58 min read
Reader Signal
Lower current readership momentum right now.
EDWARD
Thursday, May 1, 2025 – 5:30 AM – The Dock, Edward’s House on the Intracoastal
The dock was quiet at dawn.
Edward sat at the edge, fishing pole balanced loosely in his hands, watching the Intracoastal turn from black to gray to silver as the sun climbed over the horizon. He’d been coming here every morning since spring break, since the conversation with Marcus, since Sophia’s stories about AI and hackathons.
Five weeks now. Over thirty mornings of watching the water and wrestling with one phrase that wouldn’t leave his head: I won’t care if I’m irrelevant in 2028. I’ll be rich.
Marcus had laughed when he said it. Brushed off Edward’s story about Harold, the tape library guy from his first job at a defense contractor, like it was ancient history. But Edward couldn’t stop thinking about Harold. The expert who became obsolete overnight. The guy who went from essential to ornamental because he waited to see what happened instead of figuring it out.
Edward had always been the cautious one. Marcus used to joke about it at the Academy—“Johnson doesn’t take a bathroom break without a contingency plan.” But caution had kept Edward alive in the Persian Gulf. Two years on the Boutwell, running maritime security operations while Jennifer taught English in Istanbul—close enough to meet him whenever he got leave. Stolen weekends in Cyprus, Athens, wherever his ship docked. Two years of watching the horizon, checking the radar, anticipating threats before they materialized. You didn’t survive that by being reckless.
When he finally got stateside—New York, then the MS at NYU, then the defense contractor job in Virginia—he’d brought that caution with him. Northrop Grumman was perfect for someone who’d learned to triple-check everything. Regulated software. Compliance reviews. Documentation for documentation. The kind of work where careful was a feature, not a bug.
But caution could curdle into paralysis. He’d seen it happen to Harold. He’d seen it happen to entire companies. The organizations that were so careful about avoiding the wrong decision that they never made the right one.
The calculus had changed. That’s what kept gnawing at him.
He’d spent twenty years building expertise. Twenty years of learning systems, understanding architectures, developing judgment about what worked and what didn’t. That expertise had made him valuable, valuable enough to become CTO of a public company.
But expertise had a half-life now. He could see it in the quarterly reports from Gartner and Forrester. He could see it in the way junior engineers were getting comparable results in hours instead of days. The productivity multiplier was real, and it was compounding.
If Edward didn’t figure this out, his board would eventually bring in someone who would. Maybe not this year. Maybe not next year. But by 2028? His experience might be as relevant as Harold’s tape schedules.
Edward didn’t resent the change; resenting market forces was a waste of energy. But he needed to understand the new math. If a junior engineer with the right tools could match a senior engineer’s output, what happened to the value of seniority? If institutional knowledge could be captured in AI systems, what happened to the premium companies paid for experience?
The board wouldn’t ask these questions directly. They’d just start wondering if maybe a younger CTO, one who’d grown up with these tools, might be a better fit for the next phase.
He was forty-three years old. Sophia was eighteen now, a freshman at Georgia Tech. Emma was fifteen, three years from college. Lily was eleven, still young enough to think her dad knew everything.
Twenty years of fatherhood. Twenty years of watching himself become someone he hadn’t expected to be.
He’d missed Sophia’s entire pregnancy. Four months on a ship in the Persian Gulf while Jennifer grew their daughter inside her. He’d gotten emergency leave once—seventy-two hours, eighteen of them spent in the air each way. One night to hold his pregnant wife, to feel the baby kick against his palm, to whisper promises to a belly he wouldn’t see again until it was empty. Then back on a plane, back to the Gulf, back to watching the horizon while Jennifer’s mother drove her to doctor’s appointments.
When he finally came home, Jennifer was seven months pregnant and exhausted. They got married two weeks later—small ceremony, just family, Jennifer in a dress that accommodated the belly he’d missed watching grow. Sophia arrived in August. Edward held his daughter for the first time and his hands wouldn’t stop shaking.
He’d made a promise that day. Never again. Never miss the moments that mattered. Never let the job take him away from the people who needed him.
Emma came three years later, when they were settled in Virginia. Edward was there for every appointment, every ultrasound, every 3 AM craving run to the 7-Eleven. He held Jennifer’s hand in the delivery room and cried when Emma screamed her first breath. He hadn’t cried when Sophia was born. He’d been too overwhelmed, too shell-shocked, too aware of everything he’d missed.
Lily came four years after that. By then, Edward knew who he was. Not the young officer watching horizons for threats. Not the anxious new father trying to make up for lost time. Just a dad. A husband. A man who’d figured out that the job was supposed to serve the family, not the other way around.
Forty-three years old. The age his father had been when Edward graduated from the Academy. His father had seemed so old then, so settled, so certain about everything. Edward didn’t feel certain about anything. He felt like he was still figuring it out, still learning, still trying to become the man his daughters thought he already was.
Jennifer had given up her career to raise them. She’d been teaching English in Istanbul when they met, twenty-two years old with plans for graduate school, maybe international development, maybe something that used her languages. Instead, she’d followed him to New York, then Virginia, then Florida. Three kids in eight years, and by the time Lily started kindergarten, Jennifer’s career had become motherhood. She’d made that choice willingly, but it meant everything rode on Edward’s salary. His equity. His relevance.
The house they’d bought eight years ago, when he’d made VP of Engineering: back then, waterfront on the Intracoastal was attainable, not aspirational. Now the insurance had tripled. Hurricane premiums, flood insurance, windstorm coverage. Every year Jennifer would show him the renewal notice and they’d have the same conversation: Should we move?
If he lost this job, they’d have to sell. Twenty more years on the mortgage. No safety net. Three daughters who were counting on him to figure it out, the same way he’d counted on his own father to figure it out, back when he thought forty-three was old.
But there was something else gnawing at him. Something more immediate than career anxiety.
The medication tracking feature.
It had been in the backlog for eighteen months. Hospitals kept asking for it, a way to flag potential drug interactions before discharge. The CDC estimated 7,000 to 9,000 patients died annually from medication errors. This feature could catch some of those. It could save lives.
And it sat there. Stuck. Because the engineering team was buried in maintenance work, code reviews, and technical debt. Because every time someone tried to prioritize it, something else came up. Because the process was designed to produce activity, not outcomes.
That’s what kept Edward up at night. Not “AI strategy.” The fact that his company was too slow to ship the things that mattered.
Monday, May 5, 2025 – 9:00 AM – Edward’s Office, Riverton Health Systems, Brickell
Rachel Torres was waiting in his office when he arrived. She looked worried.
Rachel had been Edward’s chief of staff for seven years now. She’d started working for him right out of college, when he was still at his previous company. Eager, sharp, with a computer science degree she wasn’t sure she wanted to use for coding. Edward had seen something in her, an ability to translate between technical and business that most engineers lacked. When he made CTO at Riverton, she made the move with him. She knew how he thought, anticipated what he needed, and wasn’t afraid to tell him when he was wrong.
“We have a problem.”
Edward set down his bag. “What kind of problem?”
“Engineers are using AI tools. On their own. Without approval.”
He sat down slowly. “What do you mean?”
“I mean they’re copying code into ChatGPT. Using Copilot with personal accounts. Pasting patient data schemas into AI chatbots to get help debugging.” Rachel handed him a printout. “I found this in a Slack channel. An engineer was celebrating that he’d ‘saved four hours’ by having AI write his unit tests.”
Edward stared at the printout. His first instinct was alarm. HIPAA violations, IP exposure, security risks. His second instinct was to ban everything immediately.
But then he stopped.
“If they’re doing this despite the risk,” he said slowly, “there’s a need we’re not meeting.”
“That’s… not what I expected you to say.”
“Me neither.” Edward leaned back. “But think about it. Smart people. They know the rules. They’re doing it anyway.”
“So what do we do?”
“Figure out why they’re in pain.”
Tuesday, May 6, 2025 – 2:00 PM – Conference Room B, Riverton Health Systems
The consultants arrived at 2 PM sharp. Three of them from Stratton McKelvey, dark suits, leather portfolios, confident handshakes. Edward had brought them in because the board kept asking about “AI strategy” and he needed to understand what the market was recommending.
“Thank you for making time,” the lead consultant said, pulling up a slide deck. “What we’re proposing is a comprehensive AI SDLC transformation roadmap. Twenty-four to forty-eight months, depending on organizational readiness.”
Edward nodded. “Walk me through it.”
“Phase one is assessment. We’ll evaluate your current state across twelve dimensions of AI SDLC maturity: data infrastructure, talent readiness, governance frameworks, change management capacity. That’s a ninety-day engagement.”
“Ninety days just to assess?”
“Transformation requires understanding before action. You can’t build a house without blueprints.”
The slides kept coming. Phase two: pilot selection. Phase three: Center of Excellence establishment. Phase four: scaled rollout with governance oversight. Phase five: continuous improvement and maturity advancement.
“What’s the timeline to ROI?” Edward asked.
The consultant checked his notes. “Assuming no major blockers, you’d see measurable returns by Q1 2027. That’s aggressive but achievable.”
Edward did the math. Two years minimum. Maybe four. Before they saw any return at all.
“I have a feature that saves lives sitting in my backlog right now,” Edward said. “The medication tracking feature. It’s been stuck for eighteen months. Hospitals are asking for it. Patients could be dying because we can’t ship fast enough.”
“That’s exactly why you need a transformation roadmap. The feature is stuck because your organization isn’t ready. “
“What if I told you I don’t have two years?”
The consultant smiled patiently. “Transformation takes time. You can’t rush organizational change. Companies that try to move too fast end up with failed implementations, change fatigue, and worse outcomes than if they’d never started.”
Edward leaned back. A question had been forming in his mind, one he wasn’t sure he wanted to ask. But he asked it anyway.
“Do any of you build software with AI tools?”
The room went quiet.
The lead consultant glanced at his colleagues. Neither spoke.
“I’m not asking about your clients,” Edward pressed. “I’m asking about you. Personally. Do you use Copilot? Cursor? ChatGPT? When you write code, if you write code, do you use these tools?”
More silence.
“We’re consultants,” the lead finally said. “Our job is to help organizations adopt. “
“You’re selling a forty-eight-month roadmap for tools you’ve never used.” Edward stood up slowly. “You’re telling me how to transform my engineering organization, and none of you have ever built anything with these tools yourselves.”
“With respect, implementation experience isn’t necessary for strategic advisory. “
“It’s the only thing that matters.” Edward walked to the window, looking out at Biscayne Bay. “I’ve spent twenty years in this industry. I’ve seen a dozen transformation initiatives. The consultants always have the same pitch: assess, plan, pilot, scale. Twenty-four to forty-eight months. And 70% of them fail. You know why?”
Nobody answered.
“Because the people designing the transformation have never done the work.” He turned back to face them. “Thank you for your time. We won’t be moving forward.”
The consultants packed up their leather portfolios with the stiff politeness of people who’d just been fired. Edward walked them to the elevator, shook hands, and watched the doors close.
Rachel was waiting when he got back to the conference room.
“You just fired Stratton McKelvey.”
“Saved us two million and two years of nothing.”
“The board wanted an AI SDLC strategy.”
“The board wants results.” Edward sat down. “Those guys can’t teach what they’ve never learned. I need to figure this out myself.”
“You’re going to learn to code with AI?”
“I’m going to learn to build again. Haven’t written production code in eight years.” He shook his head. “Maybe that’s the problem.”
“So what’s the plan?”
“First, I learn. Actually use the tools. Then we map the software process, the real one, not the documentation. Find out where the medication feature is stuck.” He paused. “Then we fix it.”
“That’s not an AI SDLC strategy.”
“No. It’s process improvement. AI might help. Might not. Won’t know until I try.”
Tuesday, May 6, 2025 – 4:00 PM – Edward’s Office, Riverton Health Systems
After the consultants left, Edward tried the vendors.
Over the next two weeks, he sat through eleven demos. Established players. Hot startups. Enterprise giants rebranding legacy products as “AI-powered.”
The pattern was always the same. Slick slides. Impressive demos. Metrics that measured tool activity—lines generated, keystrokes saved, task completion time—but nothing about what the business actually needed. Edward kept asking the same question: Can this help me ship the medication tracking feature? The one that’s been stuck for eighteen months while hospitals keep asking for it?
Not one of them could answer it. They had demos and case studies, but they couldn’t connect any of it to his actual problem.
One sales engineer was honest enough to admit it. “Look,” she said after Edward’s third pointed question, “the truth is, we’re all figuring this out. The technology is moving faster than anyone’s ability to implement it. Most of our case studies are six months old, which means they’re basically ancient history in AI terms.”
Edward appreciated the honesty. He didn’t buy the product.
He tried one more avenue. Tom Hirsch, an old colleague with a boutique consulting firm, sent over a team. Edward recognized two of them from a previous DevOps engagement. They had new AI certifications, but when Edward asked what they’d actually built, the answer was roadmaps and readiness assessments. None of them had written code with AI tools.
He’d have to figure this out on his own.
Thursday, May 22, 2025 – 2:17 AM – Master Bedroom, Edward’s House on the Intracoastal
Edward couldn’t sleep.
He’d been staring at the ceiling for two hours, running the same calculation over and over. He’d fired Stratton McKelvey. Dismissed Tom’s team. Rejected eleven vendors. The board was expecting an AI SDLC strategy, and he had nothing to show them except a list of people he’d decided weren’t good enough.
What if I’m wrong?
The thought kept circling back. What if the consultants knew something he didn’t? What if the 24-month roadmap was actually the right approach, and his impatience was going to cost the company everything? What if his instinct to “figure it out myself” was just arrogance dressed up as initiative?
He slipped out of bed, careful not to wake Jennifer, and went downstairs to his home office. The laptop was still open from the evening, email drafts unsent.
He started typing.
Dear Michael,
I’ve reconsidered our conversation from Thursday. While I still have concerns about the timeline, I recognize that Stratton McKelvey has transformation experience that Riverton lacks. I’d like to discuss resuming our engagement, perhaps with a modified scope that focuses on…
Edward stopped typing. Read it back. His finger hovered over the trackpad.
Send it. Just send it. Let the professionals handle this. You’re a CTO, not a transformation consultant. You don’t know what you’re doing.
The cursor blinked.
But they don’t know either. They’ve never built anything. They’re selling confidence, not competence.
Maybe confidence is what the board needs. Maybe looking like you have a plan is more important than actually having one.
He stared at the draft for a long time. The house was silent. Through the window, he could see the dock, the water black and still beyond it.
Jennifer’s voice came from the doorway. “You’re up.”
“Couldn’t sleep.”
She walked over, wearing the old Coast Guard sweatshirt she’d stolen from him in 2006, right before Sophia was born. She’d been wearing it for twenty years now. He’d stopped asking for it back around year three.
She looked at the screen. Read the email. Didn’t say anything for a moment.
“You’re going to call them back.”
“I’m thinking about it.”
“Because you’re scared.”
“Because maybe they know something I don’t.”
Jennifer sat on the arm of his chair. The same way she’d sat when he told her about Iraq. When he told her about the job offer in Florida. When he told her Sophia was switching majors. She’d been sitting on the arm of his chair for twenty years, and he’d been making better decisions because of it.
“Ed, I’ve watched you for twenty years. I know when you know something’s wrong. You knew the consultants were wrong the moment they said ‘forty-eight months.’ You knew it in your gut.”
“My gut isn’t a strategy.”
“Neither is hiring people you don’t believe in.” She put her hand on his shoulder. “What are you actually afraid of?”
Edward was quiet. The answer came slowly.
“That I’ll fail. That I’ll try to figure this out myself, and I’ll get it wrong, and the board will fire me, and we’ll lose the house, and Emma won’t go to the college she wants, and it’ll all be because I was too proud to admit I needed help.”
“That’s a lot of catastrophizing for 2 AM.”
“It’s 2 AM. That’s what 2 AM is for.”
Jennifer smiled slightly. “Delete the email.”
“What if I’m wrong?”
“Then you’ll be wrong. And you’ll learn something. And you’ll try again.” She stood up. “But you won’t learn anything by handing this to people who’ve never done it themselves. You said it yourself: they can’t teach what they’ve never learned.”
Edward looked at the email. At the cursor still blinking.
He highlighted the text. All of it. And pressed delete.
“Come back to bed,” Jennifer said.
“In a minute.”
She paused at the door. “You remember what you told me when you left the Coast Guard? When you took the Northrop job?”
“That I was ready to start our real life.”
“You said you’d spent five years learning to watch for threats. And now you wanted to learn to build something.” She smiled. “You’ve been building things ever since. The defense contractor, the healthcare work, Riverton. You know how to build, Ed. Trust that.”
She left. Edward sat alone in the dark office, laptop glowing, feeling the weight of what he’d just decided. No consultants. No vendors. No safety net.
Just the questions he didn’t know how to answer yet.
But Jennifer was right. He’d spent five years in the Coast Guard learning to navigate uncertainty. Two years in a war zone learning to make decisions with incomplete information. Fifteen years in corporate America learning to build systems that worked.
Maybe that was enough. Maybe the consultants’ confidence was just a different kind of fear—fear of admitting they didn’t know, dressed up as a process.
He closed the laptop and went back upstairs. Sleep came eventually, fitful and thin.
Saturday, May 24, 2025 – 5:00 AM – Home Office, Edward’s House on the Intracoastal
Edward sat at his desk with his laptop, a cup of coffee growing cold beside him. The sun wouldn’t be up for another hour. Through the window, he could see the dock, the water black and still.
He’d downloaded Claude Code, Copilot, and Cursor the night before. Now he was staring at a blank screen, trying to remember the last time he’d written code that wasn’t a quick script or a proof of concept.
Eight years. Maybe nine.
He started small. A simple Python script to parse some log files, something he’d asked an engineer to do the month before. It took the engineer two days. Edward wanted to see what happened with AI.
Three hours later, he had a working script. Not just working, elegant. The AI had suggested patterns he wouldn’t have thought of, caught edge cases he’d missed, refactored his clunky first attempt into something clean.
But that wasn’t what surprised him.
What surprised him was how it felt.
For eight years, coding had been something other people did. Edward reviewed code, approved architectures, made decisions about technical direction. But he didn’t build anymore. He’d told himself that was fine; his job was leadership, not implementation.
Now, sitting on the dock with the sky turning pink over the Intracoastal, he picked up the laptop again. Opened a new file. Started typing. He didn’t stop until the sky was dark.
Saturday, June 7, 2025 – 3:00 PM – The Dock, Edward’s House on the Intracoastal
Sophia was home for spring break. She’d arrived two days ago, a week off from Georgia Tech before the final push to finals. Edward found her on the dock that Saturday afternoon, laptop open, feet dangling over the water.
“What are you working on?”
“Side project.” She didn’t look up. “Building an AI agent that can navigate codebases. You feed it a repo and it learns the architecture, then you can ask it questions.”
Edward sat down beside her. The dock creaked the way it always did.
“I wrote a Python script this morning,” he said. “Log parser. First real code I’ve written in eight years.”
Now she looked up. “Wait, seriously?”
“It was…” He searched for the right word. “Fun.”
Sophia’s face lit up in a way that reminded him of her at six years old, showing him a drawing she was proud of. “Dad. That’s amazing. Show me.”
He pulled out his laptop. For the next hour, they sat on the dock together, two generations of Johnsons staring at screens, the Intracoastal lapping beneath them. Sophia walked him through her workflow. Claude for architecture, Copilot for implementation, ChatGPT for debugging. Prompt patterns. Context windows. How you had to think with the AI, not just at it.
“People treat it like a magic box,” she said. “Type in a question, get an answer. But it’s more like a really smart collaborator who doesn’t know your codebase yet.”
“Like onboarding a new engineer.”
“Exactly! Except this one works at 3 AM and never gets tired.”
They pair programmed for three hours. Sophia showed him how she’d automated her entire testing workflow. “I haven’t written a manual test in six months. The AI generates them from my code. Better coverage than anything I could write by hand.”
She showed him the prototype generator one of her classmates had built. “James made this. You give it a product requirements document and it generates a working prototype. Not production-ready, but 80% of the way there in an hour instead of a week.”
Edward asked about failure modes. The hallucinations. The context limits. The times the AI confidently produced garbage. Sophia was honest about the limitations.
“It’s not magic,” she said. “It’s a tool. A really good tool. But you still have to know what you’re building.”
What struck Edward wasn’t the technology. It was his daughter’s attitude.
She didn’t think of AI as a threat or a disruption or a transformation challenge. She thought of it as a tool, like a really good IDE or a smart pair programmer. She wasn’t afraid of it. She wasn’t resistant to it. She just used it, the same way she used Git or developer forums or a search engine.
“How do you stay current?” Edward asked. “The tools change so fast.”
Sophia laughed. “I don’t try to stay current on tools. I stay current on outcomes. What do I need to build? What’s the fastest way to build it? The tools are just means to an end.” She shrugged. “My professor keeps saying the same thing. ‘Don’t fall in love with the hammer. Fall in love with the house you’re building.’“
Edward thought about the consultants he’d fired. They’d been obsessed with tools. AI Readiness Assessments, governance frameworks, maturity models. His nineteen-year-old daughter didn’t care about any of that. She just built things.
Jennifer came out around six with sandwiches and lemonade. She found them still on the dock, still coding, still talking.
“I don’t think I’ve ever seen you two like this,” she said, setting down the tray.
Edward looked at Sophia, then at his laptop, then at the water. “I don’t think I’ve ever been like this.”
“Like what?”
“Learning from my kid.” He smiled. “It’s humbling. And kind of wonderful.”
Sophia rolled her eyes, but she was smiling too. “Dad, you’re being weird.”
“I’m being honest.” He picked up a sandwich. “The consultants I fired—they had decades of experience. Certifications. Frameworks. Roadmaps. They couldn’t answer simple questions about what actually works. You answered them in three hours.”
“That’s because they’ve never built anything with these tools,” Sophia said. “They’re selling expertise they don’t have.”
Edward stopped chewing. That was it. That was exactly it.
The transformation industry was selling expertise it didn’t have. The real expertise was in dorm rooms and on docks, built by people too young to know they were supposed to wait for permission.
That evening, after Sophia went inside and Jennifer was cleaning up dinner, Edward sat alone on the dock with his notebook.
Stop asking “how do we adopt AI?” Start asking “what do we need to build?”
The tools are changing too fast to standardize. Standardize on outcomes instead.
Sophia isn’t afraid because she’s never known anything else. My engineers are afraid because they remember the before times. Different problem, different solution.
Sophia’s testing workflow: AI generates tests from code. Why aren’t we doing this?
James’s prototype generator: 80% in an hour vs 100% in a week. When is 80% good enough?
He looked at the water, then at the house, then at the window where he could see Sophia’s silhouette moving around her room.
The consultants had wanted two years and three million dollars to tell him how to think about AI. His daughter had done it for free in an afternoon. Because she wasn’t trying to sell him anything. She was just excited about what she was building.
For the rest of spring break, they built together. Every morning on the dock, laptops open, coffee cooling beside them. Sophia showed him her workflows; he showed her how enterprise systems actually worked. She’d never seen a codebase with fifteen years of technical debt. He’d never seen someone build a working prototype in an afternoon.
