A Thesis on Discovery

Lazy AI People Have Codes

Your customers aren't personas. They're characters with codes — forged by experience, invisible from the outside, and governing every decision they make about your product.

Jamil Jadallah — February 2026
Scroll
01

I own three pairs of the same Air Jordans. I have the same pants in five different colors. If you asked ChatGPT to generate a persona of me as a consumer, it would never — could never — predict that. It would give you something rational, evenly distributed, demographically plausible. It would describe a person who doesn't exist.

Real people are specific. They're contradictory. They're irrational in ways that make perfect sense once you actually talk to them.

But right now, across thousands of product teams, someone is typing "create a persona for a B2B procurement manager" into ChatGPT, getting back a polished, entirely fabricated profile, giving it a name, putting it on the wall, and building a roadmap around it. Without ever talking to a customer.

This isn't a technology problem. It's a curiosity problem. And the technology just made it frictionless to stop being curious.

People Have Codes The best fictional characters are memorable because they operate by an internal logic system that governs every decision they make. Your customers do too. And no LLM can crack them for you.
Code #001

Omar Little

"A man got to have a code."

Omar doesn't rob civilians — ever. That's not a demographic attribute. It's not a psychographic segment. It's an internal logic system that governs every decision he makes and every interaction he has.

If you were building a product for Omar's world and you didn't understand his code, you'd build for the wrong person entirely.

Code #002

Hickman's Worlds

Codes don't just define characters. They build civilizations.

When Jonathan Hickman established Krakoa, he didn't just create a nation — he established its operating system. Three laws. Three codes that an entire civilization would run on:

I. Make more mutants. II. Murder no man. III. Respect this sacred land.

And then immediately, every character's individual code started grinding against those collective laws. Magneto's code evolved from separatist to statesman. Xavier's code — always about integration — somehow led him to build the very separatist nation he once opposed. And Apocalypse — the character whose code is literally "survival of the fittest" — became a diplomat. Hickman showed that codes aren't static. They're living systems that evolve under pressure.

Code #002b

In East of West, Death's code is built around a love that literally defies the apocalypse. Wolf's code is survival through loyalty to a system that betrayed him. Hickman uses these personal codes to drive an entire geopolitical narrative — proving that codes aren't just personal attributes, they're structural forces that shape systems.

The tension between Killmonger and T'Challa in Black Panther tells the same story. Same goal. Codes forged by completely different lived experiences. Opposite strategies from the same stated intention. The intergalactic run takes it further — the code evolves as the context changes, but it never becomes generic. It stays specific.

Code #003

Complicated Codes

Codes aren't simple binaries. They're living contradictions.
Magneto
Code forged in the Holocaust. "Never again" applied to mutants. Except on Krakoa, where the separatist became a builder.
Xavier
Code of integration and coexistence. Until he built a nation that proved separation was the answer.
Apocalypse
"Survival of the fittest" sounds simple. But "fitness" keeps getting redefined. The destroyer became a diplomat.
Ned Stark
Honor as code. It got him killed. But his code outlived him — it shaped every decision his children made.

These aren't personas. They're operating systems — built from specific experiences, specific wounds, specific beliefs about how the world works. They evolve. They contradict themselves. They produce unexpected behavior that makes perfect sense once you understand the code beneath it.

Your customers are the same.

Your customers aren't personas. They're characters with codes — and the job of discovery is to crack them.

This is what the persona conversation keeps missing — in both B2B and B2C. The mechanical frameworks of persona creation — demographics, psychographics, jobs-to-be-done — are scaffolding, not the building. The building is the code. The internal logic. The thing that only reveals itself when you sit with someone long enough for them to trust you with the story behind the behavior.

What Sitting Next to People Teaches You

Early in my career, I worked in financial services. I sat next to the people who ran the processes — the operations teams, the back-office staff, the ones who actually moved money through systems every day. And I learned something that reshaped how I think about product work: the people running the processes were not the bankers making the deals.

