Claude, A to Z

Creating your account

Rosa Mendez owns Panadería Rosa, a Mexican bakery on the west side of Columbus. Her nephew, home from college, showed her Claude on his phone one Sunday after church. "Tía, you just ask it things. Like, for the bakery." Rosa was skeptical, she'd tried one of those chat things before and it gave her a concha recipe calling for ingredients she couldn't buy anywhere in Ohio.

Signing up took her four minutes on a Tuesday morning before the ovens were hot. She went to claude.ai, typed her email, clicked the verification link, and picked a password. That was it. No credit card, no download, no app store. The free plan opened right up in her browser.

The free plan, she learned, is genuinely useful: you get a solid number of messages each day, enough to write, brainstorm, and analyze documents. The paid plan, Claude Pro, twenty dollars a month, raises those limits substantially, handles bigger uploads, and unlocks Projects, which you'll meet in Module 4. Rosa's rule, which is worth stealing: start free. Use it for two weeks on real tasks. If you keep hitting the daily limit right in the middle of something important, that's your signal to upgrade. She upgraded in week three, on the afternoon she was rewriting her catering menu and the free limit cut her off mid-sentence.

One small thing that matters more than you'd think: when it asks for your name, use your real first name. Claude addresses you by it, and the whole thing stops feeling like software and starts feeling like an assistant who knows who she's talking to.

The takeaway: creating the account is trivial. The only real decision is free versus Pro, and "start free, upgrade when the limit annoys you" is the right call for nearly every owner.

Claude vs. chatbots you've tried

Devon Carter runs Devon's Heating & Cooling outside Dayton. Two years ago he put a cheap chatbot on his website. A customer asked, "Do you offer free furnace replacement with a tune-up?" and the bot answered, "Yes! We offer free furnace replacement!" It had hallucinated an entire promotion. A furious customer showed up with a screenshot. Devon ripped the bot off his site that same night.

So when a buddy told him to try Claude, Devon's first question was blunt: "Will it make stuff up about my business?" The honest answer: any AI can, but Claude is built to be careful, and more importantly, you can fence it in.

He ran a test. He pasted his actual service list, sixteen services, plain text, and wrote: "A customer with a 20-year-old furnace asks what we recommend. Answer using ONLY the list I gave you. If the answer isn't in the list, say so." Claude recommended a replacement consultation and a tune-up, both on his list, and said it couldn't quote a price because no prices were listed. It did not invent a free furnace, a discount, or a service he doesn't offer.

Three differences Devon noticed versus every bot he'd tried before: first, Claude asks clarifying questions instead of guessing, "Is the furnace gas or electric?" Second, it holds long, detailed instructions without forgetting them halfway through the conversation. Third, when it's unsure, it says so plainly instead of bluffing with total confidence.

The takeaway Devon now gives every owner who asks: Claude's carefulness is real, but the fence is yours to build. Give it your real documents, tell it "use only this," and it stays honest. Skip that step and any AI will happily invent your business for you.

Your first tasks

Priya Nair owns Luxe Locks, a hair salon in Westerville. Her first evening with Claude, she didn't try to revolutionize anything. She picked three small, real, low-stakes tasks, the kind of thing she'd normally do at 10pm with a glass of wine and a sigh.

First: her no-show policy sounded like a legal threat. She pasted it in and asked for "warm but firm, like I'm talking to a regular I actually like." Second: a supplier email thread, eleven messages about backordered color, which she pasted with the question "what's the actual delivery date, and what do I need to decide?" Third: Tuesdays were dead, so she asked for five slow-day promotion ideas "that don't discount my best services."

All three came back usable in under two minutes. She used the policy nearly verbatim. She learned the color would arrive Thursday, the date had been buried in message eight of eleven. And she ran a "bring-a-friend Tuesday" that filled four chairs the very next week.

Her rule now, and it's a good one: judge Claude like a new employee's first draft. Fast, smart, needs your review. Start with tasks where a mistake costs you nothing, wording, summaries, ideas, and save the customer-facing numbers for later, once you've learned its habits.

