How to Make AI Actually Useful (Instead of Frustrating)

Written by

in

I tried using AI to draft a project update email last month. After 20 minutes of back-and-forth, I gave up and wrote it myself.

The AI kept giving me something too formal, then too casual, then too long. I spent more time editing its output than if I’d just started from scratch.

That’s when I realized the problem. I wasn’t giving the AI clear enough instructions about what I actually wanted.

After weeks of testing different approaches, I found five patterns that work. Here they are.

1. Specify the Role and Format

Instead of this: “Write an email about the project delay”

Do this: “You’re a project manager. Write 3 bullet points (max 12 words each) explaining the delay, then one 30-word paragraph with next steps.”

Why it works: The AI knows exactly what you need. No guessing, no editing, ready to use.

When to use it: Emails, summaries, any routine communication


2. Show an Example

Instead of this: “Summarize this article”

Do this: “Summarize this article using this format: [Paste an example summary you like]

Now apply the same format to this article: [Paste new article]

Why it works: AI copies examples better than it follows descriptions.

When to use it: Anything where you need consistent output—meeting notes, reports, weekly updates


3. Include the Source Text

Instead of this: “Summarize our Q3 numbers”

Do this: “Here’s the Q3 report: [Paste full text]

Summarize the top 3 metrics in bullet points.”

Why it works: Without source text, AI invents plausible-sounding details. With it, answers stay accurate.

When to use it: Any task where facts matter—reports, meeting notes, data analysis


4. Save Templates for Repeat Tasks

Once you find a prompt that works, save it.

I have templates for:

  • Meeting notes
  • Email drafts to my team
  • Weekly progress summaries
  • Breaking projects into tasks

Now when I need meeting notes, I don’t start from scratch. I use my template, drop in the transcript, done.

How to start: Build one template this week for one task you do regularly. Test it. Save it. Reuse it.


5. Keep It Simple First

Instead of this: A 200-word prompt with detailed step-by-step instructions

Do this: “You’re a business analyst. List the 5 most important insights from this data.”

If the output isn’t good enough, then add detail.

Why it works: Simple often works better. You can always make it more complex if needed.

When to use it: When starting any new type of prompt


Real Example: My Project Status Template

This is what I actually tried:

You're a project manager preparing a weekly status report for leadership.

I'm pasting updates from my team below. Create a status summary with:

1. Overall status: On Track / At Risk / Blocked (pick one and explain why in 15 words)

2. Progress this week: 3 completed items (one line each)

3. Blockers: Top 2 issues preventing progress (include what's needed to unblock)

4. Next week priorities: 3 critical items

Format: Use bullet points. Keep total under 200 words. Be specific, not vague.

[Paste team updates here]

This turns scattered updates from my team into one executive-ready report in 60 seconds. Before this template, I spent 30 minutes every Friday doing it manually.


How to Start

Pick one task you do every week:

  • Email drafts
  • Meeting summaries
  • Task lists
  • Status updates

Create one prompt for it. Test it twice. Adjust. Save.

That’s one template that saves you 10 minutes each time.

Next week, make another.

In a month, you’ll have 4-5 templates that actually work.

The goal isn’t using AI for everything. It’s using it well for repetitive work—so you can focus on things that need your actual thinking.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *