Hey, it's Cindy 🌱 You commented PAID, so here is the whole first-project playbook: the tasks businesses already pay for, the prompts that build the workflow with you.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
🧠 The whole idea in one line: pick a task a business already pays money for, build a repeatable Claude workflow that does it well, run it once on a real example, and sell that documented outcome. You are selling a finished process, not your hours, which is why this works before anyone knows your name.
step one
Pick a task businesses already pay for
Notice the word "already". You are not inventing demand, you are taking something with an existing price tag and doing it faster and more consistently. Here are six that small businesses reliably pay for, with what a first project realistically sells for. These are the numbers I would quote at the start, not agency rates.
The weekly performance report Pull ad, sales or web numbers into a plain-English one-pager with what changed and what to do next. Sells at $300 to $800 a month. Buyers: ecommerce stores, agencies, clinics.
Product descriptions at volume 100 listings written to one brand voice with proper structure and keywords. Sells at $3 to $6 each, or roughly $400 per 100. Buyers: online stores, wholesalers, marketplace sellers.
Long video into short clips Take a webinar, podcast or livestream and hand back the pulled quotes, clip timestamps, captions and post copy. Sells at $250 to $600 a month. Buyers: coaches, podcasters, B2B founders.
Enquiry and FAQ replies Turn a year of repeat customer questions into a reply library and a response template set. Sells at $400 to $900 once, then a retainer. Buyers: service businesses drowning in the same five questions.
Meeting notes into actions Transcripts converted into owners, deadlines and a follow-up email, in their format. Sells at $250 to $700 a month. Buyers: small sales teams, consultants, property agents.
The content repurposing pass One blog post or newsletter turned into a week of platform-native posts. Sells at $300 to $700 a month. Buyers: anyone publishing consistently on one channel and nowhere else.
⚠️ How to choose between them, honestly: pick the one where you can already tell good from bad without asking anyone. If you have never read an ad report, do not sell ad reports. Your taste is the actual product here, because Claude will happily produce something plausible and wrong, and the client is paying you to be the person who catches it. Pick where your judgment is strongest, not where the money looks biggest.
step two
Build the workflow with Claude
Do not prompt your way through this once and hope. You are building a process you can run again next week without rethinking it. Open Claude (the free plan is enough to build and test this), start a fresh chat, and paste the prompt below. Answer its questions properly, because vague answers here produce a vague workflow.
🛠️ build my service workflow
I want to sell one specific service and I need you to help me build the repeatable workflow behind it before I sell it. The service is: [task, for example "a weekly Meta ads performance report for small ecommerce stores"]. My buyer is: [who they are and what they care about].
Start by interviewing me. Ask what inputs the client would hand over, what the finished deliverable looks like, how long I want it to take me each time, and what would make a client say this was worth the money. Ask everything before you write anything.
Then build me four things. One: the workflow as numbered steps, from receiving the client's files to sending the deliverable, with a realistic time estimate per step. Two: the exact prompts I paste at each step, written so I only have to swap in the client's details. Three: a quality checklist of the specific things that must be true before I send it, including the things you are most likely to get wrong. Four: the deliverable's structure, section by section, so every client gets the same shape.
Write it so a beginner could run it start to finish without asking you anything, and flag any step where a human has to check the work rather than trust it.
💡 When the workflow settles, save it so you never rebuild it. In the Claude app you can package it as a skill: go to Customize, then Skills, then Add, then Upload a skill, and from then on Claude runs your process the same way every time. Skills are available on every plan, Free included. Anthropic's walkthrough is here.
step three
Run it once for real, and keep the receipts
This is the step people skip, and it is the only one that turns effort into income. You need one real before and after. Not a hypothetical, not a screenshot of a chat.
📥 Find real input, without asking anyone for a favour. Use your own business if you have one. If not, use public material: a real store's live product listings, a podcast episode on YouTube, a company's published FAQ page. It is your practice run, so nobody needs to sign anything.
⏱ Time yourself and write the number down. How long the workflow took you, start to finish. "Four hours of work in 35 minutes" is a sentence that sells. A vague "saves time" is not.
🖼 Save the before and the after side by side. Screenshot the messy input and the finished deliverable. Keep both at full quality. This pair is your entire portfolio for the next month.
