Pilot Purgatory - Part 1
Pilot Purgatory
A Story of AI Transformation
By Scott Weiner (AI Lead at NeuEon, Inc.), inspired by conversations with Erwann Couesbot (CEO of FlipThrough.ai)
Note: This is a work of fiction. All characters, companies, and events are fictional composites created for illustrative purposes. While the industry statistics cited are real and sourced, the narrative is designed to illuminate common patterns in enterprise AI adoption, not to depict any actual organization or individuals.
This is Part 1 of a serialized story exploring why enterprise AI initiatives fail, not from lack of technology or talent, but from invisible organizational dynamics that doom them from the start.
Prologue: The Announcement
The traffic on Route 9 moved in fits and starts, brake lights blooming red ahead of Marcus Chen’s windshield like warning indicators on a production dashboard. December mornings in Massachusetts had a particular quality of gray that made everything feel like it was waiting for something.
The tech news podcast had been running in the background for twenty minutes, familiar voices discussing familiar topics. Funding rounds he didn’t care about. Product launches that wouldn’t affect Thornfield Manufacturing. Then a name he recognized.
“…and the big story this week is Strathmore Industrial’s announcement that their AI-driven supply chain optimization is now fully operational across all seven facilities. The company claims a 23% reduction in procurement costs and, get this, reports a 40% improvement in demand forecasting accuracy.”
Marcus’s hand found the volume control before he’d consciously decided to turn it up.
“What makes this particularly interesting,” the host continued, “is that Strathmore started this initiative roughly eighteen months ago. They brought in a dedicated AI leader, someone they called a Chief AI Officer, and that seems to have made all the difference.”
The words landed in Marcus’s chest like stones dropped in still water.
Strathmore. Their direct competitor. Same industry, similar size, serving adjacent customer segments. Eighteen months ago, both companies had been in the same position: watching AI transform other sectors and wondering when it would reach manufacturing. Both had felt the same board pressure. Both had allocated budget. Both had capable teams.
And now Strathmore was reporting results while Thornfield was preparing for another board meeting where “we’re making progress” would have to carry the weight of a $1.4 million investment that had yet to produce a single system in full production.
Marcus watched a delivery truck merge ahead of him, its logo cheerful against the gray December sky. The podcast had moved on to another topic, but the numbers echoed. Twenty-three percent cost reduction. Forty percent accuracy improvement. The kind of metrics that made board members smile and competitors sweat.
He had numbers too, but they told a different story: three engineers who had left with AI skills trained on Thornfield’s dime, a chatbot pulled from production after its confident hallucinations created legal exposure, four pilots stalled at various stages of development with none ready for prime time, and an observability platform purchased six months too late.
The traffic broke, and Marcus accelerated through a yellow light, his mind calculating the gaps. Eighteen months of effort. A team that had tried everything they knew how to try. Technology that worked brilliantly in demos and failed quietly in production.
He thought about calling Sarah, his security engineer. She had warned him, back in month three, that they were moving too fast without governance. He had added her concerns to the backlog and kept pushing. The backlog was where priorities went to wait, and security had waited until a customer screenshot proved the cost of waiting.
He thought about Daniel, the engineer who had become genuinely skilled at machine learning during his time at Thornfield. Daniel who had left for a 40% raise and pure AI work, no legacy maintenance. Daniel who had thanked Marcus for the opportunity on his way out the door, genuine gratitude mixed with something that looked like escape.
He thought about Linda in procurement, who had evaluated thirty-two vendors and picked the one with the most confident demo. The demos had been spectacular. The production implementation had been something else entirely.
And he thought about Jennifer Park, the board member who would be sitting across from him in three hours, asking questions she already knew the answers to. Jennifer who had been a CIO herself, who understood exactly how these initiatives could stall, and who would not be satisfied with progress reports that didn’t progress.
The familiar exit for the Thornfield campus appeared ahead, and Marcus signaled his turn. The building’s glass facade caught the weak winter sun, its surfaces clean and modern, a manufacturing company that had modernized its facilities even if it hadn’t quite figured out how to modernize its relationship with AI.
In the parking lot, he sat for a moment with the engine off, the car cooling around him in the December air. The podcast had ended, leaving silence where Strathmore’s success had been.
“What would have been different?”
He said it aloud, to the empty car, to no one. The question he had been circling for months without quite letting himself ask.
The answer, he suspected, wasn’t about budget since they had spent more than enough, wasn’t about talent since his team was capable, and wasn’t about technology since the tools existed and improved weekly.
The answer was about something else—something structural, something he hadn’t seen clearly until he was already eighteen months into learning it the hard way.
He reached for the door handle, then paused.
It had started eighteen months ago, in a Q4 board meeting, with a slide deck about competitive threats and a mandate that felt achievable. In that conference room, with the autumn light coming through the windows and Jennifer’s questions sharp but not yet accusatory, Marcus had felt the familiar confidence of a challenge understood.
“We’ve modernized before,” he had thought. “Every transformation has complexity. I’ve navigated ERP migrations, cloud transitions, security overhauls. Technology problems are what I solve.”
That was the moment, he realized now—that was where the path had forked.
He opened the door and stepped into the cold morning air, the building waiting ahead of him with all its accumulated decisions. Three hours until the board meeting. Eighteen months of answers he would have to give.
But to understand how he had arrived at this particular December morning, with Strathmore’s success ringing in his ears and his own slides feeling thin in his briefcase, he would have to go back.
Back to that first board meeting, to the confidence that had felt so reasonable, to the beginning.
Chapter 1: The Mandate
The Q4 board meeting had already run three hours when Jennifer Park pulled up her slide deck.
Marcus Chen shifted in his chair, trying to find a position that didn’t acknowledge how long they had been sitting. The autumn light through the conference room windows had shifted from morning gold to afternoon flat, and his coffee had gone cold sometime during the facilities report. On his desk back in his office, a stress ball from some vendor conference sat waiting, its logo worn smooth on one side from years of unconscious use. He wished he had it now.
“I want to talk about competitive positioning,” Jennifer said, her voice carrying the particular quality that made everyone in the room pay closer attention. As board members went, Jennifer was one of the more engaged. Former CIO of a Fortune 500, genuinely understood technology, asked questions that revealed she had done her homework. Marcus respected her, even when her questions made him uncomfortable.
The slide appeared: “AI Adoption: Competitive Landscape Analysis.”
“Three of our direct competitors announced AI initiatives in the last quarter,” Jennifer continued. She clicked to the next slide. Logos, timelines, stated outcomes. “Strathmore has a pilot in supply chain optimization. Redmont Manufacturing claims 15% efficiency gains in quality control. Grimkey just hired their second machine learning engineer.”
Marcus felt David Kim, the CEO, glance in his direction. The familiar look that said: this is about to become your problem.
“We’ve been exploring AI applications,” Marcus said, pulling up his own presentation on the conference room screen. The slides felt thin compared to Jennifer’s research, but they were honest. “Customer service automation is the obvious first target. We handle about 2,000 product specification inquiries a month. A well-trained chatbot could reduce that load significantly.”
“What else?” Jennifer asked.
“Predictive maintenance for factory equipment. We have sensor data going back five years that we’ve never really analyzed. And our procurement team spends considerable time on demand forecasting that could benefit from pattern recognition.”
Jennifer nodded, but she hadn’t stopped clicking through her slides. The next one showed a chart: AI investment by manufacturing sector. The trendline went up and to the right in a way that made comfortable companies uncomfortable.
“What’s our timeline for production deployments?” she asked.
Marcus did the mental math. Scoping, proof of concept, testing, integration, training. “Six months for initial pilots. Twelve to eighteen months for production-ready systems.”
“Exploring isn’t shipping,” Jennifer said, not unkindly, but with the precision of someone who had managed enough technology initiatives to know the difference. “I’m not criticizing what you’ve done, Marcus. I’m asking what we’re going to do.”
The room waited. Marcus could feel David processing, the other board members exchanging the kind of glances that happened when strategic direction was being set in real time.
“If I may,” Jennifer continued. She clicked to her final slide. “I’m not suggesting we panic. But the window for being an early mover in AI is closing. In two years, this won’t be competitive advantage. It will be table stakes. The question is whether Thornfield wants to be ahead of that curve or behind it.”
David cleared his throat. “What are you recommending?”
