The Year Marcus Almost
Lost His Best Team to
an AI They Never Asked For
A story about a 400-person logistics company, a CEO who thought he was doing the right thing, and what happened when his people decided — quietly, together — that they had had enough. And what he did next that changed everything.
Marcus had been CEO of Clearfield Logistics for eleven years. He had built it from 40 people to 400. He had navigated two recessions, a global supply chain meltdown, and one very bad acquisition decision in 2021 that he had spent three years quietly fixing. He was not the kind of executive who panicked. He was the kind who read the room, made deliberate decisions, and trusted his people.
Which is why the board meeting in March landed so hard.
“Your competitors are deploying AI in their dispatch and routing operations,” the lead board member said, setting down the research report with the particular authority of someone who had just paid $40,000 for it. “Fifteen percent efficiency gains. Twenty percent reduction in delivery exceptions. You need to move.” Marcus nodded. He went back to the office. He called his COO, his head of operations, and his IT director. “Find me something,” he said. “Something that works.” They found three vendors in two weeks. The contract was signed in six.
— A composite story drawn from real transformation conversations over 25 years. All names and identifying details are fictional.If you are a CEO or a senior leader who has ever felt the board’s breath on the back of your neck about AI, you already know where this story is going. You have felt that particular mixture of genuine strategic conviction and quiet urgency that comes from watching competitors move and knowing that every quarter you wait has a cost. Marcus was not wrong to move. He was wrong about how. And the difference between those two things is the entire story of workforce transformation in the AI era.
The system went live on a Tuesday. By Friday of the same week, Marcus’s most senior dispatcher — a woman named Diane who had been with the company for fourteen years and who knew the quirks of every driver, every route, and every client relationship in the portfolio — had stopped eating lunch with the operations team. She ate at her desk. Alone. With the particular focused silence of someone who has made a decision they have not announced yet.
Marcus did not notice. He was busy reviewing the adoption dashboard the vendor had helpfully provided. Seventy-one percent active users in week one. “Better than average,” the vendor’s success manager told him cheerfully on their weekly call. Marcus sent the number to the board in his monthly update. Progress, he wrote. Early adoption strong.
He did not know that Diane had started taking calls from a competitor on Wednesdays. He did not know that two of his three best operations managers had connected with each other on LinkedIn the previous weekend and spent forty-five minutes discussing whether the culture at Clearfield was changing in ways they were uncomfortable with. He did not know that the workarounds his team had quietly developed to bypass the AI’s routing recommendations were now so embedded in daily operations that the system’s actual utilization rate — not the login rate the dashboard measured, but the rate at which its outputs were actually acted on — was sitting at 23%.
— The gap between dashboard adoption and real utilization is one of the most consistent patterns in failed AI deployments.The story above is not unusual. It is not even particularly dramatic by the standards of what I have seen in 25 years of working alongside organizations through significant change. What makes it worth telling is not the crisis. It is what Marcus did when he finally found out — and the five specific things he changed that transformed not just the technology deployment, but the entire relationship between his leadership team and the people doing the work.
What the Numbers Look Like When You Dig Below the Dashboard
“The technology did not fail Marcus. His people were not obstinate. The deployment failed because it asked his most experienced people to trust a system that had never been introduced to them as something built with their expertise — only as something replacing their judgment.”— Tim Booker, President & CEO, MindFinders
And What Every One of Them Has in Common
None of these mistakes were born from bad intent. They were born from urgency, a vendor’s confidence, and the particular blind spot that affects most leaders when they conflate deploying technology with transforming the organization. Click each to see what went wrong — and why it matters.
Diane knew things about Clearfield’s routing operations that were not in any system. She knew that Client 47 always needed a morning delivery on Tuesdays because their loading dock was unavailable after noon. She knew that the route through the Eastside industrial corridor ran 12 minutes faster on Thursdays because a school traffic pattern shifted. She knew which drivers were reliable under pressure and which needed structured dispatch support. That knowledge — built over fourteen years — was more valuable to an AI routing system than anything the vendor brought to the table.
Nobody asked her for it. Nobody explained that the system needed it to work well. She was shown a demo, given a login, and told the system would handle routing going forward. She interpreted this, correctly, as: your expertise is no longer the asset it was.
Before any AI deployment, your most experienced people should be interviewed as subject matter experts — not informed as system users. Their knowledge is the data the AI needs to be good at your specific business.
Seventy-one percent login rate in week one. The vendor called it “better than average.” Marcus put it in his board update. What nobody measured was whether the team was actually acting on the system’s outputs — or logging in, seeing something that did not match their on-the-ground reality, and then doing the job the way they had always done it while the system ran in a separate window.
The real utilization rate — AI outputs actually acted on without manual override — was 23%. Marcus discovered this four months in, by accident, when a logistics audit flagged that the system’s routing recommendations were being overridden at a rate that was erasing the efficiency gains entirely.