Tone control

Marcus Webb runs Marcus & Sons Plumbing with his two sons. His appointment reminders went out sounding like a bank: "Dear valued customer, this is a reminder of your scheduled service appointment..." Customers didn't reply. His youngest son texted customers like "hey it's marcus, still good for tomorrow at 9?", and got instant replies.

Marcus pasted three of his son's actual texts into Claude and wrote: "Write my reminders to sound like this. Short. Friendly. Like a neighbor, not a corporation." The next reminder went out as: "Hi Dana, Marcus here. Still good for tomorrow at 9 for the water heater flush? Reply YES to confirm." His confirmation rate tripled in a month.

What Marcus learned about tone control: describing the tone with adjectives alone is nearly useless, "professional but friendly" means nothing to anyone. Show, don't tell. Paste two or three examples of writing that sounds like you. Name the audience. And ban the phrases you hate, Marcus added one line to his standing instructions: "Never say 'valued customer' or 'per our records.'"

Here's the part worth knowing: matching a writing style from examples is one of Claude's standout strengths, it's genuinely better at it than most AI tools. Feed it your voice once, and it holds it across everything you ask for.

Long-form drafts

Elena Ruiz is the office manager at Bright Smile Dental in Dublin. The practice needed a new-patient welcome sequence, three emails, plus a one-page "what to expect at your first visit" handout. She'd been putting it off for six months because every attempt came out sounding like a brochure.

She gave Claude everything factual up front: the address and parking ("park behind the building, entrance on 5th Street"), all six insurance plans they accept, and the first-visit steps in order (paperwork, X-rays, cleaning, doctor exam). Then she wrote: "Write three welcome emails, warm and reassuring, many of our patients are nervous. Use ONLY these facts. Do not invent anything."

The draft came back structured and genuinely soothing, it even opened the first email with "It's normal to feel nervous. Here's exactly what happens, so there are no surprises." Elena's job was then pure verification: she corrected one insurance name, added the Saturday hours, and cut a paragraph that drifted into marketing-speak. Forty minutes total, for something she'd avoided for half a year.

Her takeaway, and the lesson of this module: for long drafts, you supply the facts and the audience; Claude supplies structure, transitions, and flow. And the "do not invent anything" line isn't optional decoration, it's the entire game whenever the writing goes to a customer.

Editing and tightening

Sam Okafor owns Greenline Landscaping. His website's About page was 600 words of ramble, his grandfather's first mower, the 2016 truck, an entire paragraph about mulch varieties. A customer told him, kindly, "Your website made me tired."

Sam pasted the page into Claude with this: "Cut this to 150 words. Keep: the family-business story, our service area, the satisfaction guarantee. Cut: everything else. Keep it sounding like me, not like an ad." What came back kept the heart and lost the ramble. Then he ran his service descriptions through the same treatment, "we do mulching" became two sentences about what mulching actually does for the customer's yard, and his quote requests went up.

Sam's rule, learned the hard way on the first try: tell Claude what to KEEP, not just what to cut. His first attempt was just "make this shorter," and he got back shorter mush, technically fewer words, still no spine. "Keep these three things, cut the rest" is what produces a page with a point.

Claude is a strong editor because it preserves voice while cutting, it doesn't flatten everything into corporate speak the way simpler tools do. Give it the keep-list and it will defend those parts while trimming around them.

Uploading documents

Gus Marino owns Gus's Auto Care, a three-bay repair shop. His new diagnostic tablet arrived with a 90-page manual, English translated from German, dense as a brick. It sat on his toolbox for a month while he guessed his way through the menus.

One slow afternoon he dragged the PDF straight into the Claude chat window and typed: "How do I run an ABS bleed test on a 2018 F-150? Give me numbered steps from this manual, in plain language." Claude walked him through it, referencing the manual's own procedure. The job took twenty minutes. The manual had been sitting there for thirty days.