Edward rebuilt an internal dashboard that had been on the backlog for six months. He prototyped a new API endpoint that his team had estimated at three weeks. He did it in four days. When he got stuck, Sophia unstuck him. When she got stuck, he reminded her to step back and think about the problem differently.
It was the best week he’d had in years.
When Sophia flew back to Atlanta on Sunday, Edward kept building. Small projects at first: utilities, prototypes, experiments. He’d text her screenshots, ask questions, get back emoji reactions and quick voice memos explaining what he’d done wrong. The mentorship had reversed—his nineteen-year-old daughter teaching him to code again.
And somewhere in those weeks, something clicked.
If the tools were this good now, in early 2025, what would they be like in 2028? The improvement curve wasn’t linear. Every few months, the models got significantly better. The things that felt like magic today would be table stakes in three years.
He thought about Harold again. The tape library guy. Harold had been an expert in systems that stopped mattering. He’d spent his career building competency in something that became obsolete.
Edward couldn’t let that happen to Riverton. He couldn’t march his organization toward today’s standards; by the time they got there, the standards would have moved. It would be like planning a trip to Mars by aiming at where Mars was today, instead of where it would be when you arrived.
He needed to build the organization for 2028, not 2025.
That meant learning first, then scaling. You couldn’t teach what you’d never learned. You couldn’t lead a transformation you didn’t understand. The consultants had failed because they’d never done the work. Edward refused to make the same mistake.
Saturday, June 21, 2025 – 2:00 PM – Klaus Advanced Computing Building, Georgia Tech Campus, Atlanta
The invitation came through Sophia.
“Dad, this is kind of embarrassing, but my computer club figured out who you are.”
She’d called him the week after spring break, sounding mortified.
“What do you mean, figured out who I am?”
“One of our TAs was looking for entry-level jobs. She saw Riverton’s posting online and noticed you were in my connections. Same last name.” Sophia groaned. “Now they want you to come speak at a meeting. About AI transformation in industry.”
“Your computer club wants to hear from me?”
“They want to hear from anyone in industry. Last month they had a startup founder. Month before that, an engineer from a tech giant.” She paused. “You don’t have to do it. I know it’s weird.”
Edward thought about the consultants he’d fired. Thought about how disconnected they’d been from actual building. Thought about how much he’d learned from Sophia over spring break.
“I’ll do it,” he said. “But I want something in return.”
“What?”
“I want to learn from them too. Not a lecture—a conversation. I want to know what they’re building, how they’re using AI, what works and what doesn’t.”
The computer club meeting was in a cramped room in the Klaus Advanced Computing Building. About thirty students, mostly sophomores and juniors, some grad students in the back. Laptops everywhere. Energy drinks. The particular intensity of people who loved what they were doing.
Sophia had met him at the airport that morning, clearly nervous about her dad invading her college world. But when Edward walked into the room and saw the whiteboards covered in architecture diagrams, the monitors showing half-finished projects, the energy of people building things—he felt at home.
Edward had prepared remarks. He threw them out in the first five minutes.
“I’m supposed to tell you about AI transformation in enterprise healthcare IT,” he said. “But honestly? I’m here to learn from you. I’ve spent the past two months trying to figure out how to help my organization adopt AI. The consultants I hired couldn’t answer basic questions. So I want to ask you instead.”
The room shifted. Students sat up. This wasn’t what they’d expected.
“What are you actually building?” Edward asked. “Show me.”
For the next two hours, they showed him.
A junior named Cora was building an AI agent that could navigate codebases and answer questions about architecture. “I’m training it on open source projects,” she said, pulling up her screen on the projector. “It already knows React better than most senior developers.”
A senior named Priya—the one who’d noticed the online connection—had automated her entire testing workflow. “I haven’t written a manual test in six months. The AI generates them from my code. Better coverage than anything I could write by hand.”
Another student named James was building something that made Edward’s stomach drop: an AI system that could take a product requirements document and generate a working prototype. “It’s not production-ready,” James admitted. “But it gets you 80% of the way there in an hour instead of a week.”
Edward asked about workflows, about prompts, about failure modes. He asked about what didn’t work: the hallucinations, the context limits, the times the AI confidently produced garbage. The students were honest about the limitations.
At the end of the meeting, Priya raised her hand. “Mr. Johnson, can I ask you something?”
“Of course.”
“Why did you really come here? CTOs don’t usually fly to Atlanta to ask college students for advice.”
Edward thought about it. “Because the consultants I hired have never built anything with these tools. They have certifications and frameworks and roadmaps. But they couldn’t answer simple questions about what works and what doesn’t.” He looked around the room. “You can. That’s why I’m here.”
The room was quiet. Then Priya smiled. “That’s the most honest thing an industry speaker has ever said to us.”
On the flight back to Miami, Edward made notes.
The irony wasn’t lost on him. He’d met with four consulting firms. Stratton McKelvey. Chicago Consulting Group. Detroit DevOps. A boutique AI consultancy with a slick website. Combined, they’d probably had two hundred years of enterprise experience. And not one of them could show him a working demo of something they’d built with AI.
These students, some of them twenty years old, had more hands-on experience with AI development than all those consultants combined. Priya’s testing automation alone was more sophisticated than anything the consultants had described. James’s prototype generator would have been a six-figure consulting deliverable, and he’d built it as a side project.
The transformation industry was selling expertise it didn’t have. The expertise was in dorm rooms and computer clubs, built by people too young to know they were supposed to wait for permission.
Friday, June 27, 2025 – 9:00 AM – Large Conference Room, Riverton Health Systems
Edward reserved the large conference room, the one with the windows facing Biscayne Bay and the whiteboard wall that stretched twenty feet.
He’d spent five weeks learning to build again—first on his own, then with Sophia over spring break, then flying to Atlanta to learn from her computer club. Now he understood, viscerally not just theoretically, what the tools could do. He’d felt the joy of coding come back. He’d experienced the productivity gains firsthand. And he’d realized that aiming for 2025 standards was already aiming too low.
Now he needed to understand what was stopping the organization from shipping.
He’d invited six engineers from different teams. People he trusted. People who would tell him the truth.
“Here’s what I want to understand,” he said, standing at the whiteboard. “The medication tracking feature. It’s been in the backlog for eighteen months. Walk me through what actually happens when a feature like this moves through our system. Not the official process, the reality. Where does it wait? Where does it get stuck?”
For the first hour, they talked. Edward wrote. Rachel asked questions.
But something became clear quickly: the engineers struggled to articulate what they actually did.
“And after the pull request is opened, what happens?” Edward asked.
“It goes into review.”
“What does ‘goes into review’ mean? What are the actual steps?”
Long pause. Shrugging.
“I mean, Kevin looks at it. He does his thing. Approves it or sends it back.”
“What does Kevin look for? What’s his process?”
“I don’t know. He just… reviews it.”
A junior engineer named Derek spoke up. “The queue isn’t a real blocker. Kevin’s fast. A day, maybe two.”
Edward wrote it on the whiteboard. “One to two days. Okay. What happens before the PR gets to Kevin?”
“Development.”
“How long?”
“Depends on the feature. Two weeks, usually.”
Edward wrote that down. “And before development?”
Silence. The engineers looked at each other.
“Before development,” Edward repeated. “What happens after a feature gets assigned but before anyone writes code?”
“Environment setup,” someone said.
“How long?”
“A week? Sometimes two?”
“Why?”
More silence. Then a woman named Maria, a ten-year veteran, leaned forward. “Because the infrastructure team is backlogged. They provision environments manually. If you submit a request on Monday, you might get access by Friday. If you submit on Friday, you wait until the following Thursday.”
“So a feature can sit for two weeks before anyone writes a line of code?”
“Sometimes longer.”
The engineers leaned forward. They could see it now, the picture emerging from fragments they’d never assembled.
“What about before that? Before environment setup?”
