Book Brew 176: The Trophy Belongs to You (But So Does The Pile of 💩)

There is a concept in behavioral psychology called self-serving bias.

The TL;DR: we are wired to credit ourselves for wins and blame external factors for losses (not inverse paranoid enough for me….).

You crushed that presentation? Your preparation, your expertise, your instincts.

The client didn't renew? The economy, bad timing, a competitor who lowballed you, the moon was in retrograde. (Okay, maybe not that last one. But you've heard someone say something close.)

This cognitive shortcut is something your brain uses to protect you from your thoughts.

Most marketing advice will tell you to use self-serving bias in your favor →

• make the customer the hero,
• position your product as the tool that enabled their brilliance

That is fine advice for selling things, but for something like AI adoption/enablement, I think we have the problem backwards.

The Inversion We Miss

"When faced with a difficult question, we often answer an easier one instead, usually without noticing the substitution." - Daniel Kahneman

The conventional self-serving bias story tends to go like this: people over-credit themselves for good outcomes and under-credit themselves for bad ones.

With AI, I have seen people do both → sometimes in the same conversation:

• The output is great → "AI is incredible, it did everything." (Credit goes entirely to the tool. You are now apparently useless.)
• The output sucks → "This tool is garbage." (Blame goes entirely to the tool. Your prompt had nothing to do with it.)

Neither of those is an accurate picture of what truly happened.

These tools require collaboration to obtain quality output. You don't get to opt out of having contributed to the result….good, bad, ugly.

If you’ve never developed an accurate read of which variable failed, you will never actually get better at this.

Education from Chronic Migraine

"WYSIATI: What You See Is All There Is. System 1 excels at constructing the best possible story that incorporates the ideas currently activated, but it cannot (cannot) allow for information it does not have." - Daniel Kahneman

I have chronic migraine.

I am not sharing this for sympathy (though I do accept snacks and animal videos). I share it because living with a chronic neurological condition has given me a skill that I did not realize was a skill until I started applying it elsewhere.

Over the many years with this condition, I have had to get very, very good at accurate internal self-assessment.

On a bad day, everything feels harder.

• • My processing slows (yay, brain fog).
  • My tolerance for complexity drops (what is 3+5 again?).
  • My judgment feels intact, which is sneaky because it feels fine, right up until the moment I look at what I produced and realize it is not fine at all. (transient expressive aphasia is fun).
    • It's a textbook example of WYSIATI. My brain constructs the best possible story with the energy it has available, completely blind to its own deficits.

I used to spend a lot of energy blaming the wrong variables. On an attack day I was annoyed at everything, so clearly the task was too hard, or the deadline was unreasonable, or the document I was working from was poorly written.

Over decades, I developed what I can only describe as a calibration practice. Before I sit down to work, I do a quick inventory:

• Is this a neurological symptom, or am I just hungry?
• If it's an attack day, which kind? I color-code them to know when I should or shouldn’t operate heavy machinery,
• What does that mean for what I can produce today, and what should I defer?

Having the right information allows me not to place the blame on the output that was actually compromised by my own neurological state.

AI Skill Transfer

"On any team, in any organization, all responsibility for success and failure rests with the leader. The leader must own everything in his or her world. There is no one else to blame." - Jocko Willink

This same calibration applies to AI collaboration.

When I get a shitty output from an AI tool, I now run through a version of the same internal inventory before I place blame (and honestly, my goal is more about root cause analysis than laying blame…):

• Was it the model? Sometimes, yes. Models have real limitations, and some tasks are genuinely outside their current capability. Using something like Opus 4.8 to draft an email response is like using a jackhammer to hammer that 1 penny nail into trim work.
• Was it my prompt? Also sometimes yes. A vague, context-light (or even empty), energy-low prompt produces vague, context-light output. Garbage in, garbage out has not stopped being true just because the tool is impressive.
• Was it me? Because of our egos, this is the one most of us want to skip. Was I tired when I wrote that prompt? Did I paste in a rough, unedited brief and expect the tool to read my mind? Did I accept the first output because I was exhausted and "good enough" felt like a finish line?

The introspective discipline I learned with dealing with chronic migraine taught me that good enough because I'm hanging on to that rope that is one thread away from breaking (I see you Wile E. Coyote), is a decision I make sometimes, but I am also aware that is what I am doing.

I know not to dress it up as "AI produced a shitty output." Rather, take Jocko’s Extreme Ownership view of it: I produced a shitty prompt expecting it to read my mind, on a low-capacity day, and accepted the first thing that came back.

What Accurate Self-Assessment Actually Looks Like

"Compete against yourself. When you look outside — your rivals, your industry, your luck — there is always something to blame. When you look inside — your process, your effort, your rate of learning — there is always something to improve. Average looks out. Elite looks in." - James Clear

To be clear, I don’t intend for you to go out and be harsh on yourself for the sake of being hard on yourself. But I do recommend that you take a pause when using the AI tools to ensure you are performing an accurate self-assessment and spending time creating a well-crafted prompt before hitting that enter key

When you get a bad AI output, a few questions worth asking are:

• Was the prompt specific enough, or did I leave the AI to guess at my intent?
• Did I give it the context it needed, or did I expect it to fill in what I already know?
• Did I review the output with my actual standards, or did I skim it because I was just so ready to be done?
• If I asked the same thing tomorrow with fresh eyes, would I get a different result?

And when you get a great AI output, it’s also worth asking:

• What did I do that made this work? (Write it down. That is a repeatable system.)
• Was this genuinely a high-quality output, or am I so relieved it exists that I haven't actually evaluated it? (Don’t fall into to this trap….do your own due diligence to ensure you have a true quality output).

The trophy does belong to you (and I’m not talking participation trophies…that is a whole other soap box I won’t climb on to for this post…). But so does the shitty output. Both of them are data…use them wisely

Ponder This

1. Think about the last time an AI output disappointed you. What percentage of that outcome do you honestly attribute to the tool versus the prompt you gave it versus the energy and attention you brought to the review?
2. If you had to describe your "calibration practice" before sitting down to work (the internal check of what you can actually produce today) what would that look like? Do you have one?

Books / Newsletters

• Thinking, Fast and Slow — Daniel Kahneman
• Extreme Ownership - Jocko Willink
• 3-2-1 Newsletter - James Clear