The practical notes Gus would give you: Claude accepts PDFs, Word documents, text files, even spreadsheets, just drag and drop, and you can attach several at once. It reads the actual text of the file, so a proper PDF works beautifully while a phone photo of a printed page won't. And it performs best when you point it at the document: adding "using only this manual" keeps it from mixing in generic internet advice that might not match your equipment.

This is the capability that separates Claude from the chatbots most owners tried first. Long documents don't confuse it, they're its home turf.

Summarizing reports

Lena Fischer runs Lena's Table, a catering company doing about forty events a month. Her bookkeeper sends a twelve-page profit-and-loss statement every month. Lena's old method: skim the bottom line, file it, move on. Then food costs started creeping upward and she couldn't tell why, or where.

She uploaded March's P&L and asked: "Explain this to a non-accountant. Then tell me the top 3 things that got more expensive versus February, and give me one question to ask my bookkeeper about each." Claude's answer: chicken up 18%, a new linen supplier fee she'd never noticed buried in the detail lines, and overtime creeping up in the prep kitchen. Three questions, one phone call, and she'd renegotiated the linen contract by Friday.

Lena's lesson, and it's the big one for this module: "summarize this" is the weakest question you can ask a report. A generic summary gives you a generic paragraph. Ask for the decision hiding inside the document instead, what changed, what's driving it, what you should ask next. The longer the document, the more specific your ask should be.

She does this on the fifth of every month now. Twenty minutes with Claude replaced two hours of squinting at spreadsheets, and she actually understands her numbers for the first time in six years of business.

Finding the key points

Ray Delgado owns Ray's Roofing. A shingle manufacturer sent him a 25-page warranty and dealer agreement. Ray's old method: sign it, file it, hope for the best. His new method takes eleven minutes.

He uploads the agreement and asks three things: "Flag anything unusual for a small contractor. List every deadline with its exact date. Tell me exactly what voids the warranty, and quote the exact lines." Claude came back with three findings that paid for the effort a hundred times over: warranty registration is required within 30 days of installation (Ray had never registered a single job in his life), coverage is void unless the work is done by a manufacturer-certified crew (two of his guys weren't certified), and a price increase was buried on page nineteen with thirty days' notice.

He registered every 2026 job that week, got the crew certified the next month, and locked in his material pricing before the increase hit. One unsigned deadline would have cost him a warranty claim; the price lock saved him thousands.

Ray's takeaway: for contracts, don't ask for a summary. Summaries smooth over the teeth, the whole point of a contract is the teeth. Ask for flags, dates, and exact quotes. The expensive stuff in any agreement lives in single sentences, and Claude is tireless at finding single sentences.

Setting up a Project

June Park owns Sudzy Paws, a dog grooming salon in Dayton. For her first month with Claude, she retyped the same background in every single chat: "We're a dog grooming salon in Dayton, here's the price list, don't promise same-day appointments..." Then a friend showed her Projects, and she felt the way you feel when someone shows you the dishwasher after a year of hand-washing.

A Project is a dedicated workspace inside Claude. It holds two things: custom instructions (how Claude should behave in this Project) and your documents (what it should know). Every conversation you start inside the Project begins with all of that loaded. June named hers "Front desk" and wrote these instructions: "You are the front-desk assistant for Sudzy Paws, dog grooming in Dayton, Ohio. Quote ONLY from my price list. Never promise same-day appointments. Keep replies under 60 words. Warm tone, like talking to a fellow dog lover."

Setup took ten minutes. The payoff is permanent: every chat in that Project now answers like a trained receptionist instead of a stranger. "How much for a full groom on a golden retriever?" gets the right price, the right add-ons mentioned, and "we're usually booking about a week out", without June retyping a single word of context.

Think of it this way: regular chats are conversations; a Project is a trained employee. You do the training once, and it shows up to every shift already knowing the job.

Adding your business docs

Theo Alvarez runs Theo's Electric, a one-truck electrical shop. His Project got genuinely useful the day he fed it. He uploaded four documents: his price book as a PDF, his service-area town list, his warranty terms, and a document of the ten questions customers ask most.

