Co-Workers Think I Suddenly Got Smart, But I'm Just Letting GPT-4 Ghostwrite All My Slack Messages

The Slack Message That Started It All

It began innocently enough. After sending yet another message with a typo ("Let's circle back when you're free to dicuss"), followed by the obligatory asterisk correction ("*discuss"), I'd had enough. As a developer who spends all day writing precise code, why was my workplace communication so sloppy?

That's when I remembered that GPT-4 integration had been added to our Slack workspace. It was intended for summarizing long threads and answering basic questions, but what if I used it for something more... personal?

The Experiment Begins

I created a simple system: instead of typing messages directly into Slack, I'd first ask GPT-4 to craft them for me. My first attempt was for a routine update to my manager:

My draft: "Fixed the login bug. Users were getting locked out because session timeout was set wrong. Should be good now."

My prompt to GPT-4: "Rewrite this message to sound more professional and detailed, but keep it concise:"

GPT-4's version, which I copied and pasted: "I've resolved the authentication issue that was causing user lockouts. Root cause identified: incorrect session timeout configuration. I've implemented and tested the fix, and user login flows are now functioning as expected. Full documentation added to the ticket."

My manager's response was immediate: "Great work! Very thorough fix and excellent documentation. This is exactly the kind of detailed communication I've been asking for."

From Good to "Extraordinary"

Encouraged by this positive feedback, I expanded my experiment. Soon, I was having GPT-4 craft all my Slack messages, from simple lunch plans to complex technical explanations. For each message, I'd provide the basic information and a style guide:

Rewrite this message about delaying the feature launch. 
Make it sound strategic and thoughtful, but not pretentious.
Include a brief explanation and a forward-looking statement.

Within weeks, people were responding differently to me. Comments like "great insight" and "well articulated" started appearing under my messages. A senior director even commented, "This is exactly the kind of strategic thinking we need more of on this team."

Little did they know I was just providing bullet points and letting an AI do the heavy linguistic lifting.

The Office Reputation Transformation

The change in how coworkers perceived me was both subtle and dramatic:

Before GPT-4:

• "He's knowledgeable but struggles to explain things clearly"
• "His messages are brief to the point of confusion"
• "Good developer, but needs to work on communication skills"

After GPT-4:

• "Such a clear communicator, always gets right to the point"
• "His explanations make complex issues easy to understand"
• "Really emerged as a thought leader in recent months"

The Slack Persona I Never Knew I Wanted

As my experiment continued, I noticed GPT-4 had developed a consistent communication style for me—one that was apparently far more impressive than my natural voice:

The Unexpected Promotion

Three months into my AI-assisted communication experiment, I was called into a meeting with my manager and the department head.

"We've been really impressed with your growth recently," my manager began. "Your communication skills have improved dramatically, and you've demonstrated exceptional strategic thinking in how you approach problems."

The department head nodded in agreement. "Your slack messages, in particular, have stood out. The way you articulated the challenges with the authentication service last week was impressive—clear, thorough, and with excellent recommendations for moving forward."

By the end of the meeting, I had been offered a team lead position—a role that would require even more communication, planning documentation, and strategic thinking.

As I walked back to my desk, I wondered if I should confess that my newfound communication prowess was actually GPT-4's doing. But then again, wasn't I the one providing the core information? The AI was just helping me present it better... right?

When the System Fails

Of course, no system is perfect. There have been a few... incidents:

The Vocabulary Mishap

GPT-4 occasionally used words I wouldn't normally use, leading to awkward moments like this exchange with our UX designer:

Her: "Did you just use 'juxtaposition' in a message about button placement?"

Me: "Uh... yes?"

Her: "Since when do you talk like that? Last month you called our dropdown menu 'the clicky down thingy.'"

The Tone Confusion

Sometimes I'd forget to specify the tone, resulting in messages that were jarringly formal for casual situations:

What I asked GPT-4 to rewrite: "Pizza in the break room, come grab some"

What GPT-4 produced and I absentmindedly sent: "I wish to inform all team members that pizza has been procured and is currently available for consumption in the communal break area. Please feel free to partake at your earliest convenience."

Team's response: A thread of 20+ confused/laughing emoji reactions

The Copy-Paste Catastrophe

My worst moment came when I accidentally copied not just GPT-4's response but my entire prompt:

What I actually pasted into Slack: "Rewrite this message to sound more authoritative but still friendly: I think we should reconsider the timeline for the reporting feature since we're running into more bugs than expected."

My cover nearly blown, I quickly followed up with: "Sorry, was testing a new code documentation tool. What I meant to say was: Given the higher-than-anticipated bug count, I believe we should reevaluate our timeline for the reporting feature implementation."

The Philosophical Questions

As my experiment continued, I found myself wrestling with some unexpected questions:

1. Is this actually dishonest? I'm providing all the core information and ideas; GPT-4 is just helping with expression.

2. Am I becoming dependent on AI for basic communication? I've noticed my non-AI-assisted messages becoming even worse by comparison.

3. Does it matter if my "voice" isn't authentic if the information is accurate? Many executives have speechwriters, after all.

4. If people respond better to my AI-enhanced communication, isn't that a net positive for everyone?

The Gradual Revelation

Eventually, I began selectively revealing my secret to close colleagues. The reactions were mixed:

• My work friend found it hilarious and immediately started sending me messages she knew were obviously written by GPT-4
• Another colleague felt slightly deceived: "So all those thoughtful messages weren't really... you?"
• My manager, when I finally confessed, had a surprising take: "I figured you were using something like that. The important part is that you're communicating the right information clearly. The tool is secondary."

Lessons Learned About AI and Communication

This strange experiment taught me several valuable things:

1. Clear communication matters more than we admit - People respond dramatically differently based on how information is presented, not just what information is provided.

2. AI can be a powerful communication bridge - It can help translate our thoughts into clearer expression, particularly valuable for those who struggle with written communication.

3. Consistency is key to credibility - One reason people responded well to my AI-crafted messages was their consistent quality and tone.

4. Authenticity still matters - While the AI helped me communicate better, maintaining some of my natural voice and style prevented the messages from feeling completely foreign to those who know me.

"The best use of AI isn't to replace your voice, but to amplify the best version of it."

The New Balance

Today, I've found a middle ground in my communication approach. I still use GPT-4 for important messages or complex explanations, but I'm more selective. I've also started learning from the AI's techniques—paying attention to structure, clarity, and tone—to improve my non-assisted communication.

In a way, GPT-4 has become like a writing coach, showing me better ways to express my thoughts. My natural messages are improving through this strange form of AI apprenticeship.

And as for that team lead position? I accepted it, with one condition: we integrate AI writing assistance tools officially into our workflow, available to everyone on the team who might benefit from them.

After all, if there's one thing I've learned, it's that sometimes the smartest person in the room is the one who knows when to ask for help—even if that help comes from an AI.