“A serious investment. Not a hobby project. Dedicated resources, clear milestones, board-level visibility.” She looked at Marcus. “You have a capable team. The question is whether you have the bandwidth and focus to make this real.”
Marcus felt the weight of the moment. Twenty years in enterprise IT had taught him how to read rooms, how to navigate the politics of budget requests and resource allocation. This was different. This was a mandate wrapped in a question.
“We can make it happen,” he said. “I’ll need budget for infrastructure, potentially some external expertise, and permission to reallocate some engineering resources.”
“Define budget,” David said.
“Let’s start with $250K for the first phase. Pilots across three use cases. We’ll have measurable results by Q2.”
Jennifer’s expression was like someone waiting to see if a bridge would hold weight before crossing it.
“That seems reasonable,” David said, making a note on his pad. “Get me a detailed proposal by end of week. We’ll approve what makes sense and track progress quarterly.”
“Keep me updated,” he added, which Marcus had learned over six years meant something closer to: make this work so I don’t have to think about it again until it’s either successful or too late to save.
The meeting adjourned twenty minutes later, the usual mingling and small talk that followed board sessions. Marcus gathered his laptop, exchanged nods with people as they filtered out, and made his way toward the door.
Jennifer caught his arm in the hallway. “Marcus. A moment.”
He stopped, the hallway suddenly feeling longer than usual.
“That wasn’t a setup,” she said. “I’m genuinely trying to help.”
“I know.”
“Do you?” She studied him the way she studied balance sheets. “I’ve seen a lot of technology initiatives. The ones that fail usually fail because of resource fragmentation. Everyone’s excited at the start, but no one has the dedicated focus to push through the hard parts.”
“I hear you.”
“I hope so.” She released his arm. “You’re a good CTO—that’s exactly why I’m worried. Good CTOs are stretched too thin to give AI the attention it needs. It’s not a character flaw; it’s math.”
Marcus nodded, filed the warning somewhere between useful and unsettling, and continued toward the parking lot.
David was leaning against Marcus’s car when he got there, jacket off despite the October chill, the posture of a CEO who wanted a conversation off the record.
“Well,” David said. “That was something.”
Marcus clicked his key fob, the car chirping its acknowledgment. “Jennifer made some good points.”
“Jennifer always makes good points. That’s what I pay her for.” David pushed off from the car but didn’t move toward his own. “What do you actually think? Not what you said in there. What do you think?”
Marcus considered the question. The honest answer was complicated.
“I think AI is real,” he said finally. “I think we have to do something. I also think we don’t really know what we’re doing yet, and that’s going to be uncomfortable for a while.”
David laughed, a short sound without much humor. “At least you’re honest.”
“We’ve modernized before,” Marcus said, and he meant it. “The ERP migration was supposed to be impossible. The cloud transition had everyone convinced we’d lose critical data. Zero-trust security was a year of pain. We figured all of it out. This is another technology challenge, and technology challenges are what I do.”
“You sound like you’re convincing yourself.”
“Maybe.” Marcus opened his car door. “But I also sound like someone who’s done this before. AI isn’t magic. It’s infrastructure and data and integration, like everything else. We just need to approach it systematically.”
David nodded slowly. “I hope you’re right. Because Jennifer’s not the only one watching. The whole board wants to see results, not just plans.”
“Then I’ll give them results.”
David finally moved toward his own car, parked three spaces down. “Keep me informed. Not the sanitized version. I need to know if we’re getting stuck.”
“You’ll know,” Marcus said.
He watched David drive away, then sat in his own car for a long moment before starting the engine. The parking lot was emptying, the building’s glass facade catching the late afternoon sun.
“How hard could this be?” he thought. And even as he thought it, he recognized the arrogance in the question—but it wasn’t really arrogance. It was pattern recognition. Every technology transformation started with uncertainty and ended with integration. The middle part was just problem-solving.
He pulled out of the lot, already making mental lists. GPU instances to provision. Engineers to reallocate. Documentation to review. The familiar rhythm of turning mandate into execution.
That evening, Marcus sat in his home office with his laptop open and a beer he’d forgotten to drink. His wife had long since gone to bed, understanding without needing explanation that some nights the work followed him home.