Login rate is not adoption. Adoption is the rate at which people trust the system’s outputs enough to act on them. Measure that — and when it is low, treat it as information about the system’s credibility, not the team’s compliance.
The AI was handling routing recommendations. But Diane’s job title was still Dispatcher. Her KPIs were still measured the same way. Her performance review framework still evaluated her on the same criteria it had for twelve years. The tools had changed. The role description had not. She was being asked to do a fundamentally different job — monitor AI outputs, handle exceptions, apply contextual judgment — while being evaluated on the old job she was no longer doing.
That ambiguity is one of the most reliable early warning signs of impending departure. When the job changes but the role does not, experienced people feel like they are being made redundant in slow motion.
Every AI deployment that changes workflows must include a role redesign. What does excellent performance look like in the new environment? What new skills matter? How does the performance framework reflect the actual job the person is now doing? These questions must be answered before the tools go live.
The vendor ran a two-day onboarding session. It covered the platform thoroughly — how to navigate the interface, how to read the routing outputs, how to flag exceptions. What it did not cover was the organizational why. Why was Clearfield doing this now? What did it mean for each role over the next 12 months? What was Marcus’s vision for what the company looked like when this transformation was complete? Those are not vendor conversations. They are leadership conversations. And they were never had.
When the people most affected by a change hear its rationale from a vendor — rather than from the leader who decided to make it — they draw the most reasonable conclusion available to them: leadership does not think this conversation is important enough to have personally. And they respond accordingly.
The vendor trains the team on the tool. The leader communicates the vision, the rationale, the timeline, and — most importantly — what it means for the people in the room. Delegating that conversation to a vendor is not an efficiency gain. It is an abdication of the most important leadership responsibility in a transformation.
Diane gave her notice on a Thursday in June. She had been offered a role at a competitor — a role that, she told Marcus in the exit interview with the particular honesty of someone who has already decided to leave, had the same AI tools but a manager who had spent three hours with her before the launch explaining exactly how the role was evolving and what her expertise was worth in the new environment.
Marcus sat in that exit interview with the particular stillness of someone absorbing information that should have arrived months earlier. He had had every opportunity to have that conversation. He had not known he needed to. The dashboard was green. The board was satisfied. The vendor was cheerful. The signals that mattered — the solo lunches, the LinkedIn connections, the overrides accumulating quietly in the system log — were in a language he had not been reading.
The retention conversation in an AI transformation is not the exit interview. It is the conversation you have three months before you launch — and every month afterward — with the people whose expertise is most at risk of feeling displaced. By the time someone is already leaving, you are not managing retention. You are managing aftermath.
Marcus cancelled his Monday morning executive team meeting and replaced it with something he had never done in eleven years of running Clearfield. He walked the floor. Not to check on operations. Not to inspect anything. He walked it with the specific intention of listening. He sat with dispatchers, drivers, and operations supervisors and asked three questions: What is the AI getting right? What is it getting wrong? And what do you need from me that you are not currently getting?
He took notes by hand. He thanked people specifically — not generically. He did not defend the system or explain the strategy. He listened for three hours. By noon, he had filled twelve pages of a legal pad with things he had not known. By the end of the week, he had cancelled three vendor feature requests, initiated a role redesign for every operations position affected by the system, and personally scheduled one-on-one conversations with every team member Diane had worked closely with.
It was the most productive week of the entire transformation. None of it required a new technology purchase.
— This moment — the walk, the listening, the twelve pages — is what the 70% looks like in practice.The Five Moves That Turned the Transformation Around
He Made His Senior People Co-Architects of the System — Not Users of It
Marcus convened a working group of his six most experienced operations people — including the two managers who had been quietly connecting on LinkedIn — and gave them a specific mandate: find everything the system is currently getting wrong and tell us why. No consequences for candor. He told them their expertise was the most valuable input the system needed to improve. He meant it and they could tell.
Within three weeks, the working group had identified fourteen routing assumptions the system was making incorrectly — including the Client 47 Tuesday pattern that Diane had never been asked to share. Fixing those fourteen assumptions improved the system’s output quality measurably. More importantly, it changed the team’s relationship to the system entirely. They were no longer subjects of the AI. They were its collaborators.
He Redesigned Every Role That Had Changed — With the People in Those Roles
With his HR director, Marcus documented what excellent performance looked like in each AI-augmented position. Not what the old job had required — what the new one actually demanded. Exception management. Contextual override judgment. System feedback contribution. Client relationship quality that no AI could replicate. These were the new core competencies. He rewrote performance frameworks. He updated job descriptions. He made the new role visible.
For the first time, his dispatchers had a document that said explicitly: your expertise is worth more in this environment, not less. Your judgment is the quality check the system needs. The role had been redesigned from “person who does the routing” to “expert who ensures the routing is right.” That reframe changed how people felt about showing up.