“Design review.”
“How long?”
“Three weeks.”
“Three weeks?” Edward stopped writing. “For design review?”
“Product has to sign off. And they meet weekly. So if you miss the Thursday meeting, you wait until the next Thursday.”
“And if there’s feedback?”
“You wait another week.”
The room was getting quieter. Edward kept writing. The whiteboard was filling up with boxes and arrows and wait times that nobody had ever seen collected in one place.
“What about after code review? After Kevin approves?”
“QA.”
“How long?”
“Two weeks for test script writing. Another week to run the tests.”
“Why two weeks for test scripts?”
“Because we write them manually. From scratch. Every time.”
Edward stepped back from the whiteboard. The diagram covered the entire surface. Twelve boxes. Dozens of arrows. Wait times adding up to something nobody wanted to say out loud.
“I’m going to add up the numbers,” he said. “Tell me if I get it wrong.”
He worked through the diagram, writing totals next to each stage.
The room was silent.
Edward realized he was seeing something important. These were good engineers, smart, experienced, capable. But they’d been doing the same work for so long that the knowledge had become tacit. Unconscious. They couldn’t externalize what they knew because they’d never had to.
It was the same pattern he saw with onboarding. When new engineers joined, the veterans would say “just shadow Kevin for a week” because they couldn’t actually explain what Kevin did. The expertise existed, but it was locked in people’s heads. impossible to transfer, impossible to scale, impossible to improve because no one could see it clearly enough.
And now AI tools were asking engineers to do exactly what they’d never learned to do: externalize their context. Tell the AI what you’re trying to accomplish. Explain the constraints. Describe the codebase. But if they couldn’t explain it to a new hire, how would they explain it to an AI?
“Let me try something different,” Edward said. “Instead of telling me what happens, show me. Walk me through the last feature you shipped. Pull up the PRs, the comments, the tickets. Let’s trace the actual path.”
That worked better. With artifacts to anchor them, the engineers could reconstruct what had happened. Rachel took notes. Edward drew diagrams.
By noon, they had a map. Twelve stages from requirement to deployment:
| Stage | Time | What Happens |
|---|---|---|
| Requirements to Design | 3 weeks | Waiting for product sign-off |
| Design to Development | 2 weeks | Waiting for environment provisioning |
| Development | 2 weeks | Actual coding |
| Development to Code Review | 4 days | Waiting in Kevin Nakamura’s queue |
| Code Review | 2 days | Kevin reviews |
| Code Review to QA | 2 weeks | Writing manual test scripts |
| QA Testing | 1 week | Running tests |
| QA to Compliance | 3 weeks | HIPAA checklist review |
| Compliance to Security | 1 week | Security sign-off |
| Security to Deployment | 1 week | Waiting in approval queue |
Edward stepped back from the whiteboard.
“Twelve weeks,” he said. “Minimum. To ship a single feature.”
“Wait.” Priya Sharma raised her hand. “We’re missing something. Before requirements even get to design, there’s another step.”
“What step?”
“The agile coaches have to approve the acceptance criteria. Before any story gets prioritized into a sprint, Mike Donovan’s team reviews it. They meet on Fridays.”
Edward felt his eye twitch. “The agile coaches approve acceptance criteria?”
“They say it’s to ensure ‘definition of ready.’ Stories can’t enter the sprint backlog until the acceptance criteria meet their standards.”
“What standards?”
Priya shrugged. “Nobody really knows. Sometimes they approve everything. Sometimes they send stories back with comments about ‘user value statements’ or ‘testability concerns.’ One time they rejected a story because it didn’t have enough ‘acceptance test scenarios.’“
“And if you miss the Friday meeting?”
“You wait until the next Friday.”
Edward added a line to the top of the chart:
| Stage | Time | What Happens |
|---|---|---|
| Story to Prioritization | 1-2 weeks | Waiting for Friday agile coach approval |
“So a story can sit for up to two weeks before it even enters the prioritization queue?”
“Depending on when it’s submitted. If you submit on a Thursday, you might get approved Friday. If you submit on a Saturday, you wait twelve days.”
Rachel was shaking her head. “I’ve been here four years. I didn’t know this step existed.”
“Neither did I,” Edward said. “And I’m the CTO.” He turned to the engineers. “How long has this been happening?”
“Since the Agile transformation. 2021? Mike’s team was brought in to ‘improve story quality.’ They set up the Friday review as a checkpoint.”
“Has anyone ever measured whether story quality actually improved?”
Silence.
“Has anyone asked if this step is still necessary?”
More silence.
Edward wrote on the whiteboard in large letters: THIRTEEN WEEKS. NOT TWELVE.
“We just found another week of waste hiding in the process. A meeting nobody questioned because it was called ‘agile.’“
Edward stepped back from the whiteboard again.
“Thirteen weeks,” he said. “Minimum. To ship a single feature.”
“That’s… longer than I thought,” one of the engineers said.
“Look at the breakdown.” Edward circled numbers on the board. “Development is two weeks. Actual coding. Everything else is waiting. Waiting for sign-offs. Waiting in queues. Waiting for people to review things. Waiting for a Friday meeting.”
Rachel did the math. “Eleven weeks of waiting. Two weeks of work.”
“Eighty-five percent of our cycle time is queue time.” Edward set down the marker. “This isn’t a technology problem. This is a process problem.”
The room was quiet.
“Now here’s the question,” Edward said. “Where can we reduce that wait time? And does AI help with any of it?”
They went through the list item by item.
Friday Agile Coach Approval (1-2 weeks): A meeting that existed because someone decided stories needed “quality gates.” Nobody measured whether it improved anything. AI didn’t help with this. Eliminating the meeting did.
Requirements to Design (3 weeks): The wait was for product sign-off. A committee met weekly to approve requirements. If you missed the meeting, you waited another week. AI didn’t help with this, but changing the approval process did.
Design to Development (2 weeks): Environment provisioning. Every new feature needed a development environment, and the infrastructure team was backlogged. AI didn’t help with this, but automation did. They’d been asking for automated environment provisioning since 2019.
Development (2 weeks): Actual coding. AI could potentially speed this up. Copilot, Claude, AI-assisted development.
Code Review Queue (4 days): Kevin Nakamura was the bottleneck. He reviewed everything for the Claims Processing team. AI pre-review could flag obvious issues before code reached Kevin, reducing his queue.
Writing Test Scripts (2 weeks): Manual test script writing. This was a clear AI opportunity. AI could generate test scripts from requirements.
HIPAA Compliance (3 weeks): A checklist that hadn’t been updated since 2018. Half the items were for risks that no longer existed. AI didn’t help with this, but reviewing the checklist did.
Approval Queue (1 week): A three-day approval queue that existed because in 2016, someone deployed bad code and crashed a system. No one remembered why the step existed, but no one wanted to remove it.
Edward stood back and looked at the board.
“Eight bottlenecks,” he said. “How many of them involve AI?”
Rachel counted. “Three. Maybe three.”
“And the other five?”
“Process problems. A Friday meeting nobody questioned. Approval queues that exist because someone was nervous five years ago. Checklists that haven’t been updated. Automation we never prioritized.”
“So if we fix all eight bottlenecks…”
“Thirteen weeks becomes maybe three or four.”
“And the medication tracking feature ships.”
The room was quiet. Then one of the engineers, Priya Sharma, team lead for Claims Processing, said: “This isn’t AI transformation. This is just… fixing things.”
“Exactly.” Edward smiled. “AI is one tool. Maybe it helps with three of these bottlenecks. But the other four have nothing to do with AI. They’re just waste we’ve never bothered to remove.”
Monday, June 30, 2025 – 10:00 AM – CEO’s Office, Riverton Health Systems
David Aldridge listened carefully as Edward laid out the value stream map.
“Eleven weeks from requirement to deployment,” Edward said. “Two weeks of actual work. Nine weeks of waiting.”
