AI isn't just disrupting work — it's exposing the organizational immune system. Tyson Heaton reveals who's blocking progress and who's quietly building it.
AI isn't just disrupting work — it's exposing the organizational immune system. Tyson Heaton reveals who's blocking progress and who's quietly building it.
For decades, the phrase that killed more improvement efforts than any budget cut or failed software rollout was: “Wait, that’s not my job.”
You know the scene. Someone spots a gap at a handoff. Maybe it’s between sales and delivery, or between setup and operations. They try to close it. Suddenly there’s a job description. Then a supervisor. Then a turf war. Then a steering committee. The gap remains. The customer suffers. The process survives.
Lean practitioners built entire careers navigating this. We learned to map the value stream precisely because the work never respects org chart lines. We learned to walk it from shipment back to inception because the problems hide in the handoffs. We built small, cross-functional teams on purpose. We knew that proximity to the work, and genuine humility about it, was the only way to see clearly.
And then AI showed up.
Here’s what nobody saw coming: the moment a single capable person with access to frontier AI tools can reach across domain boundaries and actually do meaningful work, the organizational immune system doesn’t relax. It goes into hyperdrive.
The thing that six months ago required pulling seven people into a room for three weeks can now be sketched, mocked, and iterated on by one motivated person in an afternoon. And instead of celebration, you get 14 people, including their managers, furious that someone touched their process without a committee.
This is the new era. Welcome to “Wait, that’s my job.”
To survive this organizational angst, or better, to actually extract value from AI before your competitors do, you need to know the cast of characters about to show up in your organization who will very likely disrupt progress with AI.
The Kingdom Keeper didn’t set out to be a blocker. They spent years building deep expertise, earning trust, and slowly accumulating domain ownership. Their team handles a specific slice of the value stream, and they handle it well. The SOC 2 Compliance report is clean. The documentation exists. The handoffs are defined.
The Kingdom Keeper’s deepest fear, one they will never name in a meeting, is that the value of their domain is about to collapse faster than their team can retrain. They will not say this. What they will say is: “We need guardrails. We need governance. We need to make sure this doesn’t break what we’ve already built.”
Every one of those concerns is legitimate. None of them are the actual concern.
In lean terms, the Kingdom Keeper is optimizing for the department. The customer is three steps removed from their daily calculus. Their improvement instinct is real, but it runs upstream, away from flow and the customer.
You will instinctively want to put the Deep Expert on your AI task force. They are the most technically credentialed person in the room. They have the deepest knowledge of the stack, the process, the edge cases. They have earned this reputation, and they have paid for it in years of sacrifice.
They are also, statistically, your highest-risk selection.
The Deep Expert has spent their career defining what requires expertise. The arrival of AI is a direct challenge to the value of that expertise. Watch them in early AI sessions. They will not explore the depths of what the tool can do. They will probe for its failures. Every hallucination is a data point. Every limitation is a boundary they can draw around themselves. Like most healthy intelligent people, the Deep Expert is interested in what AI can’t do.
This is not laziness or malice. It is self-preservation operating at a level below conscious decision-making. The Deep Expert’s identity is more baked into their role than into the outcome their role was created to produce.
In lean terms, they are confusing the process with the purpose. They have optimized the method so long that the method became the goal.
You have met this person. They opened the all-hands with an AI slide. They mentioned ChatGPT four times in the last leadership offsite. They may have even said the words “We need to move with urgency” and “This is a platform shift.”
They have a Claude.ai tab open in their browser. It has not been used this week.
The AI Evangelist Executive is enthusiastic, and their enthusiasm is real. They have read the articles. They have seen the demos. They believe, genuinely, that AI is going to change their business.
What they have not done is stay up until 1:00 am arguing with a frontier model about a process problem they couldn’t solve in three meetings. They are not burning tokens. They are not building anything. They are narrating a transformation they are not participating in.
This is the most dangerous persona on the list, not because they obstruct, but because they create the appearance of organizational permission without providing the actual conditions for experimentation. Teams under AI evangelists frequently get the mandate without the runway, the enthusiasm without the executive cover when things get weird and territorial, which they always do.
In lean terms, they are confusing the declaration of intent with the gemba. They have not walked the flow.
The Token Burner is not hard to find if you know what to look for. They are using tools the IT department hasn’t sanctioned. They are building things in their personal time that touch work problems. They found the loophole in the permissions structure and have been quietly pulling threads on integrated process for six weeks.
They may not have a prestigious title. They are probably not on the AI task force. In fact, if they raised their hand to be on the AI task force, they probably didn’t get selected because they didn’t look like the obvious choice.
The Token Burner has something the other personas do not: genuine curiosity about customer value and a low tolerance for waste that doesn’t care whose domain the waste lives in. They have seen the handoff failure from both sides. They have the frustration of someone who’s been close enough to feel it.
They are also, in most organizations, about to get in trouble.
The organization’s first instinct upon discovering the Token Burner is not to promote them. It is to ask what they accessed, who gave them permission, and whether any data was at risk. By the time the governance conversation is over, the Token Burner’s energy has often been converted into compliance anxiety.
Find them before this happens.
