AI documents are making you stupid

Video Transcript

I miss the days when humans wrote documents. I miss reading something at work and thinking, "Wow, this person is really smart and I want to work with them." I miss the small grammatical errors that points to someone writing in a fever dream of inspiration and understanding. I miss the little rabbit holes and side quests that the human brain can't resist exploring. I miss the proof that a human being was actually at the core of an idea being presented in a document.

I miss being genuinely impressed by someone who wrote something I wish I'd written. I read a lot of documents and I don't mind that at all. But I miss reading human thought, human intuition and human ideas presented by a human in human speak. Call me a lite, call me old school, but if I have to read another 10 to 15page document that just regurgitated whatever they asked Claude, I'm going to lose my mind.

So recently, I've received documentation from all sorts of people and businesses and organizations, and I've noticed a very depressing trend. None of these documents were touched by a human. These documents are long. They're bloated and structured exactly like an AI slop LinkedIn post, only about 14 pages longer.

And you know what I'm talking about. You can spot it in your sleep by now, right? And the thing that really got me, the real tell in these documents is my name is misspelled a lot. And look, I know that Nick Haramus isn't the easiest of names in the world to get right, but do me a favor and figure out how to spell it.

It feels like document etiquette 101 to me. In fact, this feels like human etiquette 101. No, I find it so damn insulting and dismissive. So, let me get this right.

You want me to spend time reading a document that you didn't write, that you didn't read, and you didn't check, while also reading my name misspelled 20 different ways. You are a bold person, my friend. It was clear that nobody looked at this document before it was sent to me, not even once. But I still read it because I like to rock it old school and actually read the that people send me.

I didn't want to read it, but I read it to make a point. I pulled out the one or two sentences that actually mattered and I replied to those specifically. And then I ever so calmly wrote another email to the person and told them that if they ever sent me another AI document, I would have to come around and kill them. No, I'm just kidding.

I wouldn't read it. I I wouldn't kill them. I think this is an incredibly important step right now for everybody. I blame the readers as much as the AI writers of slop docks because if you are using AI to read their AI slop and don't call it out, then you are part of the problem.

Unsurprisingly, I received no push back, no dismissive response, nothing at all, in fact, from the writer of that document. And honestly, that's the part that moved me to make this video. That's the part that actually bothered me. It's not that someone sent me a lazy document, but the fact that calling it out didn't even register as a big deal really got to me.

So, here's my hot take. AI didn't ruin document writing, and the person who sent it to me didn't either. It's too easy to expect AI to be better or humans to act better. That's only 2/3 of the problem.

Answer me honestly here. How many documents have you read that actually were good before AI existed? How many documents were genuinely someone's deep and critical thinking on a page versus someone assembling the minimum requirement to look busy and tick a box? Not many, I bet.

But I still miss a bad document being written badly by a human. You can still see the effort they put in or at least identify that they just maybe aren't cut out for writing a document like this. That's fine. AI doesn't create hollow documents.

It just makes producing a hollow document effortless. It removed the last bit of friction that used to force at least some thinking to happen along the way. The real culprit is not artificial intelligence. The real culprit is organizational incentives and engagement.

If you incentivize people to work faster and use AI more, that's exactly what they'll do. People will use AI to work faster, which leads to slop. And the second issue is also a really interesting one. Are people engaged enough in the work to really contribute to it or are they just phoning it in and using AI to get the work done?

There's actual research on this now and it's got a name. Harvard Business Review working with BetterUp and Stanford's Social Media Lab coined the term work slop. Their definition is AI generated content that masquerades as good work but lacks the substance to actually move anything forward. Does this next part sound familiar to you?

I bet it does cuz it's exactly what I'm witnessing. The researchers found that slop doesn't actually save time. It just moves the work downstream. Somebody still has to interpret the slop.

Somebody still has to correct the slop and somebody still has to redo the slop. The sender got their box ticked, but the receiver inherited the actual work. And it gets worse still. Their data showed work slop drops trust in co-workers by 42% and perceived effort drops by 49%.

And it doesn't matter if the person is upfront about using AI. Colleagues recognize the low value and downgrade the person anyway. The researchers are clear that this isn't really an individual failure. It's a management failure.

People are being pressured to produce more, to produce faster, and told to use AI wherever possible with zero training and zero standard for what good actually looks like. This is also why I hold the reader as responsible as the writer of the slop. Readers who call it out create friction in a badly incentivized system. And writers who stop and call it out too do the same.

And that means that both sides of this horror show are complaining and creating friction. So the person sending me that document with my name misspelled a million times in a million different ways probably isn't lazy. They're probably drowning. And the organization handed them a tool that lets them look like they're keeping up without ever asking whether the output was worth anything at all.

Here's some totally insane research that scared the out of me. And listen carefully because I might be describing you. MIT Media Lab ran a study called Your Brain on Chat GPT. Yes, you can see where this is going.

They split people into three groups, writing essays over four months. One group using Chat GPT, one using search engines, and one using just their own brains while measuring all of their actual EEG activity. Obviously, the chat GPT group had the lowest brain engagement of the three. Duh.

But the people in that group also had the lowest sense of ownership over their own essay. Some of them couldn't even accurately quote back what they'd supposedly written just minutes after completing the essay. Let's think about that again for a second. People couldn't accurately recall their own writing.

That's my misspelled name, but at a neurological level. And it's not just the document that wasn't checked before it hit my inbox. It's that increasingly the person sending it never truly owned the document in the first place. The researchers even found something else.

A kind of sameness across the writing these tools produce in the phrasing, in the structure, in the topics chosen, which is exactly what I've been feeling. Everything sounds the same. Everything looks the same. Everything reads the same because increasingly none of it is actually coming from anywhere specific.

So look, here's my true feeling. The value of a document was never the document itself. It was always the humans sitting with the information long enough to interpret it, extrapolate something nobody else would have seen, and turn raw ideas into thoughts and actual insights for other people to consider. and challenge and bash around.

We're losing the skill. Don't let AI steal this step from you. This is where the magic lives. This is where the joy lives.

This is where your challenge lives and where your growth should live. And it's precisely this step that organizations downplay with incorrect incentives. Because when a document only needs to exist so another AI can read it and another AI can take a box and move a Jira ticker to the next column, you don't need a human in the loop at all. You just need something that looks like thinking work slop ticking boxes for other machines.

Where's the fun in that? So where two from here? I don't think AI killed the document. I think engagement in the work we find interesting is tanking because we're not incentivized to work on the things we find interesting or are curious about.

We're incentivized to move tickets and tick boxes and produce output and move faster. I think most of our workplaces probably never really valued the deep thinking that went into an incredible document. They valued the appearance of progress, the work around the work. And AI just made that appearance cheaper and faster to produce.

We can all tell the difference between someone who thought something through and someone who typed six words into an AI chat box to get an answer. So, here's the question to ponder. If your place of work will never notice and will never reward the actual thinking you're capable of and enjoy doing, is that a reason to give up thinking? No.

Of course not. Own your thoughts. Embrace the difficulty of discovering new information and producing new ideas. Lean into your AI tools to improve the way you think, but not think for you.