Don’t Outsource Your Thinking

I still remember the first time I watched a reservoir engineer in a meeting read off a spreadsheet output and treat it like scripture. The numbers on his screen said the well would produce a certain decline curve. Somebody at the table asked him where the assumptions came from. He paused. He pointed at the screen. He said, the model says this. And everybody in the room just nodded and moved on.

I sat there thinking about that meeting for a long time afterward. The model said it. That was the whole argument. The model said it, so we were going to drill it. Nobody in the room had looked behind the model. Nobody had asked which analogs it was pulling from. Nobody had asked what happened last time the software was wrong. They had skipped the thinking because the spreadsheet had saved them from having to do it.

Something very similar is happening right now with AI, and when Alex Economides got to it on the show, he articulated the problem better than anybody I have heard in a while.

What Alex Actually Said

Kevin asked Alex where he sees AI helping in his daily work. Alex said, right up front, that AI is going to end up everywhere, and he is not wrong about that. But he then said something I want every founder and every operator in our industry to sit with. He said the concern he has with AI is that people use the tool to replace their own thinking. To replace their own responsibility to actually analyze something properly. And he said, in his words, he does not think AI is all that great at that.

Then he walked through the reason why, and it is the clearest explanation I have heard. What most people think of as AI today is really a large language model. What an LLM does is take essentially all the information that is sitting out there on the internet and synthesize it down to an answer. The way it picks the answer is probabilistic. It is weighted toward what the most popular answer to your question is. Not the most correct answer. Not the most useful answer for your specific situation. The most popular one.

Stop and think about what that means if you are running a business. You are asking a machine to tell you what the average person on the internet thinks you should do. And then, if you are not careful, you are going to do it.

The Average Answer Is Not the Right Answer for a Leader

Alex's next observation is the one that really landed for me. He said, if you are an entrepreneur or a CEO and you are trying to figure out how to differentiate your company, the answer AI gives you is probably the one that does not differentiate you at all. It is the one that makes you the most in the pack possible.

That is a hard thing to hear if you have been leaning on AI to help you think through strategy. But read it twice. An AI that is trained on the whole internet, asked how to position your company, is going to give you the position the whole internet would pick. Which is, by definition, the most common one. Which is, by definition, the one your competitors are going to land on too. You just spent an afternoon asking a very fast research assistant to generate a positioning statement that every other company in your space is about to arrive at the same week.

Alex reached back to a business school principle to make the point stick. He said if you are the leading company, the smart play is to watch every move the pack makes and match it. You are already ahead. You just need to not lose ground. But if you are not the leader, copying the pack is how you stay in the pack forever. You have to do something different. That is the whole job of a founder. That is the whole job of a company that is trying to become something rather than blend in.

And then he landed it. Unless you are the leading company and you are getting your ideas from AI, everybody else is going to end up following you. Which means, for most operators and founders in this industry, leaning on AI for differentiation is mathematically the opposite of what you actually need.

What This Looks Like in Our Business

See, the other guy is asking ChatGPT how to structure his carbon capture pitch deck. You figured out what is different about your capture technology by spending three weeks in the field with the customer who is actually going to use it. Six months from now, his deck looks like everybody else's deck. Yours has a diagram nobody has seen before and a cost structure nobody else can match. That is not a tool problem. That is a thinking problem.

I have been guilty of the opposite mistake too. I have drafted marketing copy with AI, looked at it, and realized it sounded exactly like what every other B2B company in our space has on their website. Fine words. Clean sentences. Nothing anybody would remember, because nothing in it was actually about us. The tool had done exactly what Alex said it does. It gave me the answer the middle of the internet would give me. That is my fault, not the tool's. But it took me seeing it a few times before I understood why it kept happening.

The discipline is not to stop using the tool. Alex made that clear. He uses it all the time. He uses it for quick marketing material. He uses it to get, in his words, a product out that is good enough. The discipline is to know what you are getting from the tool and what you are still responsible for yourself. Average drafts, fast. That is what it gives you. Not direction. Not positioning. Not judgment. You are still on the hook for those.

The Smartest Five-Year-Old in the World

This is the part of the conversation I want to frame on the wall. Alex told Kevin about a conversation he had with the CEO of Occam's Technologies a few months ago. He told that CEO: treat the AI like a five-year-old. But the smartest five-year-old in the world. Put all the safeguards in that you would with an immature five-year-old, but one who is brilliant.

Think about that. A brilliant five-year-old can do some remarkable things. She can read above her grade level. She can do arithmetic most adults have to look up. She can remember more facts than you can keep up with. But you would not hand her your bank card and send her to the grocery store. You would not let her sit in on the partner meeting and speak for the company. You would not ask her whether you should take the contract or walk away. She is smart. She is not mature. She does not know the weight of what she is saying. She does not know when she is wrong. She does not know what she does not know.

That is the AI at your fingertips right now. It is brilliant and it is immature. It does not know when it is making something up. It does not know when it is about to give you bad advice. It does not know the weight of what it is suggesting. Alex's words, not mine. Do not assume the AI is itself brilliant and mature and knows the importance of what it is recommending, because it does not.

Where It Actually Earns Its Keep

Alex was not anti-AI. That is not the point of what he was saying. He pointed out that at Occam's, AI is helping them act like a team of thirty or forty people when they are actually only a handful. That is the right use. It multiplies a small team. It takes care of the generic work that would otherwise eat their week, so the few people they have can spend their time on the work that actually requires judgment.

He also pointed out two applications he thought were genuinely powerful. He mentioned a robotics company that uses AI to help downhole tools place themselves in a well. He mentioned methane mitigation work in West Texas, where AI is an invaluable tool for finding and addressing leaks at scale. His point, and it is the right one, is that the value of AI depends entirely on the application. Some applications, it is genuinely transformative. Others, it is dead weight pretending to be insight.

The companies I have respected the most over four decades of watching this industry have always had this kind of clarity about their tools. They used Lotus when Lotus was the tool. They used Excel when Excel was. They used reservoir simulators when the simulators were worth using, and they walked away from them when the answer the simulator gave did not match what the well was actually telling them. They did not marry the tool. They married the outcome. AI is one more tool in a long line of them. It is a good one. It is not magic, and it does not think for you.

Final Thought

I have watched engineers in this industry become better engineers because they learned how to use a new tool. I have also watched engineers become worse engineers because they let a new tool do their thinking for them. The tool was not the difference. The engineer was. The ones who got better used the tool to take the drudge work off their plate so they could spend more of their brain on the parts of the problem that actually required them. The ones who got worse used the tool as a substitute for the parts of the problem that actually required them, and they stopped noticing the difference.

AI is going to sort the industry the same way. The operators who get better over the next five years are the ones who figure out how to use it as a multiplier on top of real thinking. The operators who get worse are the ones who start trusting the answer on the screen without asking where it came from. And by the time they notice they stopped thinking, they are going to look around and realize they stopped differentiating too. They are going to look exactly like every other company in their sector. They are going to wonder why their margin is collapsing. And they are going to be very slow to understand that the tool they let do their thinking for them is the same tool every competitor is using, and it gave everybody the same answer.

The tool is useful. Your judgment is what decides whether it makes you better or just louder.

Alex Economides is a fractional CFO at Occam's Technology and an energy transition advisor to New Tech Global. He holds two US patents on carbon market design. On Wisdom at the Wellhead, Kevin Fischer and Philip Richard sit down with him to walk through where AI is genuinely useful in oil and gas, where it quietly makes companies more average, and why he tells every founder to treat the AI like the smartest five-year-old in the world.