I launched an experiment with AI. I was fresh off working for a client in the field and wanted to determine if those neural layers could develop actual thought leadership of their own (which is how many are trying to use the technology today). The results still seem to hold true, even as the algorithms continue to advance.
As part of this test, I presented the LLM with three topics:
#1: Please define the advantages of Managed Detection and Response in Cybersecurity.
#2: Please describe what makes Winston Churchill an effective leader.
#3: Please describe what is unique about a Flathead Ford.
My thinking in the choice of topics was to present the LLM first with a topic of interest to the business community; one where there should be plenty of information available to build a dependable premise, and where I had enough personal knowledge to evaluate the result.
The Winston Churchill topic was to establish a sort of baseline. This would be a generic topic that might be the subject of a college essay. It was just a basic writing test in general.
And why the Flathead Ford topic? I wanted something very obscure, where I still had a level of personal knowledge to evaluate LLM performance when dealing with a non-mainstream subject.
My opinion heading into this experiment was that AI would rock the MDR topic, be somewhat insightful on the Churchill topic and possibly bomb the Flathead Ford topic.
And perhaps predictably…I was wrong.
An almost inverse result occurred contrary to my expectations:
The MDR topic was bland, lacking depth and any substantive opinion on the subject.
The Churchill topic did give me a college essay—written at 3 A.M. after the bar closed. It had eloquence, some measure of style, while working very hard to obscure the fact that it also didn’t really say anything.
At this point, when my expectations were at their lowest, the LLM decided to shine. It knocked the Flathead Ford topic out of the park. The response was extremely detailed, presented with authority, and really seemed to know its stuff.
So, what happened?
After giving it some thought, I realized this result distills down to access of both information and insight for the algorithm.
In cybersecurity (and just about any b2b tech industry) useful, detailed opinions reside locked behind some form of lead capture or paywall.
There is so much material available on Winston Churchill, but much of it has been rehashed ad nauseum, requiring unique insight to identify a clean, breakout take.
But the subject of the Flathead Ford…as someone who’s both worked in, and heavily involved with, car culture for many years: I can tell you it’s a topic where there is deep subject matter expertise living on in the hands of a few committed enthusiasts. They share their knowledge freely, and when they do, they tend to go far down the rabbit hole.
Why does this matter? Because AI is all about pattern recognition. When you have freely available, clear signals expressing a strong opinion, AI can carry that strong clear message through its generated communications. But when that message is hidden or diluted through oversaturated noise, AI tries to still pull out an authoritative conclusion anyway. The result can be a hallucination.
What can this tell us about AI and true thought leadership?
The lesson is that AI is more powerful when the input is clean. It’s one reason why aerial drones excel and yet self-driving cars, while making monumental performance gains in recent years, still have some significant failures: In the air, you’re not dealing with the same level of congestion, dirty sensors and unpredictable scenarios.
So, what’s the answer?
Clean input in b2b thought leadership comes down to discovery. And effective discovery doesn’t stop with web research (even when that includes GEO). The precious gem in defining a value proposition as well as clear differentiation is most often found deep in the expertise, unique perspective and wisdom of the principal architects of a company. While the C-suite titles may vary, there can be “power definers” within an organization that truly understand why the company exists, how it improves the performance of other companies and how it can simply make a customer’s life better.
But there is a nuance to mining these gems that goes beyond a prompt. While a conversation with an LLM can point a user in directions they may not have previously considered (it’s certainly happened to me), this is still very sophisticated pattern recognition combined with predictive analytics. The resulting limitation is the primary data points coming from a variably flawed source known as the Internet. It’s not the same as applying intuition and wisdom in a genuinely human conversation to tap into the hidden value of why and how a b2b organization has an offering with the potential to shine in a very crowded marketplace.
Addressing one last, obvious question
At this point, you might be asking yourself: “So, why not just let the CEO write out some thoughts directly and let an LLM polish it?” Because after decades of doing this, I’ve learned that to polish the gem in one perspective requires the contrast of another perspective. It’s how you uncover the definitive value compared to other brands, how you connect that value to prospect’s pain, and often most importantly, how you structure the unified narrative of that value from a jumble of bulleted selling points.
An LLM is very sophisticated at uncovering perspectives, which is one of the reasons it can be such a powerful tool. But does it have a perspective of its own? I suppose what you can call a point of view is the combination of pattern recognition, model training and/or neural layer powered generation. But that makes it a sophisticated facsimile or possibly even a randomized perspective in some cases. Is completely relying on an LLM the same as uncovering a gem or is it just taking a potentially flawed or unpolished perspective, tumbling it around a few different ways and then mass producing it? What I’m trying to do here is provide more definition around the idea of “AI is the tool, not the total solution.” Like other advancements throughout human history, it’s the humanity underpinning the tool that determines the success or failure of the final result.