🔍 Run your own quality checklist on it. Every fact checked, every number traced to its source, every bit of tone matched. If you would not put your name on it, fix the workflow, not the file, and run it again.
📄 Write it up as a one-page case study. The prompt below does this from your notes. One page, five sections, no fluff.
📄 write my first case study
Turn this into a one-page case study I can send to a prospective client. Here is what I did: [describe the task, the input you started with, the workflow you ran, how long it took, and the finished output].
Write it in five short sections. The situation, in two sentences, describing the kind of business this is for and the problem. The old way, with the realistic time or cost of doing this manually. What I did, in plain language, with no tool-worship and no jargon. The result, leading with the concrete numbers I gave you and nothing invented. And what it would look like for them, in one short paragraph.
Two hard rules. Do not state any result I did not give you, and do not use words like revolutionary, seamless or game-changing. If a number is missing, write a bracket telling me what to measure rather than filling it in yourself.
⚠️ Never inflate the result. Say "this was a practice run on a public listing set" if that is what it was. Small honest numbers convert better than big vague ones, and one exaggeration that gets caught on a call costs you the client and the referral behind them.
step four
Pitch three businesses
Three, not thirty. Personal beats volume every time, and three is small enough that you will actually send them today. Pick businesses whose work you can genuinely see, so your message can name something specific.
🎯 Choose the three. Local businesses you already buy from, brands whose ads keep finding you, or businesses run by someone in your existing network. The criterion is that you can see the exact problem your service fixes on their site, their listings or their feed.
👀 Find the one specific thing. Their product descriptions are two lines long. Their podcast has 40 episodes and no clips. Their FAQ answers three questions and their comments ask twenty. That specific observation is the whole message.
✍️ Send a short message, not a proposal. Four lines: the thing you noticed, what you did about it already, the outcome in numbers, and one low-pressure question. Attach the before and after. Do not attach a PDF deck.
💰 Name a price when they ask. Use the ranges in step one. Quote a fixed price for a fixed deliverable, never an hourly rate, because hourly punishes you for being fast, and being fast is the entire advantage.
🔁 Follow up once, seven days later. One line, adding something new (a second example, a small fix you spotted). Then leave it. One good follow-up doubles replies; five make you the person they avoid.
📨 draft my three outreach messages
Write me three short outreach messages, one for each of these businesses: [business 1, what they sell, and the specific gap I noticed], [business 2, same], [business 3, same]. My service is [service] and my proof is [the result from my case study, with the real numbers].
Each message must be under 90 words and follow this shape: the specific thing I noticed about their business, the fact that I already did a version of it, the concrete result, and one easy question that is simple to answer with a yes or no. Write them so they could not be copy-pasted to any other business.
No flattery openers, no "I hope this finds you well", no talk about AI as if it is the point. The point is the outcome. Give me each message twice: once as an email with a subject line, and once trimmed for an Instagram or LinkedIn DM.
💡 If all three say no, that is information, not failure. Ask the friendliest one what would have made it a yes, then change one thing (the offer, the price, or who you picked) and send three more. Most people quit at three nos, which is the exact point where this starts working.
worth knowing
Why the finished project is the asset
The market is already paying more for AI-enabled work, but it is rewarding judgment rather than button-pushing. Upwork's Future Workforce Index 2026 found that freelancers performing AI work on its marketplace earn 34% more per hour than those not incorporating AI. The same research notes that basic AI execution is losing pricing power while higher-judgment work gains it, which is exactly why your workflow plus your quality checklist is the thing worth selling, not access to a chatbot.
Proof travels One documented before and after can be sent to every prospect, posted publicly, and reused as a portfolio piece for a year.
Process compounds The second client costs you a fraction of the first, because the workflow, prompts and checklist already exist.
Certificates do not do this A completion certificate says you watched something. A case study says you produced something someone would pay for.
It picks your niche for you Whichever of the three replies tells you which market wants this, which is far better information than another month of deciding.
this week
The five-day version
🗂 Monday. Pick your one task from the six above and write down who buys it.
🛠️ Tuesday. Run the workflow-building prompt and answer its questions properly. Save the output somewhere you will find it again.
▶️ Wednesday. Run it end to end on a real public example. Time it. Screenshot the before and after.
📄 Thursday. Write the one-page case study and check every number in it.
📨 Friday. Send three messages. Then close the laptop, because the follow-up is next week's job.