You saved the 5 ChatGPT prompts. Now what?
Saving a reel is not learning. It is the same move as bookmarking gym videos and expecting muscle tomorrow. The video does not train you. Reps do.
This is a 90-day plan from beginner to your first paid AI client. The promise is simple. Show up daily, and by day 90 you have a first paid project or a paid pilot. No income numbers. No magic prompt. A system you can repeat.
The system picture
Four words. Four steps.
- Learn. Short and specific. One topic, one source.
- Build. Same day you make something, not only notes.
- Show. Proof. Before/after, portfolio, a short case.
- Sell. One offer, real conversations, first paid work.
Keep the loop short. A week ends with an artifact, not a folder named “read later”.
Daily rule
On weekdays, block 60-90 minutes. Not “when I have time”.
Doing means three things together: (1) a learn block, (2) a build block, (3) one line in a notebook or Notion about what you shipped and where you got stuck.
Track it simply. A notebook works. A Notion checklist works. Pick one place and stay there.
Month 1. Foundations (Learn + Build)
The first four weeks are foundation. Accounts, prompt structure, real work practice, one reusable assistant.
01Week 1. ChatGPT and Claude basics, plus an honest workflow diary
Goal. Open ChatGPT and Claude accounts. Finish one honest workflow diary of your real job tasks.
What to learn. What each tool is for. Free vs paid in plain words. Chat vs working with files.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You are a patient workflow coach for someone starting with AI at work.
Goal: Turn my messy notes into a clear 1-page workflow diary.
Context: My role is [your role]. Industry: [industry]. Tools I already use: [list]. Time sinks this week: [list].
Output format:
1) 10 recurring tasks (verb + object)
2) Which tasks are copy-paste / research / writing / decisions
3) Top 3 tasks where AI can help this month
4) One honest risk (what I might avoid)
No filler. Ask up to 5 clarifying questions if needed.Role: You are a triage tutor.
Goal: Help me rank tasks for AI practice this week.
Context: Here is my task list: [paste 10 tasks]. I can practice [60-90] minutes on weekdays.
Output format:
- Must practice this week (3)
- Later (rest)
- For each "must": what "done" looks like in one sentence
Keep it concrete.Role: You are a plain-language product tutor for ChatGPT and Claude.
Goal: Explain which tool fits which of my tasks, without hype.
Context: My tasks: [paste]. Constraints: [budget / device / language].
Output format: a table with columns Task | Better starting tool | Why in one line | First 10-minute exercise.
Do not push paid plans unless a free path is clearly blocked.Daily actions.
End-of-week deliverable. A 1-page workflow diary: tools, 10 tasks, time sinks.
02Week 2. Prompt engineering fundamentals (Role / Goal / Context / Format)
Goal. Write every prompt with Role, Goal, Context, Output format.
What to learn. Why structure beats a one-line “magic” prompt. How to rewrite a weak prompt into a strong one.
Free resources.
- OpenAI prompt engineering
- Anthropic prompt engineering
- ChatGPT Prompt Engineering for Developers
- DeepLearning.AI short courses
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You are a prompt engineering coach.
Goal: Rewrite my weak prompt into Role / Goal / Context / Output format.
Context: My weak prompt: [paste]. My real task: [one sentence]. Audience: [who reads the result].
Output format:
A) Improved prompt (ready to copy)
B) What was missing (bullet list)
C) One test I should run next
No fluff.Role: You run a bad-to-good prompt lab.
Goal: Show 3 upgraded versions of my prompt with different strictness levels.
Context: Task: [task]. Draft prompt: [paste].
Output format:
1) Light structure
2) Strict RGCO
3) Strict RGCO + quality checks
Then tell me which to use first this week.Role: You teach output formats.
Goal: Design a reusable output format for my recurring task.
Context: Task: [e.g. weekly status email / Instagram caption set / meeting notes]. Constraints: [length, language, tone].