Then he tested it the way a skeptical boss tests a new hire. "Do you travel to Riverside?", yes, it's on the town list, and Claude said so. "What's the warranty on a panel upgrade?", ten years, parts and labor, straight from the warranty doc. Then the important one: "Can you install a chandelier I bought online?" That policy wasn't in any document, and Claude said exactly that, instead of inventing an answer. That's the day Theo started trusting it.

His maintenance rule is taped to his monitor: "Prices change → update the Project the same day." A knowledge base is only as trustworthy as its newest file. He learned this the pointed way, when a February price increase didn't make it into the Project until March and a customer got quoted the old rate. Theo honored it, five minutes of updating would have saved him the difference.

The takeaway: uploading is the easy part; currency is the discipline. A Project with last year's prices is worse than no Project, because it answers with confidence.

Asking across everything

Amara Diallo owns Spark & Shine, a residential cleaning service. A client emailed: "How much for a deep clean of our 3-bedroom rental, plus the inside of the oven, and what happens if we need to cancel?" The answer lived in three different documents: the price book (deep clean rates by bedroom count), the services doc (oven cleaning as an add-on), and the policy doc (48-hour cancellation terms).

Amara pasted the client's email into her Project and asked one question. Claude came back with a single quoted answer: deep clean $289, oven add-on $45, total $334, plus the cancellation terms in plain language, noting which document each number came from. Amara copied it into her reply, double-checked the arithmetic herself, and sent it in four minutes. The client booked that afternoon.

That is the real power of a Project, and it's worth naming clearly: not one document, but one question across everything you've uploaded. Before Projects, Amara would have opened three files and done the assembly herself. Now the assembly is instant, and her job is the part machines shouldn't do, verifying the numbers and hitting send.

Her standing habit, added to every quote request: "show which document each number came from." It makes verification instant, and it keeps Claude honest, because it has to point at its source.

Proposals and quotes

Victor Nguyen owns Victor's Print Co., a small commercial print shop. A wedding venue asked him to quote 500 programs on a rush timeline, and they were comparing three shops. Victor had all the facts: $1,850 fixed, seven business days, includes two rounds of design proofs, excludes rush shipping.

He gave Claude the facts plus the structure: "Draft a proposal. Include: scope of work, timeline, fixed price, what's included, what's NOT included, payment terms, next steps. Professional but warm. Use ONLY my facts, do not invent prices or dates." The draft came back clear, confident, and human. It read like Victor on his best day.

But Victor caught something on review, and this is the reason this lesson exists: Claude had softened his payment terms from "50% deposit to start" into "a deposit to begin work." Vague deposits don't get paid. He fixed that one line, and nothing else, and sent it. He won the job, and the deposit arrived before he ordered paper.

Victor's rule for proposals: Claude is your structure and your wording; you are the numbers and the terms. Read every figure out loud against your own records before it goes to a customer. The draft will be 95% right, and the 5% will be the part with dollar signs.

Policies and handbooks

Hannah Brooks owns Hannah's Daily Grind, a coffee shop with eight employees. She needed three documents she'd been avoiding for a year: a late-opening policy (baristas running behind was becoming a pattern), a plain-English explanation of tip pooling, and a social media policy after an employee posted a video filmed behind the counter.

Claude drafted all three in plain language, each under 200 words, fair to the staff and protective of the shop. Then Hannah did the essential step that this lesson is really about: she sent the drafts to her lawyer for a one-hour review. The lawyer changed two sentences in the tip pooling section, tip law is state-specific, and Ohio has opinions, and blessed the rest. Total cost: one billable hour instead of six.

Hannah's takeaway is the right mental model for any policy work: Claude writes the 90% draft in minutes; the professional reviews the 10% that carries legal weight. Never skip the review on anything touching pay, safety, or liability. But never again pay a lawyer to write the first draft from scratch, that's the expensive part, and it's the part Claude does best.

One more habit: when you ask for a policy, ask Claude to flag what a lawyer should double-check. It knows where the legal landmines usually are, even though it can't clear them for you.

