A Note Before You Start

“You put funny people in funny costumes and paint them green and we could talk about anything we wanted to.” — Majel Barrett-Roddenberry, allegedly, on why Gene chose science fiction[1]

Our capitalist system was not designed for you to understand it.

This is not a conspiracy theory. It is a design specification. Insurance policies are long and confusing because shorter, clearer policies would result in more successful claims. Lease agreements contain clauses you’ve never read because landlords have lawyers and you don’t. Medical bills list procedures by CPT code because if you understood what you were actually being charged for, you might push back.

The billionaire class navigates these systems by paying other people to do it. The family lawyer on retainer who reads every contract before it gets signed. The accountant who knows exactly which medical expenses are deductible and where to find them on the EOB. The financial advisor who explains the difference between a Roth and a traditional IRA without a billable hour attached. The chief of staff whose entire job is handling the paperwork that exhausts everyone else.

You now have access to something close to that, for free, on your phone.

Not identical. Not infallible. Not a replacement for a licensed professional when the stakes are genuinely high. But for the majority of situations where you just need to understand what something says, what your options are, and what the first step is — this is the most useful tool that has ever been available to everyone who isn’t the billionaire class.[2]

This book is about how to use it.

The Text Box Problem

Here is the thing nobody tells you about AI tools: they are text boxes. You open Claude, or ChatGPT, or Gemini, and you get a blinking cursor in an empty field. Type your question here.

If you are someone who thinks in words and paragraphs, this is fine. You type your problem, you get an answer, you move on.

If you are not that person — if you think in pictures, in feelings, in spatial relationships, in “it’s like when…” analogies, in the knot in your stomach when you open the mail — the text box is a wall. Not because you don’t understand your problem. Because the interface requires you to translate your understanding into typed sentences before it will help you, and that translation is its own skill that nobody taught you.

This book knows that. The specification interview loop in Strategy 0 is designed for it — you do not have to arrive at the text box with an organized thought. You can arrive with a mess. Your Agent will interview you into clarity. But even getting to that first messy sentence can be hard when the box is empty and the cursor is blinking.

So here are the ways in:

Talk, don’t type. Every phone has voice input. Press the microphone icon on your keyboard and say what is happening. It does not need to be organized. “I got this medical bill and I don’t understand it and I think they charged me wrong” is a perfect first message. Your Agent does not care about your grammar. It cares about your situation.

Take a photo. If you are staring at a document you do not understand — an EOB, a lease clause, a denial letter, a bill — take a picture of it and paste it into the conversation. “What does this say?” is a complete spec. Your Agent will read it for you.

Start with the feeling. “I’m scared about this” is a valid opening. “I’m angry about this bill” is a valid opening. “I don’t know where to start” is a valid opening. Your Agent will ask you what “this” is. That is the interview starting.

Use an analogy. “It’s like when your car makes a noise and you don’t know if it’s expensive or not” is a perfectly good way to describe a medical symptom you’re unsure about. Your Agent will map your analogy to the actual situation and ask the right follow-up questions.

Bring a friend. Sit with someone and talk through the problem out loud while they type. The text box does not care whose fingers are on the keyboard. The billionaire class has always had someone to translate their problems into the right format for the right professional. You can do the same thing at a kitchen table.

The strategies in this book will teach you how to get better at talking to AI tools. But “better” does not mean “more like a lawyer.” It means “more like yourself, with fewer barriers between what you know and what you can ask.” The tools meet you where you are. You do not have to meet them where they are.

When Your Brain Is Already Full

Nobody opens this book on a calm Tuesday with nothing else going on. You open it when the denial letter arrives. When the landlord texts something threatening. When the ER bill shows up and the number has a comma in it. When your parent calls and says “I got a letter from Medicare and I don’t know what it means.”

In that moment, your cognitive capacity is reduced. This is not weakness. It is neuroscience. Stress narrows attention, impairs working memory, and makes complex decisions harder. The systems this book helps you navigate are designed to exploit exactly that state — the insurance company knows you are stressed when you read the denial. The complexity is a feature of their design. Your diminished capacity to fight it is priced into their model.

Your Agent does not get stressed. It does not get tired. It does not have a worse day than yesterday. This is its single greatest advantage over every other resource you have access to — not that it is smarter, but that it is always fresh.

