What AI-Assisted Communication Actually Means
There is a line between using AI to work better and using AI to avoid the work. This module is about where that line is.
You prompt, read the output, edit it, verify the facts, and send it as your considered work. The AI saved you time. The judgment is still yours.
PermittedYou copy AI output and submit it without review. You cannot explain or defend the content because you did not engage with it. This is dishonesty, regardless of quality.
Not permittedThe test: Could you explain every sentence if someone asked? If yes, you have done the work. If no, you have not reviewed it carefully enough.
The tools
ChatGPT, Copilot, and Gemini all do the same core job. The difference is where they live.
The most widely known. Strong for drafting, rewriting, and summarizing. Free tier available and widely used.
Open ChatGPT →Built into Outlook, Word, and Teams. If your workplace uses Microsoft 365, Copilot is already there. High relevance for professional environments.
Open Copilot →Integrated with Gmail and Google Docs. Same drafting capability, relevant if your workplace uses Google Workspace.
Open Gemini →Before the sandbox
Every strong prompt has these four things. Every weak prompt is missing at least one.
Who are you? What is the situation? "I am a new employee at a financial services company" gives the AI something to work with.
Be specific. "Write an email" is too vague. "Draft a follow-up email to a client who has not responded in five days" is workable.
Who is reading it? A message to your manager reads differently from one to a new client or a peer on your team.
Professional but warm? Under 80 words? These constraints turn a generic block of text into something you can actually use.
Ready? Head to the Prompt Sandbox. Three guided tasks put these four ingredients into practice, then you can write your own prompts for a real AI model.
Prompt Sandbox
Start with three guided tasks. Their responses are prewritten, so everyone sees the same example. Then write your own prompts: each one gets an instant check against the four ingredients, and a real draft from a live model if you turn it on.
Scenario: Your first week in a new role. Your supervisor needs you to email the IT department requesting access to the shared OneDrive project folder. You have never met the IT contact.
A prompt has been prepared using all four ingredients. Click to send it.
Review and Edit
The AI gave you a draft. Now the real work begins. Skipping this step is how generic text gets sent and inaccurate information gets passed on.
The frameworkIn order. Each one catches something different.
Is every factual claim correct? AI can confidently state things that are wrong, outdated, or fabricated. Check names, dates, figures, and any specific claims before sending.
Does this match the register for this person and situation? AI defaults to pleasant and slightly formal. That may not fit your relationship with the recipient or the urgency of the message.
Does it sound like you, or like a press release? Replace AI defaults like "I hope this message finds you well" with language a real person would use.
Did the AI include everything that needs to be there? AI sometimes produces technically correct but incomplete drafts because key details were not explicit in the prompt.
Interactive exercise
Click the highlighted phrases in the raw output on the left to see exactly why each one was changed.
Dear Mr. Thompson,
I hope this message finds you well. I am writing to follow up on our previous conversation regarding the upcoming project deliverable. I wanted to circle back and confirm we are on track for the 15th.
Please do not hesitate to reach out at your earliest convenience should you have any questions. I look forward to touching base soon.
Kind regards,
Jordan
Dear Mr. Thompson,
I wanted to confirm the project deliverable is on track for the 15th. Nothing has changed since we last spoke, but a quick confirmation felt useful ahead of the deadline.
If anything comes up, feel free to message me directly. Happy to jump on a call if it would help.
Kind regards,
Jordan
Phrases underlined in red are AI tells.
Reference
Replace these with direct, natural language.
In Practice
Two activities. Diagnose broken prompts, then spot the tells in a real email.
Activity 1Each prompt has a problem. Select what is missing or wrong.
What is the most significant problem with this prompt?
What is the problem here?
This prompt is partially useful but missing one critical ingredient. Which one?
Evaluate this prompt.
Activity 2
Read the email below. Click every phrase that sounds like unedited AI output.
Dear Ms. Reyes,
I hope this message finds you well. I am writing to confirm that the onboarding documents you requested are now ready. I wanted to circle back on our earlier discussion and let you know the files have been uploaded to the shared drive.
You should be able to access them using the link below. Please do not hesitate to reach out to me at your earliest convenience if you encounter any issues.
I look forward to touching base with you in the near future.
Kind regards,
Taylor
Knowledge Check
Five questions on prompting, reviewing, and professional judgment.
What is the key difference between AI-assisted and AI-generated work?
Which of the following is the strongest prompt for drafting a professional email?
During your accuracy review, you notice the AI included "the deadline of the 20th", but you never mentioned a deadline in your prompt. What should you do?
Which phrase is a typical AI tell that suggests a draft has not been properly reviewed?
Your manager asks you to write your own performance self-assessment. You use AI to generate the full document and submit it without editing. This is:
Try it yourself