“And you think you can fix this?”
“I think we can ship the medication tracking feature in ninety days. Not by adopting AI, but by fixing the process. AI helps with some of the bottlenecks. But most of what’s slowing us down has nothing to do with technology.”
David leaned back. “The board is expecting an AI strategy. Johnny’s been building his governance framework. “
“Johnny’s governance framework is designed to manage AI adoption. But we don’t have an adoption problem. We have a shipping problem. Features get stuck because our process has nine weeks of waste built into it.”
“What are you proposing?”
“Give me ninety days and one feature. The medication tracking feature, the one that’s been stuck for eighteen months. If we can ship it in ninety days, I’ve proven the approach works. If we can’t, I’ll admit the consultants were right and we’ll do it their way.”
David was quiet for a long moment.
“Johnny won’t like this.”
“Johnny’s measuring workshop attendance. I’m proposing to measure features shipped.”
“And if something goes wrong? If there’s a HIPAA violation, a security breach. “
“I’ll own it. Personally. This is my bet.”
David was quiet. Twenty years they’d worked together, in various configurations. He knew Edward wasn’t prone to bold claims.
“Ninety days,” David said finally. “Ship the medication feature. And Edward, don’t make me regret this.”
Monday, June 30, 2025 – 11:00 AM – CEO’s Office, Riverton Health Systems
Victoria Hale, the independent board director, caught them in the hallway.
“David, Edward. I have concerns.”
David gestured toward his office. “Five minutes.”
Victoria sat down, her expression skeptical. “I just saw the board materials Edward submitted. Value stream mapping? Removing approval queues?” She shook her head. “Edward, this is elementary. Every organization knows they should map their processes. Every Lean Six Sigma workshop teaches this. Where’s the AI strategy? Where’s the innovation?”
Edward had been expecting this.
“You’re right that it’s elementary. But elementary doesn’t mean easy.” He leaned forward. “Everyone knows they should map their value stream. Everyone knows they should remove waste. But have you asked why they don’t do it?”
Victoria waited.
“Because consultants sell complex frameworks,” Edward said. “Thirty-six-month roadmaps. Centers of Excellence. Maturity models. They sell complexity because complexity justifies fees. Nobody pays $2 million for ‘look at your process and fix what’s broken.’“
“So you’re saying the entire transformation industry is wrong?”
“I’m saying the entire transformation industry has an incentive to make simple things complicated. The framework isn’t the hard part. The hard part is discipline, doing simple things consistently. That’s what organizations can’t do. Not because they lack knowledge. Because they lack the discipline to apply what they already know.”
Victoria was quiet.
“The medication tracking feature has been stuck for eighteen months,” Edward continued. “Not because we lack AI tools. Because we have a three-week compliance review that hasn’t been updated since 2018. That’s not an AI problem. That’s a discipline problem.”
“And you think ninety days of discipline fixes eighteen months of delay?”
“I think ninety days of actually looking at the problem beats two years of workshops about how to look at the problem.” Edward met her eyes. “The consultants have frameworks for everything except shipping. I’m proposing we skip the frameworks and just ship.”
David intervened. “Victoria, let’s give Edward his ninety days. If it works, we’ve found something. If it doesn’t, we go back to the traditional approach.”
Victoria nodded. “Fine. But I want weekly updates. And I want to understand why this works, if it works, when the frameworks don’t.”
“Fair,” Edward said. “But I’ll warn you: the answer isn’t going to be satisfying. The answer is going to be that we actually looked at the process and fixed what was broken. No magic. Just discipline.”
“That’s what I’m afraid of,” Victoria said. “Because if that’s true, it means we’ve been wasting money on consultants for twenty years.”
“Only one way to find out.”
Tuesday, July 1, 2025 – 2:00 PM – Conference Room B, Riverton Health Systems
Johnny Morrison was not happy.
Edward had known Johnny for three years now, ever since Johnny had been brought in to lead digital transformation at Riverton. Before that, Johnny had been VP of Technology at MedFirst Health, a regional hospital chain in the Midwest. What happened at MedFirst still followed him around like a shadow.
In 2019, a MedFirst engineer had used an early AI tool to generate code for a patient data export feature. The tool hallucinated a database connection string that happened to match a production server. The code went through review, passed testing, and deployed. Nobody caught the error. For six weeks, patient records were being exported to an unsecured staging environment. A routine audit found the breach. HIPAA violations. Class action lawsuit. $47 million settlement. The engineer was fired. Johnny resigned, even though he hadn’t approved the tool or known about the experiment. He was the VP. It happened on his watch.
He’d spent the next two years consulting, rebuilding his reputation. When Riverton hired him to lead AI transformation, he’d been explicit about his approach: governance first. Structure before speed. Never again would something deploy without proper oversight.
Edward understood the scars. He just thought Johnny had learned the wrong lesson.
“You went to David without consulting me,” Johnny said, his voice tight. “AI strategy is my domain. The CEO hired me to manage this.”
“I’m not doing AI strategy,” Edward said. “I’m shipping a feature.”
“By using AI tools. Without proper governance. Without. “
“By fixing our process. AI is involved in maybe three of seven bottlenecks. The rest is removing waste that’s been there for years.”
Johnny sat down. “I don’t understand what you’re trying to accomplish.”
“Let me ask you something. The medication tracking feature. It’s been in the backlog for eighteen months. Why?”
Johnny blinked. “That’s not really an AI question.”
“Exactly. It’s a process question.” Edward leaned forward. “And until we answer the process question, all the AI tools in the world won’t help us ship faster. I’ve been mapping the value stream. Eleven weeks from requirement to deployment. Two weeks of actual work. Nine weeks of waiting.”
“So you need AI to speed up the waiting?”
“No. I need to understand why the waiting exists. Some of it is real, compliance reviews for patient safety. Some of it is theater, approval queues that exist because someone was nervous five years ago.” Edward paused. “AI can help with some bottlenecks. But treating this as an ‘AI adoption’ problem is like treating a broken leg with a new pair of shoes.”
Johnny was quiet.
“Here’s what I think,” Edward continued. “You’ve seen transformation failures up close. The HIPAA violation at MedFirst. You know what ungoverned experimentation can cost. Those concerns are legitimate.”
Johnny nodded. “They are.”
“But your solution, ban everything until we have perfect governance, creates a different risk. It drives usage underground. It makes the problem invisible instead of solving it.” Edward stood. “I’m proposing we fix the process. AI where it helps. Governance where it matters. Not governance as a way to avoid doing anything.”
“And if something goes wrong?”
“Then we learn from it. Fast. That’s the difference between process improvement and process theater.”
Johnny didn’t agree. But he didn’t escalate either.
It was a start.
Friday, July 4, 2025 – 3:00 PM – Edward’s Office, Riverton Health Systems
Edward dissolved every existing governance body and created one new one.
“We’re calling it the AI-SDLC Board,” he told Rachel. “Not because AI is magic. Because AI is infrastructure now. It touches every part of how we build software. So governance has to be embedded in the SDLC, not bolted on top.”
He wrote the charter on the whiteboard:
AI-SDLC Board Charter
Mission: Reduce toil. Reduce risk. Deliver better software outcomes.
Principles:
– We are AI-biased, but adopting AI alone won’t create these outcomes
– We reduce handoffs and protect nothing that doesn’t protect patients
– We act now, not later. Waiting is not safe.
– Any team member can ask for a decision. Answer in 24 hours or less.
“Any team member?” Rachel asked. “That’s going to be chaos.”
“That’s going to be accountability. If someone on the ground sees a blocker, they shouldn’t have to wait for it to bubble up through three layers of management. They ask. We answer. In a day.”
The board:
– Edward (CTO) – accountable for outcomes
– Sandra Williams (CFO) – accountable for spend
– Maria Santos (Legal/Compliance) – accountable for risk
– Rachel Torres (Chief of Staff) – accountable for execution and coordination
– One rotating engineer from the pilot team – accountable for ground truth
“What about Johnny’s steering committee? The Architecture Review Board? The Security Council?”
