The Secret Life of AI: Patience and Flexibility

How to prompt, think, and get results from any AI tool

#WorkingWithAI #Prompting #AIConfidence

Margaret is a senior software engineer. Timothy is her junior colleague. They work in a grand Victorian library in London — and in every episode, they'll show you exactly how to get what you want from AI.

Episode 1

The fog had settled early over the city, pressing itself against the library windows as though curious about what lay inside. Timothy arrived at his usual hour, unwound his scarf with the mechanical patience of a man performing a habit he no longer noticed, and sat down across from Margaret without a word of greeting.

She did not look up from her book immediately. She never needed to.

"Tell me," she said, turning a page.

Timothy stared at the grain of the oak table. "I've been using it for three weeks now. The AI. For work." He paused. "I don't think it's working."

Margaret closed her book — unhurried, deliberate — and folded her hands on top of it. "And by working you mean it hasn't done what you expected it to do."

"I mean it gave me something completely wrong on Tuesday. Confidently wrong. And then on Wednesday it was extraordinary. And then Thursday it was —" he searched for the word — "mediocre. Like a colleague who shows up differently every single day and you can never quite rely on them."

"Yes," Margaret said simply.

Timothy looked up. "That's all you have to say? Yes?"

"I'm agreeing with you. That is precisely what it is like." She reached for the small brass lamp at the corner of the table and adjusted its angle slightly, as though the light needed to fall differently for what came next. "What tool were you using before AI to do this work?"

"Documents. Templates mostly. Some research."

"And did your templates ever come back confidently wrong?"

"Templates don't think."

"No," Margaret said. "They don't. And that is exactly where your trouble begins."

Why AI Feels So Different From Every Tool You've Used Before

She rose and moved toward the shelves with the unhurried certainty of someone who had walked these aisles ten thousand times. "We have spent forty years training ourselves to use tools that do precisely what we instruct them to do. Click save — it saves. Sort the column — it sorts. The machine is obedient. Deterministic. It does not have good days and bad days." She returned with a slim volume and set it on the table without opening it. "Then along comes something that thinks — approximately, imperfectly, impressively — and we hand it our expectations from the old world and wonder why they don't fit."

Timothy turned this over. "So the problem is my expectations."

"The problem," Margaret said carefully, "is the gap between your expectations and the nature of the thing. That is not entirely your fault. Nobody told you what you were actually sitting down with."

What Actually Happened on Tuesday

She settled back into her chair. "May I ask what happened on Tuesday? The confidently wrong answer."

"I asked it to summarize a report for a client meeting. It summarized a report — just not the right one. Made up statistics I'd never seen. Presented them like gospel."

"And what did you do?"

"I caught it. Fortunately. Before the meeting."

"Good. That instinct — catching it — that is not a failure of the tool. That is you doing your job." She paused to let that settle. "The phenomenon you encountered has a name, in fact. People who study these systems call it a hallucination — the model generating something plausible-sounding but entirely fabricated. It happens because the AI is predicting what a reasonable answer should look like, not retrieving a fact from a verified database. It does not know what it does not know. It fills the gap with confidence."

She let that land for a moment before continuing. "This is not a flaw that will simply be engineered away. It is a structural characteristic of how these systems work. Which means verification is not optional. It is part of the practice."

"So I should never trust it?"

"You should always verify what matters," Margaret said. "There is a difference. A confident tone is not the same as a correct answer — that was true of human colleagues long before AI arrived. The skill is learning which outputs deserve scrutiny and which can be accepted at face value. That judgment develops over time."

What Made Wednesday Different

"Now tell me about Wednesday. The extraordinary day."

For the first time Timothy's expression shifted toward something almost reluctant — as though admitting the good day somehow complicated the grievance. "I was drafting a difficult letter. To a client who'd had a bad experience. I didn't know how to begin. I described the situation to the AI and it gave me a draft that was —" he shook his head slightly — "better than what I would have written alone. More measured. The right tone."

"Because you brought it something it could work with. Context. Nuance. Your knowledge of the client relationship." Margaret tapped the table once. "This is the first thing I want you to understand about this tool, Timothy. It is not a vending machine. You do not insert a coin and receive a product. It is far closer to a collaboration. And like any collaboration, what you bring to it determines much of what you get back."

"That feels like a lot of work to put on the user."

"It is," she said without apology. "And no one in any brochure or documentation will tell you that plainly. But it is the truth. The quality of what you receive is deeply connected to the quality of what you provide — not just the words in your request, but the context, the specificity, the clarity of what you actually need. Vague questions produce vague answers. Precise questions, grounded in real context, produce something worth using."

The Two Human Qualities AI Will Always Require

"The second thing — and this is perhaps the harder lesson — is that this tool will require two qualities from you that most software has never asked for before."

Timothy waited.

"Patience," she said. "And flexibility."

"Patience I understand. But flexibility —"

"The model you use today is not quite the model you used last month. Its capabilities shift. Its behavior changes. The prompts that worked beautifully in February may need adjustment in April. The use cases that felt clumsy six months ago may now be its strongest suit." She opened her hands slightly, a small gesture toward the endlessness of it. "You cannot learn this tool once and consider it mastered. You must hold your understanding loosely. Update it. Stay curious rather than settled."

She continued, more quietly now. "And patience is not simply tolerance for imperfection. It is the willingness to iterate. The first answer the AI gives you is rarely the final answer. It is a starting point — a draft, an outline, a first attempt. The users who get the most from these tools are the ones who treat the first response as the beginning of a conversation, not the end of one. They push back. They refine. They ask again with more clarity. That takes patience."

Timothy was quiet for a moment. Outside, a cart horse clopped slowly through the fog-thickened street.

"That's rather exhausting," he finally said.

Margaret almost smiled. "Yes. At first. But consider what it asks of you — not technical skill, not a certification, not a manual read cover to cover. It asks for patience when it disappoints you. And flexibility when it surprises you." She picked up her book again. "Those are not software skills, Timothy. They are human ones. And you already have them. You simply haven't thought to bring them to the table."

A Note on Which AI You Use

He sat with that for a long moment.

"This works with Claude," he said slowly, "the AI I've been using. But you're describing something broader."

"I am describing the relationship between a human being and any AI tool," Margaret said. "Claude, or any of the others. The specific name on the cover matters less than you think. The principles underneath are the same — the non-determinism, the need for context, the importance of verification, the requirement for patience and flexibility. Master the human side of this equation and you will find your footing with any of these tools, whatever they happen to be called, wherever this technology goes next." She reopened her book to the page she'd been on. "Come back Thursday. We'll talk about the gap between what you expect and what you're likely to actually get. And why closing that gap is entirely within your power."

What the Letter Taught Both of Them

Timothy rewound his scarf. The fog outside had thickened further, swallowing the street lamp whole.

"Margaret," he said at the door.

She looked up.

"Wednesday's letter. The client wrote back. Said it was the most thoughtful correspondence they'd received in years."

She held his gaze for a moment. Then returned to her page.

"Bring that with you Thursday," she said quietly. "Both of you did that."

Next episode: The Expectation Gap — why the distance between what you hoped for and what arrived is the first thing worth understanding, and entirely within your power to close.

Aaron Rose is a software engineer and technology writer at tech-reader.blog.

Catch up on the latest explainer videos, podcasts, and industry discussions below.

Search This Blog

Tech-Reader.blog