He had been reading for three hours. OpenAI’s documentation. Hugging Face tutorials. Enterprise case studies with titles like “How Company X Achieved Y% Improvement.” The more he read, the more he realized how much he didn’t know.
Transformers. Attention mechanisms. Fine-tuning versus RAG. The terminology was foreign in a way that Kubernetes and microservices had never been. Those technologies had evolved from things he understood. These felt like they had arrived from somewhere else entirely.
But they also felt learnable—that was the thing. The concepts were complex but not impenetrable. Given time and focus, he could understand them. Given a team willing to experiment, they could build something.
He made notes in the margins of a printed paper: “three use cases = faster learning,” “start with contained scope,” “measure everything.”
Around midnight, he closed the laptop. The beer was warm now, untouched. Through the window, the suburban darkness was quiet, the kind of stillness that made problems feel more manageable.
He thought about Jennifer’s warning. Resource fragmentation. Good CTOs stretched too thin. The math of attention.
But that was exactly why he would succeed where others had failed. He understood the math. He would carve out the time, protect the resources, maintain the focus. Other initiatives had demanded the same discipline. This was no different.
He went to bed convinced he understood what was ahead.
Monday morning. The team meeting.
Marcus had reserved the large conference room, the one with the whiteboard wall and the standing-height table that made meetings feel more energetic. His infrastructure lead, two senior engineers, the head of data engineering, and the customer service manager had all gathered with the slightly confused expressions of people called to an unscheduled meeting.
“I’ll keep this brief,” Marcus said. “The board has approved investment in AI initiatives. Real investment, with real resources and real expectations.”
He watched the reactions ripple through the room: interest, curiosity, and from the engineers, something that looked like excitement barely contained.
“Three pilot projects,” he continued. “Customer service chatbot, predictive maintenance analysis, and procurement optimization. Daniel, I want you leading the technical implementation. Priya, you’ll need to assess data readiness. Sarah, security considerations from day one.”
Daniel Park, the Python developer who had been reading AI papers on his lunch breaks for months, looked like he had just been handed a gift. “What’s our timeline?”
“Six months to pilots, twelve to production. Aggressive but achievable.”
“Budget?” Sarah Martinez asked. She was the security engineer, the one who always asked about constraints before getting excited about possibilities.
“$250K to start, with room to grow if we show results.”
The room buzzed with the particular energy of people who suddenly had permission to build something new. Questions flew: What models would they use? Could they access GPU clusters? Had Marcus seen the latest paper on retrieval-augmented generation?
Marcus answered what he could, deferred what he couldn’t, and let the momentum build. This was the part he was good at. Translating board mandates into team energy. Channeling excitement toward execution.
“We’ll meet weekly to track progress,” he said as the meeting wound down. “I want demos, not just updates. Show me what works and what doesn’t.”
The team filtered out, already clustering into smaller conversations about architecture and approach. Marcus hung back, straightening chairs that didn’t need straightening, feeling the particular satisfaction of potential unleashed.
Daniel caught him at the door. “Thanks for this,” he said. “I’ve been wanting to work on something like this for months.”
“Then make it count,” Marcus said. “You’ve got the skills. Now you’ve got the resources. Let’s see what you can build.”
Daniel nodded, something between gratitude and determination in his expression, and headed toward his desk.
Marcus watched him go, then started toward his own office. The morning felt promising. The team was energized. The direction was clear.
To be continued…
What happens next: Early momentum builds as the team makes real progress—an 87% accurate chatbot, excited leadership demos, genuine technological achievements. But watch for the warning signs that Marcus is too busy to notice. Chapter 2 reveals how success can be just as dangerous as failure.
Why we wrote this
Scott Weiner is the AI Lead at NeuEon, Inc., where he helps organizations navigate the complexities of AI adoption and digital transformation. This story draws from patterns observed across dozens of enterprise AI initiatives.
Erwann Couesbot is the CEO of FlipThrough.ai, specializing in AI strategy for professional services. His conversations with technology leaders and his personal experiences inspired many of the dynamics explored in this narrative.
Want to read the complete story?
Have your own AI transformation story? We’d love to hear it. Connect with Scott on LinkedIn or reach out to NeuEon at neueon.com/contact.