Output format: a template I can paste every time, with [brackets] for variables, plus 3 quality checks.Daily actions.
End-of-week deliverable. Prompt library v1: 10 structured prompts for YOUR work.
03Week 3. Research, rewrite, summarize on real work
Goal. Use AI for research, rewrite, and summary on your real materials.
What to learn. Research packs, voice rewrites, one-page briefs. Toy demos do not count.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You are a research assistant for my real work.
Goal: Build a research pack I can act on this week.
Context: Topic: [topic]. Goal decision: [what I need to decide]. Sources I already have: [paste links or notes]. Limits: [time, region, audience].
Output format:
1) Key claims (max 7) with uncertainty tags
2) What is missing
3) 5 questions to ask next
4) One-page action brief
Mark assumptions clearly.Role: You are my rewrite editor.
Goal: Rewrite my draft in my voice without inventing facts.
Context: My draft: [paste]. Voice notes: [short / direct / no hype / examples]. Audience: [who].
Output format:
A) Rewritten version
B) What you changed and why (bullets)
C) 3 lines that still sound generic (so I can fix them)
Do not add new claims.Role: You write executive briefs.
Goal: Summarize this material into one page a busy person can use.
Context: Source text: [paste]. Reader: [role]. Decision needed by: [date or none].
Output format:
- Context (3 lines)
- What matters (5 bullets)
- Risks / unknowns
- Recommended next step (one)
Max length: one page.Daily actions.
End-of-week deliverable. 3 before/after packs from your real materials.
04Week 4. Mini build: one repeatable assistant for your job
Goal. Build one assistant (doc / sheet / email) you can rerun the same way every week.
What to learn. Instructions + examples + constraints = a reusable system.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You help me spec a reusable AI assistant for my job.
Goal: Write a clear assistant spec I can paste as standing instructions.
Context: Job: [role]. One recurring workflow: [doc / sheet / email]. Inputs I will provide: [list]. Definition of done: [one sentence].
Output format:
- Name
- Role
- Steps it always follows
- Inputs required
- Output format
- Hard constraints (what it must never do)
- 3 example runs (short)Role: You build few-shot examples.
Goal: Create 3 strong input→output examples for my assistant.
Context: Assistant spec: [paste]. Real samples: [paste 1-3 messy real inputs].
Output format: Example 1/2/3 with Input, Good output, Why this is good (1 line).
Keep examples realistic, not perfect marketing copy.Role: You are my weekly run coach.
Goal: Give me a 15-minute rerun checklist for my assistant.
Context: Assistant name: [name]. Where it lives: [ChatGPT project / Claude project / Notion / Docs]. Last issues: [paste].
Output format: numbered checklist, then a "stop if" rule, then a one-line log template for my notebook.Daily actions.
End-of-week deliverable. A named assistant plus a 1-page how-to.
Month 1 checkpoint
Month 2. Specialize + proof (Build + Show)
Now you pick one path: AI content creation, AI automation, or AI agents. All three are fine. All three at once will not finish in 90 days.
05Week 5. Specialize setup: pick one track
Goal. Write 5 lines: why this track, who the client is, what proof you will build.
What to learn. What each track sells, and what proof looks like.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You are a career-track coach for AI freelancing beginners.
Goal: Help me pick ONE track for the next 60 days: AI content, AI automation, or AI agents.
Context: My background: [paste]. Proof I already have: [paste]. Time per weekday: [60-90 min]. I must avoid switching tracks.
Output format:
- Recommended track + why (5 lines max)
- Why not the other two (1 line each)
- First proof I should build in 14 days
Be direct.Role: You help me name a niche problem.
Goal: Turn my vague “I do AI” into a problem statement a buyer understands.
Context: Track: [content / automation / agents]. People I know: [list]. Pain I have seen: [paste].
Output format:
1) Who (specific)
2) Problem in their words
3) Desired outcome
4) What I will deliver in the first paid project
No buzzwords.Role: You plan proof artifacts.