Ask ChatGPT to describe a "user at a financial services firm" and you'll get a persona that looks like a banker. You won't get the person who's been at the company for fourteen years, knows every edge case in the settlement workflow, and has opinions about which fields in the legacy system are lying to you.

Real Codes — Discovered, Not Generated
The Operations Manager
Persona says: "Resistant to change"
She sat next to me and showed me the macros she'd built in Excel — not because she loved VBA, but because the system didn't match how the work actually happened. She won't adopt any tool that doesn't let her export to CSV. She got burned in 2019 when a vendor locked her data and she spent three weeks manually rebuilding her reports.
THAT'S A CODE.
The Sysadmin
Persona says: "Prefers documentation"
He documents everything in a personal wiki because he's been blamed too many times for outages caused by undocumented changes someone else made. He built entire defensive architectures around the assumption that nobody would follow the process. His trust model tells you more about the organization's real operating culture than any stakeholder interview ever could.
THAT'S A CODE.
The Procurement Director
Persona says: "Values vendor references"
She insists on calling every vendor reference personally because early in her career she approved a contract based on a case study that turned out to be fabricated.
THAT'S A CODE.
Every macro someone builds to work around your product is a feature request they've given up on filing. Every defensive architecture a sysadmin builds is a trust signal about your organization. These are codes in action. None of them exist in training data.
ChatGPT-Generated

"Sarah is a 38-year-old Director of Procurement at a mid-market SaaS company. She manages a team of 5 and is responsible for vendor evaluation, contract negotiation, and cost optimization. Her key pain points include lack of visibility into spending and fragmented approval workflows."

From an Actual Interview

"Honestly? I spend half my day in Slack convincing engineers that the vendor they want isn't SOC 2 compliant. My 'approval workflow' is me texting my CFO from the parking lot. I built my own tracker in Notion because nobody else's tool matches how we actually buy things."

One describes a role. The other reveals a code.

The Outsourcing of Curiosity

This isn't an anti-AI take. I spend my days building AI-powered analysis tools. But the most dangerous thing you can do with a powerful tool is use it to skip the work that actually matters.

Lazy AI

Asks ChatGPT to create a persona. Replaces the work. Generates from nothing.

Leveraged AI

Feeds fifty interview transcripts and asks it to surface the contradictions between what customers say and what their behavior shows. Amplifies the work.

When everyone asks the same model the same question, everyone gets the same answer. Every product starts to look the same. Meanwhile, the team that actually cracked their customers' codes is building something nobody else can copy — because nobody else did the work.

When everyone asks the same model the same question, everyone builds the same product. The competitive moat isn't AI. It's the quality of your inputs.

Where This Breaks Your Operating Model

CUSTOMER INSIGHT STRATEGY ROADMAP ENGINEERING SHIP
INSIGHT STRATEGY ROADMAP BUILD SHIP ERROR PROPAGATION

Product operating models are built on a chain of translation. Each handoff is a place where meaning can drift. When the very first link — customer understanding — is itself a hallucination, the drift starts before anyone has made a single decision.

Your Innovation Tax — the $2.30 in maintenance friction for every $1 of actual innovation — doesn't just come from engineering complexity. It comes from building the wrong thing confidently. It comes from Context Collapse that started not in the Jira ticket, but in the persona that was never grounded in reality.

You haven't adopted a product operating model.
You've automated the appearance of one.

The Work That Matters

Customer understanding is slow. It's expensive. It's messy. Sometimes you sit through forty-five minutes of someone talking about their day before they offhandedly mention the thing that reshapes your entire product strategy. You can't prompt-engineer your way to that moment.

That's where the codes live — in the offhand comment, the workaround, the story they tell about the time the system failed them.

Talk to your customers. Sit next to them. Watch them work. Record it. Transcribe it. Then use AI to help you find the patterns you missed.

Observation first, synthesis second. Signal first, abstraction second. Codes first, personas never.
Anything else is just lazy AI
wearing a persona's name tag. Beyond The Alignment  •  Product Operating Models