Lean practitioners are, in a genuine sense, the most sociotechnically equipped people in most organizations to navigate this moment. They know how to structure cross-functional teams. They know how to walk value streams. They know how to distinguish genuine humility from expertise-as-armor. They know that the messenger gets shot, and they know how to design for that.
But lean practitioners have their own relationship with technology, and it is complicated. Many came up in physical environments. Many developed their instincts in manufacturing or service operations where the tool was secondary to the method. Many are quietly skeptical about whether a probabilistic language model understands the difference between a pull system and a push system.
That skepticism, handled well, is actually useful. The lean practitioner who can stay curious about what AI genuinely can and cannot do, who does not need AI to validate their career or threaten it, who approaches the frontier model the same way they approach a gemba walk, with observation before judgment, is genuinely rare and genuinely valuable.
The lean practitioner who defaults to methodology defense is just another Deep Expert in different clothing.
Last October, in The Boundary Is Moving: AI Changes What We Expect from Team Leaders, Managers, and Executives, I wrote about a team leader at O.C. Tanner who, a year earlier, had taught his iPhone to spot aluminum finish defects in weeks, building a proof of concept for vision-based quality inspection that experienced operators and I had collectively written off as too expensive and too complex to integrate. He was six months into the role. Nobody asked him to do it. Nobody gave him a budget.
I caught up with him recently. He is a group leader now, still on nights, still going to school, now running multiple teams. And he has not stopped building.
Since that defect detection experiment, he has built a full leader-standard-work application from scratch: a daily planner with task tracking and metrics, a gemba audit tool that pushes observations directly to the quality system without copy-paste, a live production dashboard that pulls line data his team leaders were manually tallying at the end of every shift. He built the API connections himself, using dev tools and a Jira user token and a lot of late nights. He hosted it on his free Azure account and handed out beta invites to other leaders. Roughly eight of them are using it.
He did all of this without a ticket, without a sprint, without asking IT.
This is what a Token Burner looks like 18 months in. This is the return on protecting that energy rather than converting it into compliance anxiety.
But here is where the story gets interesting, and why I am watching it closely. The app works. Leaders like it. And now the organization has to decide what to do with it. Does IT review the code and host it on the network? If they do, who owns it? When a leader wants a new feature, does that request go back to him or into a sprint queue? Does the tool get absorbed into the existing quality system, rebuilt by people who were never group leaders and will make an entirely different set of tradeoffs? The person who built it understands exactly why each design decision was made. The organization may not.
This is the next trial. Not the building. The handoff.
If the integration goes well, I’ll name this Token Burner. He’s earned it.
If you are serious about extracting genuine value from AI, the org structure that works looks less like a center of excellence and more like a kaizen team with sharper tools.
Forget the five to seven person task force. The model that actually moves is multiple small cells, two to three people each, working different parts of the value stream in parallel. Three people can make a decision and iterate. Five people schedule a meeting about it. The goal is a network of small teams with enough connective tissue between them to stay oriented toward the same customer outcome, not a single cross-functional committee trying to do everything at once.
Pull them from the work, not from the credential stack. Look for people who are frustrated that the current process gets in the way of the customer, not people who have mastered it. The signal you want is someone who gets animated by the idea of a value stream that can actually respond to changing customer demand, someone with a genuine sense for what clients value. Find a coach or an outside perspective who is not tied to the social structure, someone whose job is to walk the flow, not to give technical advice. Give the group real access across domains, not screenshots of process maps. Start at the point of delivery and work backward.
Your executive steering committee needs to be small and cross-functional by biography, not by assignment. The executives who grew up through their silo will tense up when they see the boundaries move, even when they believe it is correct. Their bodies will answer faster than their values. You can inform them. You cannot build the future with them in the driver’s seat.
And when you find the Token Burners? Protect them. Give them the cover they need. They already have a sense for where the opportunity lies. They’ve been losing sleep over it.
The organizations that figure this out will not win because they bought better AI tools. They will win because they finally solved the same problem lean has been pointing at for 40 years: the customer doesn’t care about your org chart, and the work has always known that. AI just made it impossible to pretend otherwise. 
The Lean Tech & AI Journal publishes practitioner-grounded perspectives on lean thinking and AI adoption. If this article made you think of someone in your organization, forward it to them. If it made you think of yourself, sit with that.
Be sure to check out LeanTech, an online LEI initiative dedicated to helping organizations apply lean practice to the technology layer of their operations. This effort is led by Tyson Heaton, Executive Director, and anchored by a faculty of practitioners whose work spans Toyota production systems, enterprise software transformation, AI-enabled lean coaching tools, and technology-native lean practice. At its core, the initiative rests on a straightforward premise: the same discipline that made lean thinking effective in operations and supply chains (understand the work, build capability in people, improve continuously) is now both applicable and necessary in the technology environments encountered by nearly every modern organization.
Org Strategy
Tyson Heaton spent 15 years running operations in manufacturing at JBS, Schreiber Foods, and Greencore, learning lean on the floor before moving into enterprise technology leadership at O.C. Tanner. The frustration he found there became a throughline: organizations that had mastered continuous improvement on the shop floor were treating their technology layer as a black box, a dependency to manage rather than a capability to master. He now leads LEI's LeanTech/AI initiative, working with organizations done accepting that tradeoff -- the work is moving faster than most can keep up with, and he'd rather be part of sharpening that thinking than watching from the side.