I Let an AI Roast My Work Habits for a Week

I built a tool to track employees. So naturally, the first thing I did was turn it on myself.

For seven days, I ran ScreenJournal with one specific instruction to the Agentic Chatbot: "Be ruthless. Don't be nice. Tell me exactly where my time is going and where I'm wasting it."

The results were humbling.

I expected the AI to say something like, "You're very productive, Adam." Instead, Gemini took me to the woodshed. Here are the excerpts that stung the most.

The "CSS Yak Shave" Report

ScreenJournal Log:

10:34 AM - 11:19 AM: "Adjusting login button color and shadow properties in Figma."

Gemini's Roast:

"You spent 45 minutes tweaking the login button. The color changed from #3B82F6 to #4F92F7. The shadow went from 0 4px 8px to 0 4px 6px. No user impact. No A/B test. No measurement. This is called 'Yak Shaving'—you prioritized feel-good work over actual shipping. The button looked fine at 10:34. It still looks fine at 11:19. You wasted 45 minutes because you have zero design discipline."

Ouch.

The "Tax Procrastination" Receipt

ScreenJournal Log:

3:00 PM - 3:15 PM: "Opened: Tax Optimization Strategy (PDF)." 3:15 PM - 4:02 PM: "Tab visible, but no scroll activity detected."

Gemini's Roast:

"You opened the 'Tax Optimization' PDF 12 times in the past week. Total time: 1 hour 47 minutes. Total pages scrolled: 0. You didn't read it once. You opened it, looked at the title, and closed it. This is anxiety-driven procrastination disguised as work. Just hire an accountant for £2,000 and recover 2 hours of your life. You're losing money trying to save money."

It was right. I've been planning to hire someone for 6 weeks.

The "Slack-Driven Stupor"

ScreenJournal aggregated data across the week and generated this timeline analysis:

Gemini's Report:

"You check Slack every 6 minutes. On average. Even when you're not actively coding, Slack is the second window you switch to after taking a mental break. Your average focus window before a Slack check is 8 minutes. For context: the human brain enters 'Deep Work' after approximately 15 minutes of uninterrupted focus. You are systematically preventing yourself from entering deep work by checking Slack compulsively. You are not being productive. You are being reactive."

The AI then quantified it:

• Average context switch: 47 seconds
• Total context switches (5-day work week): 1,247
• Time spent just switching (not including time reading Slack): 9.8 hours
• That's more than a full workday spent on the act of switching.

ScreenJournal's interface overlaid my Slack switch times on a graph. The pattern was undeniable. I wasn't "multitasking." I was "context-switching-tasking."

The Moment It Got Real

But then the AI said something that made me actually change my behavior:

"You are shipping 12% less code this week than your moving average. Your team ships 8% more features than you do, on average, despite your seniority. If you were an employee reviewing this timeline, you would not get a raise. You would get a performance improvement plan. This is not about 'balance' or 'rest days.' You're just broken."

I wasn't broken. But I was inefficient in a way I couldn't see because I was in it.

I disabled Slack notifications that day.

What Happened Next

The insight from ScreenJournal wasn't "Adam is lazy." It was the opposite. The AI showed me I was overworking on low-value tasks and underworking on high-value ones.

After I turned off Slack notifications:

• Context switches dropped to 1 every 14 minutes
• First deep-work window extended from 8 minutes to 42 minutes
• Code shipped increased by 19%
• The PDF still sits there, and I hired a £2k/year tax person. Best decision ever.

Why This Isn't "Spyware"

I built ScreenJournal for companies, not surveillance. But here's what surprised me: The scariest part wasn't having a system to watch me. It was having the system tell me the truth about myself.

Traditional employee monitoring tools show your manager a timeline. They can see you were "in Slack for 2 hours" and make assumptions.

ScreenJournal's AI doesn't show raw footage. It shows analysis. It roasted me, not because it was malicious, but because it was honest.

The Problem with Raw Transparency

If I gave my manager access to a video recording of my 8 hours, they'd see:

• Me staring at the login button
• Me clicking on the PDF
• Me switching to Slack (a lot)

They'd think I was lazy or unfocused.

With ScreenJournal, I got:

• The exact tax procrastination trap I've been in for weeks
• The Slack-driven context-switch problem that was silently killing my productivity
• A data-backed insight that I needed to change my environment, not my work ethic

That's the difference between "monitoring" and "coaching."

What This Means for Your Team

Here's the proposition I'd make to any CTO or founder using ScreenJournal:

You could use this to micromanage. You could look at your developer's Slack frequency and get angry. Or you could do what I did and let it tell your team the truth about their inefficiencies.

"Hey team, we ran this on ourselves. Here's what we found. The system isn't judging you. It's coaching you. Use it."

The best part? Your team will probably thank you. Because nobody wants to waste 45 minutes on a button. Nobody wants to lose 2 hours a week to Slack. Nobody wants to be less effective than they could be.

They just can't see it without the data.

The catch? It works both ways. If the AI tells you you're shipping 50% more than expected, it'll say that too.

I was roasted for a week. I changed my behavior. I shipped 19% more code.

Imagine if your entire team did the same.

---

Ready to get roasted?

Join the ScreenJournal beta and be prepared to hear some truths you didn't expect to hear.

But stick with it. On the other side is better work.