He Changed What He Was Measuring — And Who He Reported It To
Marcus retired the vendor’s adoption dashboard as his primary reporting metric and replaced it with three business outcome measures: delivery exception rate, route efficiency versus manual baseline, and team confidence score — a simple monthly survey asking each operations employee how much they trusted the system’s outputs on a 1–10 scale. He reported these three metrics to his board in place of the login rate. The first month, two of them looked worse than the vendor’s dashboard had suggested. He presented them anyway. The board respected it.
The discipline of measuring actual outcomes rather than platform activity changed every conversation downstream. Vendors stopped bringing him adoption stories. He started bringing them performance data. The accountability shifted in the right direction.
He Built the Conversation Into the Calendar — Permanently
Marcus scheduled a 30-minute monthly operations review that had nothing to do with performance numbers. Its only agenda item was: what is the AI getting wrong this month, and what do we want to do about it? Every operations team member with direct system interaction attended. His presence at every session was non-negotiable. He asked questions. He took notes. He followed up on every item from the previous month in the first five minutes of the next one.
The signal this sent was unmistakable: the transformation is not finished because we launched. It is ongoing. Your input is the engine that makes it better. And I am going to show up every month to prove I mean that. In twelve months, that meeting had generated 47 system improvements, identified three new use cases the vendor had not considered, and become the most attended standing meeting at Clearfield.
He Started Having the Career Conversation Before People Needed to Have It Elsewhere
Every quarter, Marcus’s managers held structured stay conversations with every operations employee whose role had been meaningfully affected by AI. Not performance reviews. Not check-ins. Conversations with a specific, documented agenda: here is what your role is evolving toward, here is what we see in you that we want to invest in, here is what your future at Clearfield looks like from where I sit. What do you need from me?
The conversations were sometimes uncomfortable. Some surfaced concerns that required real responses. Several led to role adjustments that Marcus would not have made without them. One led to the promotion of an operations supervisor who became, within 18 months, the best AI integration manager Clearfield had — a role that had not existed before. The company did not lose that person to a competitor. They grew her into a function the transformation required.
What Clearfield Looked Like — Before and After Marcus Changed His Approach
Marcus presented to the board again in September. The results slide showed everything they had originally wanted — efficiency gains, exception reduction, cost improvement. But before he got to the results, he spent ten minutes on something he had never put in a board presentation before. He talked about Diane. He talked about the three months he had wasted measuring logins instead of trust. He talked about the twelve pages of notes he had taken on a Monday morning walk he had never planned to take.
“We had the right technology from the beginning,” he told the board. “What we were missing was the understanding that technology transformation and workforce transformation are the same thing. You cannot do one without the other. The day we understood that — really understood it, not just in a slide — is the day the numbers started moving.”
He paused. “I’m telling you this because we’re about to start phase two. And I want you to know that half the budget is going to the people work. Not because it’s the soft stuff. Because it’s the half that determines whether the technology half works at all.”
The board approved the budget. All of it.
— The ending Marcus earned. The lesson every transformation leader needs to hear before they lose their own Diane.“Workforce transformation and AI transformation are not two initiatives running in parallel. They are the same initiative. The organizations that understand this from the start do not just get better technology outcomes. They get better people outcomes. And those compound in ways that no platform upgrade ever will.”— Tim Booker, President & CEO, MindFinders
The MindFinders Difference
We Help Organizations Be More Like Marcus at Month Seven — Not Month One.
MindFinders has been sitting across from Marcus in that conference room — in one form or another — for 25 years. We know the urgency that drives the fast decisions. We know the board pressure, the vendor confidence, and the very understandable desire to move quickly on something important. We also know what the twelve pages of notes on the Monday morning walk look like — and how to get your organization to that understanding before Diane gives her notice. The entire MindFinders practice is built around the insight that technology transformation and workforce transformation are the same thing. That is how we have always worked. It is the only way that consistently delivers.
- We engage your most experienced people as co-architects before any AI system is introduced to them as users
- We redesign roles and performance frameworks to reflect what the AI-augmented job actually requires
- We build structured stay conversations into the transformation calendar — not just exit interviews into the offboarding process
- We help leaders measure what matters — trust, utilization, business outcomes — not just login rates and adoption dashboards
- We create the feedback loops that generate continuous improvement rather than a static deployment
- We advise on how to communicate the transformation in your voice — to your people, at every level — before they form their own conclusions
Before You Launch — Let’s Make Sure You’re Not Setting Up Your Own Diane Story.
Let’s talk through your AI transformation plan, identify where the people risks are concentrated, and build the workforce strategy that means your best people are still there when the results arrive — and proud they helped build them.
Schedule Your Free Consultation
