AI has become the new leadership Rorschach test.
Some people are terrified it will replace them. Others are convinced it can do anything. Most organizations are doing the same thing they always do under uncertainty: chasing certainty, chasing speed, and hoping a tool will save them from the hard work of clarity.
I recently had a conversation in which someone told me that AI will replace software developers. No school needed. No training. “My friend built a CRM with no code experience,” they said, as if that settles it.
I don’t doubt the friend built something.
I doubt the friend built a system that can survive reality.
And if Deming were in the room, I think he’d cut through the hype with one question:
What are you accelerating? Because AI doesn’t create judgment. It doesn’t create purpose. It doesn’t create quality. It creates output faster.
Deming spent his life warning leaders about exactly this pattern: confusing activity with improvement, and confusing tools with systems.
The real AI debate isn’t “can it replace developers?”
The real AI debate is whether leaders will use it to:
- improve the system, or
- amplify the mess
If your priorities are fuzzy, AI will ship confusion faster.
If your decision-making is political, AI will generate more “evidence” to justify whoever has the most power.
If your culture rewards optics, AI will produce beautiful artifacts that look credible while the underlying risks stay unnamed.
That’s not an AI problem.
That’s a leadership problem.
What Deming would notice first: fear, targets, and the illusion of control
Deming’s work wasn’t about manufacturing. It was about management behaviour and the systems leaders create.
So if you want a Deming lens on AI, start with the parts of his philosophy that are painfully relevant right now.
1) “Drive out fear” (Point 8)
Fear changes behaviour.
When people fear replacement, they hide. They stop learning in the open. They stop asking questions. They stop admitting uncertainty. They start protecting their role instead of improving the work.
When leaders fear being left behind, they buy tools before they build clarity. They mandate adoption. They push for “AI productivity” targets. They create a new layer of pressure.
And pressure doesn’t create quality.
Pressure creates shortcuts.
2) “Eliminate slogans, exhortations, and targets” (Point 10)
A lot of AI rollouts are basically slogans:
- “Use AI to be 30% more productive.”
- “Everyone must use AI daily.”
- “We’re an AI-first company now.”
Deming would ask, "What system did you change to make that possible?"
Because when you demand outcomes without changing the system, you get gaming, busywork, and adversarial behaviour.
People will hit the metric. They’ll produce more output. They’ll also create more defects, more rework, and more risk, just faster.
3) “Cease dependence on inspection” (Point 3)
One of the most common AI mistakes is using it as a late-stage inspector:
- “AI will review the code.”
- “AI will catch the bugs.”
- “AI will write the tests.”
That’s still inspection.
Deming’s point was that quality has to be built into the system upstream. AI can help, but it can’t replace the decisions that define “good” in the first place.
The “my friend built a CRM” story: output isn’t the same as capability
Yes, AI can help someone build a working app quickly.
But a CRM isn’t a demo.
A CRM is:
- Customer data
- Permissions
- Audit trails
- Uptime expectations
- Security risk
- Compliance risk
- Change management
- Ongoing maintenance
In Deming terms: it’s a system.
And systems fail when leaders confuse a prototype with a product, or a feature with a capability.
AI can generate code. It can generate screens. It can generate confidence.
It cannot generate responsibility.
The Deming question that matters: What is your definition of quality?
Deming would say your real definition of quality isn’t what you write on a page.
It’s what you protect when pressure shows up.
So here’s the AI version of that:
When AI makes it easy to move fast, what do you protect?
Because if you protect speed, AI will help you ship faster.
If you protect truth, AI can become a powerful amplifier for learning, evidence, and improvement.
A human-first operating system for AI (what to make explicit)
If you want AI to help instead of harm, you need to make a few decisions visible.
This is the part most organizations avoid, because named choices create responsibility.
1) Intent: What are we using AI for?
Not “to be more productive.”
For what outcome, in what context, with what constraints?
2) Guardrails: what won’t we sacrifice?
Examples:
- Customer data safety
- Correctness in financial calculations
- Accessibility
- Legal/compliance requirements
- Team health
3) Decision rights: who decides what “good” means?
AI can propose. Humans decide.
Name who owns the call, not for control, for clarity.
4) Evidence: what proof counts as safe enough?
A confident AI answer is not evidence.
If it can’t point to proof, it doesn’t get a confidence score.
5) Learning loops: how will we learn without blame?
If people get punished for mistakes, they’ll hide them.
If they hide them, you lose the signals you need to improve.
AI won’t replace developers. It will replace unclear leadership.
That’s the part nobody wants to say out loud.
AI will absolutely change the work. It will change the skills. It will change how fast teams can produce output.
But the organizations that win won’t be the ones with the flashiest tools.
They’ll be the ones that:
- Create constancy of purpose
- Drive out fear
- Make trade-offs explicit
- Build quality into the system
- Use evidence to learn
In other words, the ones that lead, like Deming, were begging leaders to lead.
If you’re trying to build agility without wearing your people down, book a focused call:
https://meetings.hubspot.com/jberriault. Bring one team that looks “fine” from the outside but feels stretched underneath. We’ll look at what’s contributing to the strain, system, behaviour, or both, and find the first change that helps people breathe, speak honestly, and work in a healthier way. What’s the earliest sign you notice when a team is starting to burn out?
If you want a quick starting point, run the short assessment at berriaultandassociates.com to see what your team or organization needs most. Join the community here:
https://www.skool.com/evolutionary-agilist-9478 More in-depth writing here:
https://substack.com/@jeremyberriaultReferences
