You’ve seen it. Slide decks piled up like wood. “Pilots.” “AI Centers of Excellence.” LinkedIn posts are buzzing with the word velocity, like it’s a sacred ritual. Everyone is announcing the future as if it just appeared in a Jira ticket after lunch.
This version is loud, almost as if it’s trying to cover something up.
Meanwhile, in quieter corners, something else is happening. No banners. No loud announcements. No claims that everything has changed. Someone just adds AI to the team, like pulling up an extra chair at a table that already exists.
That’s the version worth noticing.
Here’s the sign that distinguishes the noise from the real gains. The teams truly benefiting from AI often share an oddly old-fashioned trait.
They know how to teach.
Not how to enable. They don’t upskill.
They teach by slowing down, explaining the reasons, and answering the same question twice without getting frustrated.
These teams already had some “strange” habits:
– Senior engineers who discuss decisions instead of just fixing bugs and disappearing
– Code reviews that feel like discussions, not court hearings
– Documentation aimed at people who weren’t there when the key moments happened
– A willingness to take their time upfront to avoid spending months fixing subsequent messes
So when AI arrives, no one panics. No one preaches. They don’t label it “transformational.” They call it what it is.
A junior engineer.
One with an impressive memory and no real-life experience. Smart, dedicated, and wrong in interesting ways. No manifestos. Just guidelines, expectations, and a watchful eye. Same as always.
Here’s the part that makes people uneasy in budget discussions: If your organization struggled with people, it will struggle with AI, immediately and noticeably.
Teams that flounder with AI usually flounder long before AI shows up. They didn’t mentor junior staff. They exploited them. They didn’t document systems. They relied on hearsay and a few worn-out heroes. They didn’t reward teaching. They rewarded adrenaline and burnout.
So, AI enters a workplace like that and does exactly what it’s taught:
– It absorbs undocumented systems
– It reflects conflicting tribal knowledge
– It follows hidden rules no one will admit are there
Then, leadership wonders why the output is inconsistent.
That’s not an AI issue. That’s a reflection of the organization as it is.
Blaming the model in that situation is like scolding a new hire for blending in too well.
The best AI practices today aren’t in shiny innovation labs or keynote presentations with dramatic lighting. They are in small, focused teams you don’t hear about because they are busy delivering results.
They use AI to draft initial versions, not final products.
They ask it questions like “What am I missing?” instead of “Tell me what to do.”
They summarize after people agree on what’s true.
They use it to speed up learning, not to shift decision-making.
Nothing groundbreaking. Almost dull. But that’s the point.
Those teams already had a culture of mentoring. AI just fit in, as if it belonged there from the start. To them, bringing in AI isn’t a revolution.
It’s just another day.
Here’s the quieter truth no one wants to see on a conference slide: The most critical model being trained isn’t the LLM.
It’s leadership.
AI doesn’t just learn from codebases and tickets. It learns from behavior, from what gets rewarded, from what gets overlooked, and from how people react when something goes wrong at 2 a.m.
If leaders treat people like interchangeable parts, AI becomes an efficiency engine with no limits. If leaders value clarity, restraint, and shared understanding, AI enhances those traits.
That’s why two teams can use the same model and end up in totally different situations. It’s not about the prompt. It never was.
It’s about the culture being reflected, perhaps too clearly.
The companies quietly thriving with AI didn’t suddenly become smarter. They didn’t discover insights in a vendor briefing. They were already doing the unsung work that no one likes to tweet about:
– Teaching instead of hoarding
– Designing systems that endure when people leave
– Accepting that real skill takes time
AI revealed its values, like a blacklight in a messy room. It’s uncomfortable if you weren’t already cleaning. That’s why the most mature reaction to AI isn’t excitement or fear. It’s an acknowledgment.
“Oh. You’re new. Here’s how we operate. Ask questions. Don’t tackle production alone. And if you make a mistake. We’ll explain why.”
That’s not an AI strategy. That’s just solid engineering. And after all these years, it still feels a bit rebellious.