Here’s a number that stopped me cold: 77% of employees using AI say it’s actually making them less productive, not more. Not faster. Slower. All that hype, and most people are burning time instead of buying it back.
I came across this stat in a video from Dan Martell, the founder behind a portfolio of companies he says is closing in on $250 million in enterprise value in just 16 months. He laid out 40 blunt truths about AI he wishes he’d known before wasting hours and thousands in tokens. I’ve pulled together the ones that hit hardest, because the pattern behind them explains that ugly 77% number.
Why so many people fail with AI
The original poster’s core point is simple: almost nobody understands how AI actually works, so they use it wrong. It’s not the tool. It’s us.
Here’s the breakdown of what he means:
- 🧭 More context isn’t better. Dump your whole junk drawer into the prompt and you bury what matters. He calls the failure mode “context rot.” Tight, specific instructions beat a wall of text every time.
- 🎯 One good example beats the perfect prompt. Instead of wordsmithing for 20 minutes, the creator says give it three to five strong examples of what you want. That alone saves hours.
- 🔁 AI doesn’t read minds. Vague in, vague out. Tell it who to be, what you’re doing, and exactly what you want.
And my favorite reframe from him: don’t ask AI for answers, argue with it. Its default job is to make you happy, so it agrees with everything. Tell it to act as a ruthless mentor and poke holes in your thinking. You want bulletproof logic, not a cheerleader.
3 practical ways to put this to work
This is where the video gets useful. Here are three moves you can try today based on what the author shared.
- Talk, don’t type. He says voice input is about three times faster. Just ramble like you’re explaining it to a person, hit enter, and let the AI clean it up. Tools like Whisper Flow make this easy. The point is speed of thought, not typing speed.
- Introduce yourself first. The expert’s line was, “If AI sounds like a stranger, it’s because you never introduced yourself.” Want emails that sound like you? Feed it your old emails, ask it to extract how you write, then have it draft in that voice. Same trick works for your tone, your process, your standards.
- Make AI check its own work. Don’t be the first set of eyes. The creator runs the output through a second AI for feedback, feeds that back to the first, and only then reviews the polished version. You’re grading round three, not round one. He’s clear that the first draft is a rough cut, never the final.
Tips and the pitfalls to dodge
A few truths from the post that double as warnings, because this is where most people quietly lose:
- Playing with AI feels like progress, but it rarely is. Building a cool automation all week feels productive right up until you notice revenue didn’t move. Aim the tool at a real problem, not a shiny one.
- If your AI isn’t moving a number, it’s a hobby. The original poster’s gut check: is revenue up, is time saved? If nothing moves, you’re collecting logins, not results.
- You can’t automate what you’ve never done by hand. If you can’t explain every step, AI can’t run it. His rule: “If you can talk the task, AI can do the task.”
- Simple scales, complexity is procrastination with extra steps. Overengineering the setup is just a fancy way of delaying the thing that gets in front of customers.
- A prompt you can’t repeat isn’t a skill, it’s luck. When you finally get a magic output, ask the AI to write the system prompt behind it. Now you own a machine, not a one-time trick.
- One well-built agent beats ten half-built ones. Most people have a graveyard of half-finished automations. Ship one that runs daily and actually creates value.
The bigger shift
The part that stuck with me most was truth 24: go from “I use AI” to “AI runs.” Using it means you’re still copy-pasting and doing the work. Running it means the AI handles a whole workflow while you review. That’s the whole leap.
The author frames the future around directors, not doers. Be the editor, not the author. Human on the loop, not human in the loop. He even pushes leaders to check tokens before hiring, meaning ask if AI can do the job before you spend on labor. And his line about people is worth repeating: AI won’t replace your team, it frees them. His assistant stopped triaging email and became his chief of staff.
One caveat I’d add. Some of these truths lean hard toward founders and CEOs, and a few overlap. But the underlying message holds for anyone: the barrier was never the tech, it was your willingness to stick with it past the first messy attempt. The people who quit when it didn’t work on try one are the ones who fall behind.
That reframe alone might be why the 77% are struggling. They dabbled, hit friction, and blamed the model.
The full video runs through all 40 truths with the kind of tough-love detail that’s hard to fit here. If even a handful of these landed, it’s worth watching the whole thing and picking the one that stings the most. That’s usually the one you needed to hear.