Statistically Unlikely

AI is in the zeitgeist. And everywhere I go people are talking about the challenges of being human in an AI world. I won't predict what will happen next year (and beyond). I'll leave that to the clickbait on social media and news sites, and real experts if there are any left. What I can talk about is what it means for you NOW.

I'm not going to cover all AI. It's a much larger bucket than most people realize. There is plenty of AI you've been using for years, including the auto-suggest when you type an email, the search results you get from Google, and that creepily relevant ad you got served on your favorite social media site.

When ChatGPT launched in late 2022 and surged to over 100 million users within 2 months, it made people aware of a particular 'flavor' of AI called GPT (Generative Pre-trained Transformer) which we now commonly call GenAI. In layman's terms it's a model that is driven by statistics. These models (LLMs) were created by sucking up all the content available in the world. Basically the largest IP theft in history. That's a topic for another newsletter. And some lawsuits.

In a very oversimplified view, ChatGPT (and other GenAI like Claude, Gemini, Grok et al) doesn't write whole papers, paragraphs, or even sentences. It basically writes one word at a time, using statistics to determine which word is most likely to come next. It does this based on ALL the context it has for ALL the content it has consumed.

We now have the ability to generate content at scale. And it can elevate the ability of below average writers to be at least average writers. That's mass mediocrity at scale. This creates a challenge for people whose business is to read things and for people whose business it is to write things. Emails, ad copy, cover letters, and even research papers. It's a lot easier now to get decent output using GenAI.

That's powerful. And it's not going away. So what do you do?

AI + Humans

GenAI is amazing. I use it every day. It helps me think, draft, rewrite, research, and even generate ideas for this newsletter. But there’s something important to remember:

That’s not a critique. It’s just the math. GenAI doesn’t think—it predicts. It generates what is statistically most likely to come next. The result is fluent, often insightful, and incredibly useful… but also, by definition, statistically likely.

Statistically likely is another way of saying forgettable.

What you put into AI models helps influence the output. We call these prompts. Each prompt you enter adds more context to how GenAI will respond. This context doesn't exist in the model...yet. So each successive prompt makes your particular output from GenAI less statistically likely.

The output is less likely because of the unique content and context you bring to the 'conversation.' The human element you add through prompts and additional content gives you the opportunity to create something unique and differentiated.

What Makes You Trustworthy?

Let’s flip the frame. When you hire someone—an advisor, a new employee, a vendor—what makes you trust them?

It’s probably not because they echo everything you've already heard. You trust them because they bring a fresh perspective, ask better questions, or help you see something in a new way. They stand out—not just by being smart, but by being different in a way that’s valuable.

That’s where GenAI hits a wall. It’s fluent, but it isn’t insightful. And the more we rely on it to generate our messaging, positioning, or thought leadership, the more we risk sounding like everyone else.

I've written about some of this before. So let's pull together some of the pieces:

From People Like Bad Pizza:

In Real Thought Leadership Helps People Decide, I wrote:

And yet, we often confuse fluency with insight. That’s the illusion of expertise—whether it’s from AI or a human. Just because someone (or something) can talk confidently doesn’t mean they can help you think better.

In GPT Doesn’t Know What I Know, I said:

GenAI makes this worse. It’s great at sounding like it knows what it’s talking about. But just like that consultant who’s never really had to implement their own strategy, it can’t actually deliver the nuance that comes from lived experience.

Your Advantage Is Human, Not Heuristic

GenAI can remix existing knowledge. But it can’t replicate tacit knowledge—the kind of insight that comes from your pattern recognition, relationships, intuition, and timing.

That’s what I leaned on when I mentored new hires at Freshworks. As I wrote in GPT Doesn’t Know What I Know:

The connections I made, the context I held, the things I knew to ask—none of that was written down. None of that could be captured in a slide deck or scraped into a training doc. That’s what made me valuable. That’s what made me statistically unlikely.

And that’s what makes you valuable too.

If you want to future-proof your work, your company, your career—don’t just be knowledgeable. Be recognizable. Be the one who says what others won’t. Who makes connections others can’t. Who asks the question that unlocks the insight.

That’s how you create value GenAI can’t.

How to Be Statistically Unlikely

Here’s the thing about the average: it’s big, safe, and crowded.

You don’t win there.

You win by being different in a way that’s useful. 

That means:

✅Saying something new—not just summarizing what’s already been said
✅ Asking a good question—one that doesn't show off what you know but aids additional exploration
✅ Bringing unexpected clarity to a messy conversation
✅ Taking a stand others are afraid to take
✅ Translating between perspectives that don’t normally connect
✅ Helping others see differently

This isn’t about being contrarian. It’s about being memorable, trustworthy, and clear in a world of noise.

In Who Are Your Sensemakers?, I described the role of the sensemaker:

Statistically unlikely is the natural mode of the sensemaker.

The Bottom Line

GenAI is here to stay. But if all you’re doing is polishing the predictable, you’re competing with the most average machine ever built.

You don’t beat that by being a better version of average. 

You beat it by being unmistakably human.

Berkson's Bits

🧠 Fluency isn’t the same as understanding. For people or machines. 

🎭 Want to be future-proof? Learn to improvise. 

🔀 If GenAI is remixing the past, your job is to remix the present.

What I'm Listening to

I am a big fan of jazz and I got a chance to see a true legend last week. Heading back to my alma mater, Queens College, I saw Chuck Valdes and the latest incarnation of his band, Irakere, at Golden Center. I thought I had never heard him play but realized I had heard him with another Cuban jazz legend, Arturo Sandoval. Here they are playing together back in 1988.

Let me know what makes YOU statistically unlikely. Or tell me where GenAI has surprised you (for better or worse). I’d love to hear your take.

Looking forward to continuing the conversation...

Alan