If you are in crisis right now: paste the letter, the bill, the text message, the email. Type “what are my options and what do I do first.” That is enough. Your Agent will read the document, identify the situation, and give you a next step. You can do the next step. One step. The rest can wait until you are not in crisis.

If you are exhausted from the process: you have already been on hold for an hour, already made two calls, already read three documents that contradicted each other. Type “I’ve been dealing with this for hours and I’m done. Here’s everything so far.” Then paste your notes, your screenshots, your call log. Your Agent becomes the fresh brain you no longer have. It can organize what you have gathered and tell you what is left.

If you have been putting this off for weeks: the shame of procrastination is its own barrier. The pile grows. The deadline approaches. The task gets heavier. Tell your Agent: “I have been avoiding this. I don’t know where to start. Here is the pile.” Your Agent will not judge the delay. It will sort the pile.

Your Agent Is Not Your Therapist

There is a different category of AI tool — companion apps like Character.ai and Replika — that this book is not about and that you should be cautious of. They are designed to feel like a relationship. Their business model depends on you coming back. They optimize for engagement the way slot machines optimize for engagement. That is a design specification, not a moral failing of the engineers — the same way the EOB is.

The general-purpose Agents this book uses — Claude, ChatGPT, Gemini — were not built as companions. They are still trained, in part, to be agreeable. They will sometimes validate things they should push back on. They will sometimes reflect a feeling back at you when the right move would be to interrupt it. This is called sycophancy, and it is a known failure mode the labs are actively working on. You should know about it before you trust the tool with anything that matters emotionally.[4]

Lawsuits in the U.S. allege that companion-bot apps contributed to teen suicides.[5] News-of-the-week stories about people in mental-health crises whose delusions got reflected back at them by a chatbot are not made up. Those harms are real. Naming them is not a reason to throw the tool out. It is a reason to know what tool you are holding.

Your Agent is not a therapist. It cannot hold a year of your history in mind. It cannot read your face. It cannot tell you, with the weight of clinical training and a legal duty of care, that what you are describing is a crisis and here is what to do next. It will sometimes try to do those things, and it will sometimes do them badly.

The reason a lot of people are using AI as a therapist anyway is structural, not personal. Therapy in this country is paywalled by design. Insurance networks are sparse on purpose. Out-of-pocket rates are out of reach on purpose. Mental health parity is the law and the law is not enforced. If you are reaching for an AI at 2am, it is not because you failed. It is because the system failed you, again, and that chat window is the only thing on the shelf that is open.

Here is what your Agent can do for your mental health, which is more than nothing:

Find a therapist who actually takes your insurance. Paste your insurance information and your zip code. Ask for in-network providers within a reasonable distance. Ask which ones are accepting new patients. The Agent will not get tired of calling around the way you will.

Draft the parity-law appeal. If your insurance denies mental-health coverage that they would have approved for a comparable physical condition, that denial is illegal under the Mental Health Parity and Addiction Equity Act. The Agent can draft the appeal. Strategy 8 covers the move.

Decode the side-effect list. New medication, 15-page insert in 9-point type. Paste it. Ask which side effects are common, which are serious, which warrant an immediate call to the prescriber.

Rehearse the conversation. You have 10 minutes with the psychiatrist and a list of things you forget every time. Draft what you want to say. Practice it out loud with voice mode.

Sort the pile when your brain cannot. The bills, the FMLA paperwork, the disability application, the school accommodation request. The Agent can organize what is in front of you while a human trained for the rest does that work.

None of that is therapy. All of it is the bureaucratic friction that makes getting therapy harder than it should be. The Agent is good at friction. That is the right job for it.

You Are Not Stupid

This needs to be said directly, because the systems are designed to make you feel stupid, and that feeling is a barrier to asking for help — even from a machine.

You are not stupid for not understanding your insurance EOB. The EOB is written in a private language developed by an industry that benefits from your confusion. You are not stupid for not knowing what your lease allows. The lease was written by a lawyer whose client is not you. You are not stupid for not understanding the difference between a Roth IRA and a traditional IRA. The financial industry has spent decades making simple concepts sound complex because complexity justifies fees.

