Been another busy week! And at one of the conferences, I was talking with a CHRO and we got into the rabbit hole about why AI writing is so obvious and obviously bad.
So, that's what I want to cover in the newsletter, but before I do this, here are 3 events you don't want to miss with yours truly!
Then on Thursday, I'll be hosting our Toronto People Analytics meetup, Insights You Can Trust. I'm particularly excited about this one as we have not had one in a while and it is an opportunity for Toronto People Analytics Group to get together for a nice evening!
More specifically, why is AI writing so obvious and so bad.
Almost every HR leader, consultant, founder, and executive I speak with seems to have the same reaction:
"We can tell when AI wrote it. It does not meet my quality standard."
Now, whether that's entirely true is probably debatable.
But what isn't debatable is that much of the content being generated today is super generic. The grammar is on point. The structure makes sense. The ideas are clear. Yet after a few paragraphs, something starts to feel off. Very off.
Which creates an interesting paradox.
If AI is trained on billions of examples of human communication, why does it sound less human than the people whose language we used in the training data?
The more I thought about this question over the last few weeks, the more I realized the answer has very little to do with writing itself.
Because, at it's core, AI writing is an analytics problem.
Let's dive in.
The AI Average Problem
Most people think AI learns the way humans learn.
But it doesn't.
Humans learn through experience. AI learns through probability. That sounds like a small distinction, but it's actually huuuge.
Imagine you asked ten thousand HR leaders a simple question:
"What makes a great workplace?"
Some would say compensation, others would say engagement.
Some would point to management, others would talk about purpose, values, or culture.
If you averaged all ten thousand responses together, you would end up with something reasonable. Something balanced. Something broadly acceptable.
And something completely uninspiring and forgettable.
The same thing happens inside modern AI.
When a language model is trained on billions of examples of communication, it isn't learning how exceptional communicators write.
It learns what people commonly say.
It learns what is most likely to come next.
It learns what language looks like "on average."
And that is the core of the problem: average communication is inferior to memorable, specific communication.
Think about the newsletters, LinkedIn posts, and conference speakers you actually remember.
They all share something in common.
They don't sound average.
On the contrary, they have:
A clear perspective
A strong point of view
Memorable stories
Original thinking
Sometimes, even controversial opinions
And we all know AI hates sounding controversial.
But most importantly, they're specific.
And distinctive is the exact opposite of average.
AI, by contrast, is naturally drawn toward the middle of the distribution. It gravitates toward language that is statistically safe because statistically safe language is usually the most probable language.
This is what I call The AI Average Problem.
The Compression Effect
At first, I thought the problem was simply that AI was averaging too many voices together. But the more I looked into it, the more I realized something else was happening.
AI doesn't just average information, it actively compresses it.
Think about what happens when you compress a photograph, the image remains recognizable.
Most of the important details are still there. But some of the richness is well, compressed.
The same thing happens with AI-generated content.
When AI compresses communication, a lot of important things survive:
The facts
The structure
The logic
The recommendations
But a lot of important things disappear:
The stories
The emotion
The personality
The lived experiences
The subtle details that make communication unique
The result is content that is technically correct but very forgettable. Or even further--uninteresting.
And this explains why so many AI-generated articles feel strangely similar.
They're missing texture.
The Voice Density Framework
This is where things start to get interesting.
Because if AI naturally converges toward averages, how do some people consistently produce exceptional content with it?
The answer is better context.
Over the last year, I've watched many people use the exact same AI model and produce completely different results.
Think of the LinkedIn content: let's be honest, we know people use AI. But some content sucks, and other content is amazing.
At first glance, this doesn't make sense.
The technology is identical, the prompt might even be similar.
So what changed?
The answer is what I call Voice Density.
Voice Density is the amount of unique context, perspective, and experience embedded into the interaction with the model.
Most people approach AI with almost no context.
They ask:
"Write me a newsletter about leadership."
"Create a LinkedIn post about engagement."
"Summarize this article."
The model has very little to work with.
No stories.
No experiences.
No audience context.
No examples of previous work.
As a result, it defaults toward the average.
The highest-performing users do the opposite.
They provide:
Previous newsletters -- all of them
Presentation decks
Podcast transcripts and links
Customer conversations
Strategic frameworks
Years of accumulated thinking
And suddenly, the output changes.
And then, they criticize AI and ask it to re-write sections and paragraphs, teaching the machine their voice.
Think about it this way.
Imagine hiring a consultant and giving them thirty seconds of background information.
The recommendations will probably be generic if you will get anything at all.
Now imagine giving them six months of organizational history, performance data, customer feedback, strategic plans, and leadership interviews.
The quality of the recommendations changes dramatically.
The same thing happens with AI.
The output quality is often less about model quality and more about context quality.
Final Thought
The reason AI writing often sounds generic isn't because AI lacks access to human language.
In fact, it has access to more human language than any person who has ever lived and probably will ever live.
The problem is that AI learns from patterns, and patterns tend to pull us toward the average.
But business value is never created by average thinking.
K
Whenever you’re ready, there are 2 ways I can help you:
#1
If you’re still looking to get started in People Analytics, I recommend starting with my affordable course:
Practical People Analytics: Build data-driven HR programs to 10x your professional effectiveness, business impact, and career. This comprehensive course will teach you everything from building an HR dashboard for business results to driving growth through more advanced analytics (i.e., regression). Join your peers today!
#2
If you are looking for support in your human capital programs, such as engagement, retention, and compensation & benefits, and want to take a more data-driven approach, contact me at Tskhay & Associates for consulting services. Or simply reply to this email!
by Konstantin Tskhay August 12th What If AI is Training Us? ↓ Hi Friends, I was reading Sapiens the other day and came across an interesting idea that it was not humans who domesticated wheat, but rather the other way around. Then, of course, I started thinking about AI and how we use it nowdays. I don't know if you are like me, but at times, I get so frustrated with AI responses that I go with the "old fashioned" way of problem solving. Also, known as good old thinking. But then, of course...
by Konstantin Tskhay September 2nd The State of People Analytics 2026: My Biggest Takeaways ↓ Hi, Friends, Over the last 5 years or so, I have served on the HR.com's People Analytics Advisory Board. Most recently, we have built out the State of People Analytics 2026 report based on the feedback from many experts including myself about where we are at and what the future holds. You can download the report here. As I was reading through the final version, one thing stood out to me. We still...
by Konstantin Tskhay July 29th Are you planning your workforce blind? ↓ Hi Friends, I have been noodling on this idea for quite some time now: is HR aware of the macroeconomic data? Sure, we all listen to some form of news about what is going on in the world. But how much of it do we actually understand and use within our organizations to plan how to structure our team for performance. Here is a quick poll: Do you use economic data for workforce planning? Yes No I am not sure The Traditional...
