You’re Training ChatGPT Better Than Your Team
Most founders are better at prompting ChatGPT than training their people. Here’s how to flip that.
When AI Gets Better Inputs Than Your Team
Most founders today are getting more detailed with their prompts to ChatGPT…
…than they are with their team.
They’ll:
→ Add examples
→ Specify tone
→ Highlight edge cases
→ Clarify context
→ Iterate for accuracy
But when a team member asks how to make a decision?
They get a quick Slack message, a voice note, or “Just use your judgment.”
It’s not because the founder doesn’t care.
It’s because they trust their team’s intelligence, but haven’t codified their own decisions (unpacked how they, themselves, make them, step by step).
If you’re not codifying for both AI and humans, you’re probably overtraining one and under-training the other.
Codifying for AI vs. Codifying for Humans: What’s Actually Different?
On the surface, it feels like the same move: make your thinking visible.
But under the hood, the differences matter, especially if you want results you don’t have to fix later.
Let’s make this concrete.
What AI Needs From You:
AI operates best inside strict boundaries. It doesn’t “get” your tone unless you define it. It won’t know your standards unless you spell them out. It only works with what you feed it.
Here’s what that looks like in practice:
Explicit, structured input
Instead of “Make this sound more like me,” say:
“Rewrite this email in a confident, warm, slightly irreverent voice, like I’m talking to a smart friend.”
Then attach a sample it can learn from.
Well-defined constraints
AI thrives on specificity. Give it constraints like:
“Keep this under 250 words.”
“No emojis.”
“Use only examples from this quarter.”
“Don’t mention the brand until the second paragraph.”
Examples and pattern recognition
AI can’t learn taste unless you show it. One thumbs-up or thumbs-down doesn’t cut it.
“Here are 3 messages that worked. Here are 2 that didn’t — and why.”
Success criteria
AI can’t guess what “done” means. You have to tell it.
“This is ready to ship when it sounds like me, has a clear CTA, and doesn’t read like intern copy.”
What Humans Actually Need From You
Unlike AI, humans can read between the lines. But that only works when there are lines to read between. If you don’t make your thinking transparent, they’ll fill in the blanks themselves, with their own narrative, and sometimes get it wrong.
Here’s what training your people really requires:
Your interpretive logic
Teach them how you knew what you knew.
“Here’s how I sensed that last client was about to churn even though they said everything was great.”
Real filters and rationale
“When I interview someone, I’m not just checking qualifications. I watch how fast they respond. I read how they write. I look for curiosity. That’s what tells me if they’re a fit.”
Built-in nuance
Humans aren’t just task executors. They’re interpreters.
That means your instructions need to include:
“If you’re not sure, ask.”
“Here’s when to loop me in.”
“This is a time to follow the process. This other one is a time to deviate.”
So while both AI needs structure and clarity, humans need something judgment coaching.
The Cooking Analogy You Didn’t Know You Needed
Let’s bring this down to earth.
Say you’re teaching someone to make your version of the shakshuka (Google it if you’ve never had it, it’s so good!)
For AI:
You need:
Exact ingredient list
Precise quantities
Step-by-step instructions
Photos of each stage
Notes like: “Avoid extra salt, the soy sauce already covers that”
That’s how AI learns: Inputs → Constraints → Feedback → Repetition.
For humans:
You can say:
“It should smell a little nutty before you pour the broth.”
“If the edges are curling, it’s ready to come off the heat.”
“If it looks too thick, splash in some wine, not water.”
“You’ll know it’s close when it starts to pull away from the sides of the pan.”
They need:
Sensory cues
Timing intuition
Flexibility
The why behind each move
That’s how humans learn: Sensation → Intuition → Interpretation → Story.
And here’s the twist:
When you codify for AI well, you create sharper inputs for your team.
Because your team needs everything the AI does, PLUS the interpretation layer only humans can apply.
Where Codification Comes In
Here’s where the magic happens.
Codification is the connective tissue between what you know and what others can reliably replicate, with or without you.
Most founders think they’ve done it if they’ve made an SOP or recorded a Loom.
But SOPs tell you what to do. Codification tells you how to think.
Here’s what that might look like:
Instead of saying:
“Review the onboarding doc and tell me if anything’s missing.”
Codification might include:
A Decision tree:
“If it’s a referral, skip step 4. If not, follow the full sequence.”
Filters:
“If a client hasn’t replied in 3 days, they don’t need a new doc, they need a quick nudge and here’s how to do it...”
Red flags:
“If a client asks about deliverables too early, they’re unclear on the process. Flag it before it escalates.”
Exceptions:
“We don’t discount - unless it’s a partner with 3+ referrals. Then we comp onboarding.”
These are things that live in your gut, until you make them teachable.
Once you do, you give your team and your AI what they each need to operate at your level.
The Codification Checklist: Humans vs. AI
Want to make your thinking transferable today?
This is the mental checklist I use every time I brief a tool or delegate a decision. It also how I mentor my homeschooled kiddos and clients.
✅ If You’re Briefing AI
Did I clearly define the goal?
Did I specify tone, format, and constraints?
Did I include at least one example?
Did I highlight edge cases or what not to do?
Did I define success criteria?
✅ If You’re Training a Human, add these
Did I explain the context behind the task?
Did I share my filters or gut reads?
Did I give examples of good vs off-target execution?
Did I surface common exceptions or edge cases?
Did I create space for questions or reinterpretation?
Pro Tip: Codify once, adapt twice.
Once you’ve made your thinking clear for one audience (AI or human), you’re 80% of the way to the other.
All you need to adjust is:
Structure and Constraints → for AI
Narrative and Nuance → added for humans
1% Shift This Week
Have one of your team members reflect back a recent decision you made:
“Hey, I’d love to hear - based on what you’ve seen, how do you think I made the call on X? What patterns or priorities do you think I was using?”
Then compare it to how you actually made the call.
You’ll learn more in 2 minutes than in 20 Slack messages.
This is something I do constantly with my children, and in every mentorship role I hold.
It doesn’t just show you what people are absorbing.
It shows you what you haven’t made visible yet.
xo Khaïry