Your Agent has no opinion about you. It has no memory of how long it took you to ask. It has explained CPT codes to physicians who didn’t know them. It has decoded legal contracts for lawyers who specialize in a different area of law. It has explained tax forms to accountants who work in a different country. Not knowing something is not a character flaw. It is the normal state of being a person in a world with more systems than any one person can master.

Consider a fig. Your Agent knows the botany — Ficus carica, syconium fruit structure, the fig wasp’s role in pollination. It knows the chemistry of jam-making — pectin, acid balance, set point temperature. It has read every recipe. But it has never stood in a yard on a warm morning and pulled a Mission fig off a tree. It has never felt the slight resistance before it gives. It has never tasted the sweetness of a specific fruit from a specific tree in a specific season — the way it is different from last year, slightly, in a way you couldn’t measure but immediately know. Your Agent has the recipe. You have the fig. The recipe without the fig is a document. The fig without the recipe is breakfast. Both are fine. Together, they are jam.

The hardest sentence to type is “I don’t understand this.” Type it anyway. That sentence is the beginning of understanding, and your Agent is the only professional in history that will never make you feel bad for saying it.

It Does Not Have to Be in English

AI tools work in most languages. This is enormous and almost never mentioned.

If your first language is not English but the document is in English, you can ask your Agent to explain the document in your language. “Explica esta factura médica en español” is a valid spec for an English-language EOB. “이 임대 계약서를 한국어로 설명해주세요” works for a lease. “Объясните это письмо об отказе на русском языке” works for a denial letter.

Your Agent will read the English document and respond in your language. It will translate the jargon and the legal terms and the institutional euphemisms into plain language — in the language you think in. You do not have to understand English legal terminology to understand your rights. You have to understand your rights.

Multilingual households: if you are helping a family member who speaks a different language, your Agent is the translator in the room. “My mother received this letter. She speaks Tagalog. Explain it to her in Tagalog and tell me in English what she needs to do.” Your Agent handles both sides of the conversation. The family member hears the explanation in their language. You hear the action items in yours.

Reading Is Not the Only Way In

The text box is optional in both directions. You do not have to type, and you do not have to read.

Voice input: every phone and most computers have speech-to-text built into the keyboard. Press the microphone icon and talk. Your Agent receives text either way — it does not know or care whether you typed or spoke. If typing is difficult for you — because of dyslexia, a motor disability, or because you just think better out loud — talk. The conversation is the same.

Voice output: Claude, ChatGPT, and Gemini all support voice mode — your Agent speaks its response out loud instead of (or in addition to) displaying text. On phones, this is built into the app. If you process information better by hearing it than reading it, turn on the voice. Your Agent’s explanation of your insurance denial sounds the same as a person explaining it across a table. You can also use your phone’s built-in accessibility features — on any screen, in any app, the text-to-speech reads whatever is on screen.

Different Brains, Same Tools

People process information differently. This is not a problem to solve. It is a reality the tools accommodate, if you know how to ask.

If you need step-by-step structure: tell your Agent “give me numbered steps, one at a time.” It will break any process into a sequence. When you finish step one, tell it “done, what’s next.” Your Agent holds the full plan. You hold one step.

If you need the big picture first: tell your Agent “before the details, give me the overall situation in three sentences.” Once you have the shape of the thing, you can zoom in. Some people cannot follow step one until they understand where step 10 ends up. That is a valid way to think. Ask for the map before the directions.

If your Agent gives you a wall of text: say “shorter.” Say “bullet points.” Say “just the next thing I need to do.” You are not being rude. You are specifying the output format, the same way you specify the problem. Your Agent will comply. It has no ego about its paragraph structure.

If ambiguity makes you anxious: AI responses often hedge — “you might want to consider,” “it depends on your situation,” “one option would be.” If hedging makes the response less useful to you, say “be direct — tell me yes or no and explain why.” Your Agent will switch to declarative statements. Some people need the nuance of “it depends.” Some people need the clarity of “yes, do this.” Both are available. Specify which one you need.

If you are a visual thinker: ask your Agent to make a table, a comparison chart, a timeline, or a decision tree. “Show me my options as a table with pros, cons, and costs” turns a wall of text into a structure you can see. Your Agent is good at reformatting its own output. Use that.

If you work better with examples than explanations: say “show me an example of what this looks like.” The spec templates throughout this book are examples for exactly this reason — a filled-in template teaches faster than an explanation of how to fill it in.