Goal: Sketch one before/after proof I can finish in two weeks.
Context: Track: [ ]. Niche: [ ]. Materials I already have: [ ].
Output format: proof title, before state, after state, steps (max 7), what screenshot or file I will keep.Daily actions.
End-of-week deliverable. Track choice memo + niche one-liner + first proof sketch.
06Week 6. Offer skeleton and a minimal stack
Goal. Fill an offer skeleton and open free-path tool accounts.
What to learn. Minimum stack. Free paths first. Paid plans only if free blocks the one workflow you need.
Free resources.
- Content: Google Workspace AI, Zapier: how to use ChatGPT
- Automation: n8n docs, Make getting started, Zapier Learn
- Agents: Anthropic tool use, OpenAI agents guide, OpenAI tools guide
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You are an offer editor.
Goal: Draft a first-project offer skeleton (not a full sales page).
Context: Track: [ ]. Who: [ ]. Problem: [ ]. Deliverable: [ ]. Timeline: [days].
Output format:
- Who it is for
- Problem
- Deliverable (what they receive)
- What is out of scope
- First-project price band as a teaching range (label it as a starting pilot range, not a promise)
Keep it one page.Role: You are a ruthless stack minimalist.
Goal: Cut my tool list to the minimum free path for my track.
Context: Track: [ ]. Tools I opened: [ ]. What I need to ship in 2 weeks: [ ].
Output format: Keep / Drop / Maybe later. Then a 5-step setup order for this week.Role: You coach a lightweight competitor scan.
Goal: Extract patterns I can learn from, not copy.
Context: Track: [ ]. 5 examples I found: [paste links or notes].
Output format:
- Common promises
- Common deliverables
- Proof types they show
- Gaps I could fill with honesty
No insulting language. No copying their text.Daily actions.
End-of-week deliverable. Offer skeleton draft + tool accounts ready.
07Week 7. Track deep work A: first artifact
Goal. Ship the first track-specific artifact that could enter a portfolio.
What to learn. The concrete work of your path: calendar and voice, a live workflow, or a small agentic flow.
Free resources by track.
- Content. Google Workspace AI, Zapier: how to use ChatGPT
- Automation. n8n docs, n8n courses, Make getting started, Zapier Learn
- Agents. Anthropic tool use, Claude tool use build guide, OpenAI agents guide, OpenAI tools guide
Use the prompt fork that matches your track.
Role: You are a content calendar architect.
Goal: Build a 2-week calendar from my existing posts and niche.
Context: Niche: [ ]. Existing posts: [paste titles or links]. Channels: [IG / LinkedIn / blog]. Voice: [3 words].
Output format: day-by-day calendar with goal, angle, CTA, and source post to remix. No new fake case studies.Role: You extract brand voice rules.
Goal: Turn my past writing into a short voice card.
Context: 5-10 samples: [paste]. Words I never want: [list].
Output format: Voice card with Do / Don't / Example lines / Checklist before publish.Role: You remix my old posts into new drafts.
Goal: Produce 5 drafts that sound like me.
Context: Voice card: [paste]. Source posts: [paste]. Format: [caption / carousel outline / email].
Output format: Draft 1-5 with hook, body, CTA. Flag any invented facts.Role: You are an automation analyst.
Goal: Spec one live workflow I can build on a free plan this week.
Context: Tools I can connect: [list]. Pain: [paste]. Trigger event: [ ]. Desired output: [ ].
Output format: Trigger → steps → output → failure cases → manual fallback. Keep it buildable in one week.Role: You map triggers to actions.
Goal: Turn my messy process into trigger/action pairs.
Context: Process notes: [paste]. Platform preference: [n8n / Make / Zapier].
Output format: table of Trigger | Action | Data fields | Free-plan risk.Role: You design test cases for automations.
Goal: Give me 5 tests before I call the workflow “live”.
Context: Workflow spec: [paste].
Output format: Test name, input, expected result, how I will verify, when to stop and fix.Role: You help me define one small agentic workflow.
