitive details before drafting where needed.
Every piece goes through the same sequence, so the times can be compared with each other and with your baseline:
Start the timer when you start the task, before you open the assistant. Write your bullets as shown in lesson 3.2. Draft with your voice guide applied. Run one editing pass as in lesson 3.3: the assistant lists suggestions and you pick which to take. Make your own changes. Put back any details you redacted. Read the whole piece once more. Send it and stop the timer.
Start to send is everything between those two points: writing bullets, waiting for the draft, reading it, fixing it, restoring redacted names and the final read. Your baseline times included all of your own thinking and checking, so the new times must include all of it too. If you leave any of that out, the comparison flatters the assistant.
In the same sheet as your lesson 1.3 baseline, add a row for each piece with the task, the date, the minutes and "yes" in the AI column. In the quality notes, record how many edits you made to the draft and what kind: wording, tone, a fact corrected, a promise removed.
Then answer one question in one word: could this draft have gone out under your name without anyone noticing it was AI-assisted? This is the voice yes/no, and you need to answer it strictly, because a generous answer hides the problem. The question is about the draft, not the version you sent. If you had to rewrite the opening because it sounded nothing like you, the answer is no, even if the sent version was fine.
Patterns across the week show you what to fix. A run of noes means the voice guide needs work. A run of yeses with heavy edits means your bullets are too thin.
Jun Wei writes client emails. Her baseline was an average of 14 minutes per client email, timed over two weeks (an example figure, like the others here). In her test week she wrote five emails, which took 9, 11, 8, 15 and 10 minutes from start to send. That comes to 53 minutes in total and an average of 10.6 minutes. At baseline, five emails would have taken 70 minutes, so she saved 3.4 minutes per email, or about a quarter.
The 15-minute email was a reply to an unhappy client, and it took longer than her 14-minute baseline. The draft was too apologetic and she rewrote two paragraphs. She logged it as a no on voice, and decided that for complaints she would write the first draft herself and use the assistant only for an editing pass.
She marked the other four as yeses. Two needed one wording change each. One had a training date that the draft had moved by a day, which she caught on the final read and recorded in her notes.
After the week she added a line to her voice guide: "When something has gone wrong, acknowledge it once and move to the fix. Do not apologise more than once." Her next complaint reply came out much closer to her voice. For Jun Wei, the detail in her log was more useful than the average.
When the week is over you should have at least five new rows, each with total start-to-send minutes, an edit note and a yes or no on voice. Next to them, put the average baseline time for the same kinds of task and the difference between the two.
Then write one or two observations on where the method worked and where it did not, and record any change you made to your voice guide as a result.
If your times are not better yet, that still counts as a result, because early runs include the time you spend learning the method. Module 9 looks at how to read a slow start.
Now pick your five pieces and begin with the first one you need to send.
Log at least five pieces of correspondence with time taken, edits made and a yes or no on whether it sounded like you, and compare with your baseline.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).