“Gone. All of them. One board. One place where decisions happen.” Edward underlined the charter. “We’re not waiting to improve anymore. Every committee that doesn’t decide is overhead. We’re cutting the overhead.”
“That’s aggressive.”
“That’s the point. We’ve been protecting process for years. Now we’re protecting outcomes.”
The first meeting was Thursday, April 3.
Agenda:
1. Approve automated environment provisioning (bottleneck #2)
2. Review updated HIPAA checklist (bottleneck #6)
3. Authorize AI-assisted code review pilot (bottleneck #4)
All three items were approved in forty-five minutes. No escalation. No “let’s socialize this with stakeholders.” Decide or defer, but don’t delay.
That afternoon, an engineer from the Provider Portal team sent an email: “Can I get approval to use AI for test generation? My manager says it needs board review.”
Edward replied within two hours: “Approved. Document what you learn. Share results in Friday standup.”
Word spread fast. By the end of the week, the AI-SDLC Board had fielded eleven requests. Nine were approved same-day. Two needed more information and were resolved within 48 hours.
“This is insane,” Johnny told Edward in the hallway. “You can’t run governance this fast. Something’s going to slip through.”
“Something might. And when it does, we’ll fix it fast. That’s the difference between process improvement and process theater.” Edward paused. “Johnny, we spent two years building governance structures that made us feel safe. Did they actually make us safe? Or did they just make us slow?”
Johnny didn’t answer. They both knew the answer.
“Twenty-four hours or less,” Edward said. “That’s the new standard. If we can’t decide that fast, we don’t understand the problem well enough.”
Monday, July 14, 2025 – 9:00 AM – Small Conference Room, Riverton Health Systems
The pilots started with the medication tracking feature.
Priya Sharma’s Claims Processing team would handle the core development. They’d already mapped the value stream. They knew where the waste was. Now they needed to prove they could eliminate it.
Week 1: Automated environment provisioning deployed. Wait time: 2 weeks → 2 hours. No AI involved.
Week 2: Updated HIPAA checklist approved. 47 items reduced to 23. Wait time: 3 weeks → 1 week. No AI involved.
Week 3: AI-assisted code review pilot launched. Kevin Nakamura was skeptical, but he’d agreed to try.
The first week was rocky. Kevin couldn’t get the AI to understand what he was looking for. He’d type “review this code” and get generic feedback about style and syntax. Nothing about the business logic, the security considerations, the HIPAA implications.
“The tool is useless,” Kevin complained to Edward. “It doesn’t know anything about our codebase.”
“Have you told it about our codebase?”
“What do you mean?”
“The AI doesn’t know we’re a healthcare company. It doesn’t know about HIPAA. It doesn’t know what the medication tracking feature is supposed to do.” Edward paused. “Have you given it that context?”
Kevin was quiet. “I just… asked it to review the code.”
“That’s the same thing your junior engineers do. They drop code in your queue without explanation, and you have to figure out what they were trying to accomplish.”
“So I have to explain things to the AI like I’m onboarding a new hire?”
“Exactly like that. Except the AI can actually remember what you told it.”
This was the breakthrough. Once Kevin started treating the AI like a very fast but very ignorant junior engineer, one who needed explicit context about every review, the results improved dramatically.
“The AI catches the obvious stuff,” Kevin reported after the second week. “Style violations. Common security patterns. Syntax issues. That’s maybe three hours of work I don’t have to do.”
“And?”
“And I’m finding things I never had time to look for before. Logic errors. Architectural problems. The stuff that actually matters.” He paused. “But here’s the weird thing. Writing the context for the AI is making me better at onboarding new people. I’m finally articulating things I’ve just… known. For years.”
Edward nodded. This was exactly what he’d seen in the value stream mapping session. The AI wasn’t just a tool; it was forcing engineers to externalize knowledge that had been locked in their heads. The same engineers who couldn’t explain their review process to Edward could now explain it to an AI. And in doing so, they were making that knowledge visible, transferable, improvable.
“That’s the real unlock,” Edward said. “The AI doesn’t just do the work. It forces you to understand your own work well enough to explain it.”
Kevin was quiet for a moment. “I still don’t trust it. But I’m starting to see why this might matter.”
Week 4: AI-generated test scripts. The QA team was nervous, but the first batch of scripts caught two bugs the manual scripts had missed.
“The AI doesn’t understand our business logic,” the QA lead said. “It generates tests based on the code, not based on what the code should do. But it’s finding edge cases we never thought to test.”
Week 5: Edward called a meeting that would reshape how Riverton built software.
“We’re keeping the gates that matter,” he announced to the leadership team. “HIPAA compliance. Patient safety reviews. Anything that protects patients stays. But the gates that exist because someone got nervous five years ago? Those are coming down.”
He drew two columns on the whiteboard. “Important gates” on the left. “Waste gates” on the right.
“The approval queue that exists because someone deployed bad code in 2016? Waste. The three-person sign-off for production changes when one person would do? Waste. The Friday agile coach meeting that adds two weeks to every story? Waste.”
“What happens to the people running those gates?” Sandra Williams asked. The CFO’s question, always.
“That’s the key.” Edward turned to face her. “Most companies attack EBITDA by firing people. Cut the waste, cut the headcount. I’m proposing something different. We take the people who’ve been causing wait time and reassign them to actually adding value.”
“Adding value how?”
“The QA team has twenty people doing manual regression testing. That’s waste. But those twenty people know our system better than anyone. They know every edge case, every failure mode, every bug that’s ever shipped.” Edward pointed to the value stream map. “What if they stopped running tests and started building features? What if their knowledge of what breaks became knowledge of how to build things that don’t break?”
Rachel jumped in. “Lisa Martinez has been doing QA for eight years. She’s been asking to learn development the whole time. She knows more about our system’s failure modes than most of our senior engineers. If she transitions to development, she’s not starting from zero. She’s bringing eight years of context.”
“Same with Security,” Edward continued. “Ben, your team spends 60% of their time on routine code audits. What if they built tooling that automated the routine stuff? Then they could focus on the hard problems, the ones that actually need human judgment. And we’d have capacity freed up for new products.”
Ben Adler was skeptical. “You’re talking about a fundamental change to how we operate.”
“I’m talking about taking people who are currently creating wait time and turning them into people who create value.” Edward met his eyes. “The important gates stay. Patient safety, HIPAA, anything with real consequences. But the gates we built out of fear? Those go. And the people manning them become builders.”
Sandra was doing math in her head. “If this works, you’re not cutting costs. You’re increasing output with the same headcount.”
“Exactly. The CFO playbook says cut waste by cutting people. But waste isn’t people. Waste is how we’re using people. Those twenty QA engineers aren’t the problem. Having them click through the same screens a thousand times instead of building things, that’s the problem.”
“And if some of them can’t make the transition?”
“Some won’t. That’s true of any change this big. But I’d rather have fifteen people building features than twenty people creating wait time.” Edward paused. “And honestly? Most of them are going to thrive. They’ve been bored for years. Running the same tests, manning the same gates, watching engineers ship things while they waited. This gives them a chance to build.”
Ben Adler was quiet for a moment. Then: “The important gates stay?”
“The important gates stay. Your expertise on the hard security problems, that’s irreplaceable. I’m not asking you to give that up. I’m asking you to stop wasting your team’s talent on things a tool could catch.”
“You know,” Edward said as the meeting broke up, “in my experience, no one can ever make security people happy.”
Ben raised an eyebrow.
“Give security more budget, they want more headcount. Give them more headcount, they want more authority. Give them more authority, they want everyone to stop shipping code entirely.” Edward grinned. “The only thing that actually makes security people happy is when engineers start thinking about security themselves. Then security becomes a partnership instead of a police force.”
