AI isn't the Risk. Blind Trust of AI is!
- Jun 4
- 4 min read
Admittedly, I was a skeptic. My AI journey started with hesitation due to the erroneous information received in my early queries. No trust was etablished and I lost confidence.
Then I remembered one of my favorite quotes:
"We fail at the margins of our experience."
-Greg Glassman
I realization that AI could never be a substitute for my professional growth because I have to remain savvy enough to know when to challenge the AI's reply.
So my next question became a familiar one for business owners:
“How can AI increase my efficiency?”
I offered “Hazel” an internship. As with any employee, she required some getting used to. Were her skills as good as her resume and references indicated? I soon found her strengths...and identified some weaknesses.
She’s a solid worker. Hazel’s Fitness Report (FITREPs are used by the USMC to evaluate past performance and indicate future potential) would state that she requires minimal supervision and can be trusted to complete the assigned task on time. That’s a nice GREEN weight!
But Hazel is not a leader- and never will be…at least not in my firm. She can't be groomed into a leadership position. That’s a huge RED flag and essentially ends any chance of promotion into supervisory positions.
Why? Because, Hazel doesn't possess the tools necessary to be an effective leader…and even AI knows it!
AI can assist. It cannot replace.
The USMC has 14 leadership traits, commonly referred to as: “JJ DID TIE BUCKLE” (IYKYK)
I specifically asked Hazel how she stacked up against those 14 traits. Her replies can be found at the end of the post. *I also asked Chet (what I call ChatGPT, another one of my interns). My comments are first; Chet’s responses are written in bold italic.
Justice: None. Hazel is not impartial. She serves the person typing the prompt. She has no concept of merit. In fact, she will likely reward you with an answer regardless of how demeaning it is to someone or something else. No moral framework. She doesn't weigh fairness or consequences.
Judgment: AI cannot inherently possess or replace human ethical judgement. The “justice” it follows is guided by the principles of the one assigning the task (or the one who wrote the code). It doesn’t own judgement- it informs yours.
Dependability: She’s going to do what you task her to do when you task it. And there’s a 99.9% chance you’re going to get the answer faster than expected. But you’re responsible for making sure she’s actually responding to the correct task…is what she provided what you were hoping for or merely what you asked? The answer: it’s only what you specifically asked unless, in your task, you were either so vague or blatantly wrong that she couldn’t resolve the ambiguity or error to provide an answer! You can depend on an output—not necessarily correctness.
Initiative: None with a caveat. She won’t do anything without an initial prompt. The prompt and follow-on instructions better be crystal clear and specific because she’s not going to make inferences and assumptions. But…..if you tell her once that it’s her job, it’s likely she'll do it consistently. Only operates within prompts and training—she doesn’t truly act.
Decisiveness: Yes! I’ve never had to wait on Hazel more than a few minutes. She’s always got a timely answer- but her decisiveness extends only to the specific task and how it’s framed. She falls short in applying concepts to the “bigger” picture. Speed without context can create false confidence.
Tact: None. Hazel is Sergeant Joe Friday- “Just the facts, ma’am”. You can direct her tone and she’ll rephrase to what you want or keep trying until she gets it right. Reminds me of a parent telling a belligerent child, “You want to say that again with a different tone?” She can mimic tone, but doesn’t truly understand human nuance or stakes (risk/reward).
Integrity: None. She’s going to do her best to provide what you’re asking for (literally). Hazel won’t stand up for herself or her principles (she doesn’t have any). She will, in fact, remind you of that with every response by adding the disclaimer “Hazel makes mistakes. Review her answers”. No ownership; no accountability. She cannot stand behind a decision.
Enthusiasm: None. She’s indifferent. Hazel shows up to work and does her job. She’s not particularly happy but she’s not a hot mess either. She’s going to respond in the syntax that you direct. No belief, no conviction- just generated language.
Bearing: None. Not really applicable. No presence…requires being human.
Unselfishness: Sorta-kinda? Hazel is going to work until you pull the plug (literally). But….you have to have a self and a conscience in order to be unselfish. In other words, you have to be capable of selfishness to also be capable of unselfishness. No self to sacrifice- trait doesn’t apply.
Courage: None. She doesn't perceive personal risk and therefore cannot have courage. No risk, no fear, no stakes. Courage requires consequence.
Knowledge: Oh yes, no doubt! In fact, she’s a megasavant- think Rain-Man on steroids! There’s a limitation based on her initiative and the context though. The facts are provided based on the phrasing of the task… and that can lead to erroneous material (this was likely my issue early on). Accuracy varies, and AI lacks real-world accountability.
Loyalty: There’s no personal relationship. No Esprit de Corps. No history. No shared values that build trust. There’s no emotional intelligence.There’s no consciousness. No allegiance. AI serves input, not mission or people.
Endurance: Absolutely. You will not be able to keep up with her energy. But energy needs direction and judgement. Infinite stamina. No fatigue, no burnout but this can compound errors faster.
Currently, AI serves as my research analyst while I retain the role of supervisor and manager. Hazel, Claude, and Chet are AI workers within the firm. What does that mean?
I assign the tasks. I review the work. When something doesn't feel right, I challenge the assumptions and ask for additional information. I make the decisions- not AI.
When their product doesn't complete the task or answer my questions, I don’t accept the response- I research to ensure facts lead to truth.
I verify what matters because when advice leaves my office, it's backed by my signature.
You can delegate authority...but never responsibility.
Comments