Goal: Write a role spec for one agent-like workflow with tools.
Context: Job to complete: [ ]. Tools allowed: [browse / files / calculator / other]. Must ask me before: [actions].
Output format: Role, goal, tools, step policy, stop conditions, success check. Keep it small.Role: You tutor tool-use planning from official concepts.
Goal: Plan which tool calls my workflow needs, in plain words.
Context: Task: [ ]. Official ideas I read: [short notes]. Constraints: [ ].
Output format: planned steps, which step needs a tool, what I verify after each tool result.Role: You write guardrails.
Goal: Add safety and quality rails for my small agentic workflow.
Context: Role spec: [paste]. Data I must not expose: [ ].
Output format: hard rules, ask-me-first list, refusal style, logging template.Daily actions.
End-of-week deliverable. Artifact A: calendar pack, or live workflow, or agentic demo doc.
08Week 8. Track deep work B + proof card
Goal. Second artifact plus a before/after proof card.
What to learn. How to write proof so a stranger understands the result without you in the room.
Free resources.
- n8n workflows (automation examples)
- LangChain short course (optional, agents-adjacent)
- Make docs
- Zapier Help
Role: You write proof cards for portfolios.
Goal: Turn my work into one before/after proof card.
Context: Track: [ ]. Before: [ ]. After: [ ]. What I did: [ ]. Constraints: do not invent metrics.
Output format: Title, Before, After, What I did (5 bullets), Artifact link/filename, One honest limit.Role: You narrate short case stories.
Goal: Write a 150-200 word case in plain language.
Context: Proof card: [paste]. Reader: potential first client.
Output format: Situation → action → result → what I’d do next. No hype adjectives.Role: You run a weekly improvement loop.
Goal: Find one upgrade for next week without changing tracks.
Context: What shipped: [ ]. What failed: [ ]. Time left in month: [ ].
Output format: Keep / Fix / Drop. Then one experiment for 3 days only.- Content. Publish or schedule 3 pieces. Run a voice-card consistency check.
- Automation. Document the workflow, error cases, and a time note.
- Agents. Run the same flow twice. Write failure modes.
Daily actions.
End-of-week deliverable. Artifact B + 1 proof card (before/after).
Month 2 checkpoint
Month 3. First client (Show + Sell)
Now you show and talk. Portfolio, one-page offer, 10 real conversations, then a first paid project or paid pilot.
09Week 9. Portfolio of 3 case-style proofs
Goal. Three cases: before, after, what you did.
What to learn. A portfolio is proof, not a screenshot dump.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You format case studies for a first-client portfolio.
Goal: Turn my notes into 3 short cases with the same structure.
Context: Notes for case 1/2/3: [paste]. Track: [ ]. No invented metrics.
Output format per case: Title, Client-type (or self/practice), Before, After, What I did, Artifact, Honest limit.Role: You write a portfolio index page.
Goal: Create a one-screen index that links to my 3 cases.
Context: Cases: [titles]. Who I want to attract: [ ].
Output format: headline, 3-line positioning, case list with one-line each, contact line.Role: You review my portfolio like a skeptical buyer.
Goal: Find gaps that would stop a first yes.
Context: Portfolio draft: [paste].
Output format: blockers, easy fixes this week, what to stop explaining.Daily actions.
End-of-week deliverable. Portfolio page/doc with 3 cases.
10Week 10. Offer one-pager
Goal. One page: who / problem / deliverable / price for the FIRST project.
What to learn. First-project pricing is an honest pilot range, not a fake guarantee. Teaching examples only: a small pilot might be one workflow, one content pack, or one agent trial with limited scope. Local market and your proof set the number. No promises here.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You draft first-project offer one-pagers.
Goal: Write a clear one-pager for my first paid pilot.
Context: Who: [ ]. Problem: [ ]. Deliverable: [ ]. Timeline: [ ]. Out of scope: [ ]. Price band I am considering for a first pilot: [range as a starting point, not a promise].