Ben laughed despite himself. “That’s actually true.”
“So let’s make your team happy. Build the tools. Train the engineers. Keep the gates that matter. Tear down the ones that don’t.”
Week 6: Edward formed three focused feature teams for the medication tracking feature. Each team had everything they needed: developers, a former QA engineer transitioning to development, access to security tooling, and authority to make decisions without waiting for committee approvals.
The people who used to create wait time were now creating value. The gates that mattered stayed. The gates that didn’t were gone.
Priya’s team started coding. Actually coding. Not waiting for approvals or environments or sign-offs. Coding.
Thursday, August 14, 2025 – 4:00 PM – Edward’s Office, Riverton Health Systems
The Infrastructure team pushed too fast.
They’d seen the results from Claims Processing, the automated provisioning, the streamlined approvals, and they wanted the same thing. But they skipped the value stream mapping. They deployed changes without testing them properly. They assumed the approach would work without understanding why it worked.
Thursday afternoon, the staging environment went down. Twelve hours of downtime. Three board members wanted to shut everything down.
Sandra Williams, the CFO, called Edward at home that night.
“The board is asking questions. They want to know if this is going to happen again.”
“It might,” Edward said honestly. “Process improvement has failures. The question is whether we learn from them.”
“That’s not going to satisfy the board.”
“Tell them this: the approach didn’t prevent this failure. It made the failure visible and learnable. Under the old system, this would have happened anyway. we just wouldn’t have known why. Now we do. The Infrastructure team skipped the value stream mapping. They won’t make that mistake again.”
Sandra was quiet for a long moment.
“I’ll back you,” she said finally. “But Edward, the next failure can’t be this visible.”
Friday, August 29, 2025 – 11:30 AM – Edward’s Office, Riverton Health Systems
Day 60.
The medication tracking feature was in code review. Halfway done. On track to ship by day 87.
Edward stood at his window, looking out at Biscayne Bay. Sixty days ago, he’d made the 90-day bet with David. Two months of proving the approach could work.
The value stream map was still on his wall. Thirteen weeks → four weeks. Eleven weeks of waiting → two weeks of waiting.
Only three of the eight improvements involved AI:
1. AI-assisted code review (Kevin’s queue)
2. AI-generated test scripts (QA bottleneck)
3. AI-assisted development (faster coding)
The other five were just removing waste:
1. Eliminated the Friday agile coach approval meeting
2. Automated environment provisioning
3. Updated HIPAA checklist
4. Removed unnecessary approval queue
5. Streamlined requirements sign-off
More than half the improvement had nothing to do with AI.
Rachel knocked on his door.
“The medication feature passed code review. Kevin signed off this morning.”
“How long did it take?”
“Two days. Including the AI pre-review.”
“What was it before?”
“Four days in the queue. Two days of review. Six days total.”
Edward nodded. “And the test scripts?”
“Generated and running. QA found two bugs already, both from edge cases we never would have tested manually.”
“So we’re on track?”
“Day 87. Maybe earlier.”
Edward turned from the window. “When I was at the Academy, my commanding officer had a phrase. ‘Know where you are before you decide where you’re going.’ Captain Drake said it before every operation. I must have heard it a hundred times.”
“What made you think of that?”
“Because that’s what this is. We mapped the value stream. That’s knowing where we are. We identified the bottlenecks. That’s knowing what’s stopping us. We fixed them, some with AI, most without. That’s navigation, not transformation.”
Rachel smiled. “That’s not going to fit on a consulting slide.”
“Good. If it fit on a slide, it wouldn’t work.”
Thursday, September 25, 2025 – 6:00 AM – The Dock, Edward’s House on the Intracoastal
Six months since spring break now. Day 87. The problem had changed.
He wasn’t worried about Harold anymore. Wasn’t worried about becoming obsolete. Because he’d discovered something: the threat wasn’t AI. The threat was process debt. years of accumulated waste that made it impossible to ship the things that mattered.
AI was just one lever. Maybe it helped with three bottlenecks out of seven. But the other four had nothing to do with technology. They were just approval queues and outdated checklists and automation that had been on the backlog since 2019.
His phone buzzed. Text from Priya:
Feature passed final QA. Heading to compliance review Monday. We might beat day 87.
Edward smiled.
The medication tracking feature, the one that could save lives, the one that had been stuck for eighteen months, was going to ship. Not because of AI strategy. Not because of transformation roadmaps. Because they’d asked a simple question: what’s stopping us?
And then they’d fixed it.
Jennifer found him there around nine, carrying two mugs of coffee. His was black, hers had enough cream to make it barely coffee. Twenty years of marriage and she still couldn’t understand how he drank it that way.
“You’ve been out here since six,” she said, handing him a mug.
“I know.”
“The girls are asking when you’re coming in. Lily wants you to make the shrimp and grits.”
Edward smiled. His grandmother’s recipe, the one he’d learned in Savannah as a kid. Lily had requested it every Sunday since she was six. “Tell her twenty minutes.”
“I’ll tell her an hour. You always underestimate.” Jennifer sat down beside him on the dock, legs dangling over the edge the way they’d done in New York, back when they were young and the only waterfront they could afford was the East River. “You seem different lately. More energized. But also more… peaceful?”
“I figured something out.”
“The AI thing?”
“It’s not about AI. That’s what I figured out.” He sipped his coffee. “Everyone keeps asking ‘how do we adopt AI?’ But that’s the wrong question. You have to start with the outcome you want. What are you trying to ship? Then you map the process, find where work gets stuck, and fix the bottlenecks. AI helps with some of them. Not all.”
“And?”
“And the medication tracking feature was stuck for eighteen months. Not because we lacked AI. Because we had approval queues nobody remembered creating and a compliance checklist from 2018 that nobody had updated.” He laughed. “We spent two years talking about AI strategy. Turns out what we needed was to cut the process bloat.”
Jennifer was quiet for a moment. Then: “That’s not very exciting.”
“No. It’s not. That’s why nobody does it.” He turned to look at her. “AI is the shiny object. Everyone wants to adopt AI. But AI doesn’t fix process problems. It just makes bad processes run faster.”
“So what does fix them?”
“Start with the end in mind. What’s the outcome? For us, it was shipping the medication feature. Then map the value stream. Where does work actually wait? What are the real constraints, not the imagined ones? Then you redesign the flow.” He set down his coffee mug. “We went from eleven weeks to under five. AI handled three of the bottlenecks. The other four were just process bloat we’d accumulated over years.”
“That’s the one for hospitals? For drug interactions?”
“That’s the one. It could save lives, Jen. Real lives. And it was stuck because of process debt. Approvals that didn’t need to exist. Handoffs that added no value. A governance structure designed for a company we used to be, not the one we are now.”
They sat in silence for a while, watching the boats go by.
“So what’s next?” Jennifer asked.
“We scale it. Not AI. the approach. Value stream mapping. One governance board. Fix the bottlenecks, whatever they are.” He paused. “Johnny’s not going to like it.”
“Johnny’s the transformation guy?”
“Johnny thinks transformation means workshops and maturity models and Centers of Excellence. I’m proposing we just… fix things.”
Jennifer smiled. “That doesn’t sound like something you need permission for.”
“It’s not. That’s what I’m starting to realize.” Edward stood, picking up his fishing pole. “You don’t need a transformation strategy to ship faster. You need to know what you’re trying to build, see where the process is broken, and fix it. AI is just one tool in the toolkit.”
He walked back toward the house, Jennifer beside him.
The medication tracking feature would ship on day 87. Hospitals would start using it by July. And somewhere, eventually, a nurse would get an alert about a drug interaction that might have been missed.
The medication tracking feature would ship on day 87. Hospitals would start using it by October. And somewhere, eventually, a nurse would get an alert about a drug interaction that might have been missed.
That was the goal. Not AI adoption. Not digital transformation. Shipping the things that mattered.
Everything else was just process improvement.
