You ask AI to write a short email. Seconds later, the screen fills with polished paragraphs.

Then you start editing. The opening sounds stiff. The message is too long. One sentence promises something you never agreed to. By the time it sounds like you, you wonder whether writing it yourself would have been faster.

That is the question behind the excitement about AI: does it help you finish the work, or just get you to a first draft faster?

My view is that the finished result matters more than the speed of the response. A useful draft can save effort. A draft that needs a rescue mission can become another task on your list.

An hourglass beside marked-up draft pages in warm cream tones, with John Bernier branding.

Count the work after the answer

When deciding whether AI is helping, count the whole process:

  • Preparing the notes, files, or context it needs.
  • Explaining the task and refining the instructions.
  • Checking the answer for mistakes and missing details.
  • Editing it into something you can actually use.

None of that automatically makes AI a bad choice. It simply belongs in the calculation.

Imagine a task takes 15 minutes manually. AI produces a draft in one minute, but preparing the instructions takes four minutes and reviewing the result takes twelve. You have spent 17 minutes, even though the answer appeared almost instantly.

Those are illustrative numbers, not results from an experiment. The point is to measure the path to completion, including the work that happens after generation.

Three everyday tasks, three different decisions

1. The email you already know how to write

If you need to say, “Thanks for the update. Tuesday works for me,” opening an AI tool may add unnecessary steps.

For a more difficult email, the balance can change. You might need to explain a delay clearly, organize several questions, or make a firm response sound respectful. AI can offer a starting point when the wording is the obstacle.

Give it the actual message, the purpose of your reply, and the facts it must preserve. Then check the draft for unintended promises before sending it.

2. The meeting summary that sounds complete

A tidy summary is appealing. But could someone who missed the meeting use it to understand what was decided?

Separate decisions, proposed ideas, and assigned actions. An unresolved suggestion should stay unresolved. A deadline should appear only when the notes support it.

AI is more useful here when you supply clear notes and ask it to flag gaps. If you have to reconstruct the meeting to verify every sentence, the summary has created more work than its polished appearance suggests.

3. The marketing copy that could belong to anyone

“Write a post about my business” leaves a lot for AI to guess.

Before asking, identify the customer, the problem, one concrete example, and the action you want the reader to take. Those details give the draft something meaningful to say.

You may still need to remove generic claims and restore your own voice. Judge the copy by whether it communicates your business clearly, not by how professional it sounds at first glance.

Try one familiar task

You do not need a complicated productivity study. Choose a task you already understand well enough to check.

  1. Note roughly how long it normally takes and what a finished result looks like.
  2. Use AI, counting preparation, prompting, review, and editing time.
  3. Check the quality. Is it accurate, complete, and suitable for its audience?
  4. Repeat a few times before deciding whether the workflow is worth keeping.

Record initial setup separately. Instructions you can reuse may make later attempts easier, but that first investment still counts.

The question to ask is simple:

Did this help me finish sooner, with a result I trust?

Sometimes extra time is worthwhile because the result is clearer or more thoughtful. That is a quality benefit. Recognize it without calling it a time saving.

Make the task smaller before making the prompt longer

If AI keeps missing the mark, narrow the assignment. Ask it to organize your notes before writing the full article. Ask for three headline options before requesting an entire campaign.

Keep the pieces that help and adjust the ones that require too much correction. The goal is a useful routine you can repeat.

AI does not have to handle every task to earn a place in your workday. It needs to help with the right ones.

Which task does AI save you the most time on—and which creates more work?

I’d welcome your thoughts on LinkedIn. Visit my LinkedIn profile to connect and share your experience.