Customer communications

Owen Gallagher owns Owen's Happy Tails, a dog boarding kennel. The email arrived at 9pm, all caps: a client's dog had chewed another dog's bed during group play, and the owner was threatening a one-star review and then some.

Owen's old self would have fired back a defensive reply at 9:04 and regretted it for a week. Instead he pasted the angry email into Claude with the facts: what happened, what the boarding agreement says about personal items (the kennel isn't responsible), and what he was willing to offer (replace the bed, forty dollars). Claude drafted a reply that did four things in order: acknowledged the frustration without groveling, stated the facts calmly, cited the agreement exactly once, and offered the replacement as goodwill. No defensiveness. No admission spiral. No 9pm temper.

Owen tweaked two sentences to sound like himself and sent it. The client replied the next morning: apology accepted, and rebooked for Thanksgiving.

Owen's lesson: in a heated moment, Claude is your tone editor. You decide the two things that actually matter, the facts and the offer, and let it handle the temperature. The customer remembers how you made them feel; make sure "calm and fair" is the feeling.

Asking for sources

Ingrid Sorensen leads a small real estate team. She wanted a punchy line for her listing descriptions: "Home prices in our area are up 12% this year!" She asked Claude, and got the line plus a confident supporting paragraph. Before publishing, something nagged at her, so she asked one more question: "Cite your source and its date for that 12%."

The source was a 2022 article about a different metro area. In 2026. Ingrid's stomach dropped, that line would have gone on twenty listings, on the internet, with her name on it.

Her rule is now iron, and it should be yours for anything public-facing: every statistic Claude gives you gets two questions, "what's the source?" and "what's the date?", and anything that goes on your website, your listings, or your marketing gets verified against a primary source. For Ingrid, that's her MLS and the Realtor association. For you, it's whatever the authoritative source is in your trade.

Claude is a starting point for research, never the citation. Asking for sources doesn't make you distrustful; it makes you the person whose numbers hold up.

Double-checking numbers

Walt Kowalski owns Walt's Detail Shop. A property manager wanted a monthly contract: twelve vans, full detail, every month, for a year. Walt asked Claude to work up the quote and show its math. The proposal looked beautiful, line items, totals, professional formatting.

The math was wrong. Claude had multiplied the per-van price by ten instead of twelve, underquoting the monthly total by nearly four hundred dollars. On a twelve-month contract, that's a forty-eight-hundred-dollar typo wearing a confident smile.

Walt caught it for one reason: he recomputes everything. He fixed the number, sent the quote, and won the contract, at the right price. His rule is now absolute and non-negotiable: Claude drafts the quote; Walt owns the arithmetic. Every number that goes to a customer, prices, totals, dates, quantities, gets recomputed by a human with a calculator.

Here's the subtle part: the step-by-step math Claude shows you isn't proof of correctness. It's just easier to check than a bare total. AI is fast at arithmetic and occasionally, confidently wrong at it, and wrong numbers sound exactly as sure as right ones.

Knowing the limits

Sofia Reyes owns Staged by Sofia, a home staging business. She loves Claude, it writes her property descriptions, plans her install days, drafts her client emails. But she can tell you exactly where the fence is, because she learned each rail the honest way.

Claude doesn't know what happened after its training cutoff, ask it about this year's design trends and it may confidently describe last year's. It can't open her booking software, see her bank balance, or check a supplier's live website for today's prices. It can't see the room she's staging; it works from what she describes. And its confident tone means nothing about correctness, a wrong number sounds exactly as certain as a right one.

Sofia's working contract with the tool, and it's the contract this whole course has been building toward: Claude drafts, Sofia verifies. Facts, dates, prices, and anything public get checked against a real source. She treats it like a brilliant intern, remarkable output, zero authority to publish unchecked.

That one habit is the difference between AI as leverage and AI as liability. Everything in Modules 1 through 5 showed you what Claude can do for your business. This lesson is the one that keeps it safe: know what it can't do, and never ask it to.