Output format: one page with sections Who / Problem / Deliverable / Process / Price for first project / Next step.
Tone: direct, warm, no hype.Role: You are a pricing sanity coach for first projects.
Goal: Stress-test my first-project price without inventing income claims.
Context: Offer: [paste]. Hours I expect: [ ]. Costs: [tools]. My experience level: beginner with proof.
Output format:
- Is scope small enough for a first yes?
- What would make this price unfair to me or them?
- A pilot framing sentence I can say out loud
No guarantees. No “you will earn” language.Role: You write calm objection replies.
Goal: Prepare short replies for common first-project objections.
Context: Offer: [paste]. Objections I fear: [list].
Output format: objection → 3-sentence reply → one clarifying question.
No pressure tactics.Daily actions.
End-of-week deliverable. 1-page offer (PDF or Notion).
11Week 11. Outreach: 10 real conversations
Goal. 10 real conversations on LinkedIn, Instagram, or email.
What to learn. Conversation beats a pitch blast. The goal is a reply and a clear next step, not an instant close.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You write warm outreach openers.
Goal: Draft a short message that starts a conversation, not a pitch dump.
Context: Person: [name/role]. Why them: [specific]. My proof: [one line]. Ask: [15-min chat / feedback on offer].
Output format: 3 variants (LinkedIn / Instagram DM / email), each under 80 words, with one clear ask.Role: You write soft follow-ups.
Goal: Follow up once without guilt-tripping.
Context: Original message: [paste]. Days since send: [ ]. New value I can add: [one line].
Output format: one follow-up + when to stop if no reply.Role: You build discovery-call agendas.
Goal: Give me a 20-minute agenda for a first conversation.
Context: Offer one-pager: [paste]. What I must learn: [ ].
Output format: minute-by-minute agenda, 5 discovery questions, closing options (pilot / not now / referral).Daily actions.
End-of-week deliverable. Outreach log (10 sent + notes).
12Week 12. Deliver the first paid project or paid pilot
Goal. Close one paid project or paid pilot, deliver, ask for a testimonial.
What to learn. Scope lock, delivery checklist, close-the-loop message. If there is no yes yet, repeat outreach the same week with a tighter niche.
Free resources.
Prompts below are learning tutors. They teach process. They are not magic money lines.
Role: You draft simple scope-lock notes for a first pilot.
Goal: Write a short agreement outline a non-lawyer can use as a working note (not legal advice).
Context: Client: [ ]. Deliverable: [ ]. Timeline: [ ]. Price: [ ]. Out of scope: [ ].
Output format: parties, deliverable, timeline, revision limit, payment timing, out of scope, how change requests work.
Plain language.Role: You run delivery checklists.
Goal: Turn my pilot into a day-by-day delivery plan.
Context: Scope: [paste]. Hours available: [ ]. Risks: [ ].
Output format: daily checklist, quality checks before send, client update template.Role: You write soft testimonial asks.
Goal: Ask for a short testimonial after delivery.
Context: What I delivered: [ ]. Result they saw: [ ]. Channel: [email / IG / LinkedIn].
Output format: 2 short asks + 3 prompt questions they can answer in bullets.Daily actions.
End-of-week deliverable. Delivered pilot + payment note + testimonial ask sent.
Month 3 checkpoint
Common mistakes before day 90
- Collecting prompts instead of shipping a weekly deliverable.
- Switching tracks every week.
- Pitching before proof.
- A foggy offer: “I do AI”.
- Skipping the diary. No real tasks means no real assistant.
- Waiting for a perfect stack or paid plans.
Quick summary
- Saving is not learning. The loop is Learn, Build, Show, Sell.
- Weekdays: 60-90 minutes. Learn, build, log one line.
- Month 1 foundations + one assistant. Month 2 one track + proof. Month 3 portfolio, offer, outreach, delivery.
- Your first paid client comes from daily work, not a magic prompt.
