“AI slop”. The sloppiness is obvious. And I can't tell my employees: I'm firing you

I asked about him earlier. “Yes, I'm working on it, I'm just thinking about it,” I heard. I said, “Okay, take as much time as you need.”

And then I got this sketch and I thought: “This was written in three minutes by AI.”

I looked at the version history of the document and saw that the text was simply copied and pasted. Then I could see how typical AI traces were edited. There were weird mixed metaphors and perfect three bullet points in every example.

I thought, “This isn't OK.” I'm not going to wait weeks for something to get my hands on it. If I wanted chaff, I would generate it myself.

I use AI and I believe that AI can be a great tool, it can speed up all sorts of good things. We, as managers, have to figure out how to work with people who give us a hard time – and not just tell them, “You're fired.” Because if we did that to everyone, we would have no employees and no agency.

So how do we teach people to use AI effectively? We need to work with people who work with AI to achieve better results – otherwise we will all go crazy.

See also: These three skills can save you in the AI ​​era. Do you have them?

Have you even read it?

I definitely see people relying on AI a lot more today. Sometimes I get a document to review and my comments are: “It looked really good.” I ask, “Did you read this before you gave it to me?” And they say something like, “Well, I had the AI ​​summarize it.”

So I'm the first person to read this – including the author and the reviewer.

As long as a human is responsible for the thinking and it is not outsourced, some of the tedious work can be done by AI. But it still takes someone to read the final version and say, “Yes, that's what I meant.”

I can write on Slack all day, but if I have to send an email, I immediately feel anxious. So I ask the AI ​​to read my email and ask, “Is there anything awkward here? Does this sound OK?” Often he only replies: “It's fine, you can send it.” And that's it – I just needed an AI motivator.

Sometimes the AI ​​suggests, “You can adjust this to make it sound more professional.” But then it doesn't sound like me. The AI ​​will always suggest something like bland autocorrect – what it considers average or normal. You still have to have a moment where you say to yourself, “No, this is my voice. These are my thoughts.”

When I teach people to use AI, I say, “Walk me through your thinking.” This often makes them realize: “Oh, I actually didn't think at all here.” And then they start wondering, “If I were writing this from scratch, what would I want to include?” And I say, “Great. Let's add that to your prompt.”

It's about encouraging people to reclaim their own thoughts, otherwise we all start sounding the same – and that's not good for the company. I don't create blogs or case studies for fun. I do it because we have goals and KPIs, and we want to build relationships with our recipients.

Learning new ways to give feedback

If I just said “no AI”, people would still think, “Sure, sure” and do their thing. It's more about saying, “We accept that you have a tab open with the AI ​​of your choice. Let's just make sure you're using a model that's approved or one that has our brand guidelines built in.”

At some stage of writing and at some stage of editing, a human has to appear.

I spent a good part of my career learning how to give feedback to people who had fragile egos when it came to writing. And now everyone is super confident that they are great writers – because the text was written by AI. This is a new challenge: how to give feedback to people when their content was written by AI or when their planning came from AI.

So the next time I encounter situations like this, I'll say, “I could ask ChatGPT myself. I want to know what you think.”

Perhaps the biggest change in the coming years will be how to manage people who rely on AI.

The above text is a translation from American edition of Business Insider