AI Guides › Playbooks
By Nigel Guy · 6 min read
Most people with no portfolio do one of two things: polish a profile for a fortnight before applying, or fire the same pasted paragraph at thirty jobs. The first feels productive and produces nothing a client can see. The second spends money, because each proposal costs Connects, and tells the client you did not read their post.
The rule: don't ask a stranger to trust you; hand them a small piece of finished work about their own job, and make the proposal about that.
A note on the story behind this guide. We were given a claim that a beginner landed a first paid Upwork gig this week with no prior portfolio. We cannot verify that individual case, so we have not retold it. What follows is the durable mechanism, built from Upwork's own help pages, plus a clearly hypothetical example.
Five rungs. Climb them in order for every application.
Upwork's own advice on managing Connects is worth treating as a filter. Before spending any, check:
A first gig is easiest on a small, specific job from a client with a spending history, not a vague "build me an AI" brief with a hundred proposals.
Copy the job post into a document and underline three things: the outcome they want, the thing they mention twice, and any constraint (tone, tool, deadline). Your proposal answers those three, in their words. AI is good at this step, as the prompt below shows.
Produce one slice of the real work: the first five FAQ entries, a rewritten paragraph, a one-page plan. Keep it short enough to read in a minute. Do not use their confidential details if the post has none; use what is public in the listing. Upwork's guidance lets you add attachments such as work samples, so the sample goes there.
Upwork's advice is to open with the client's needs, keep the cover letter to two or three short paragraphs, include relevant proof, and close with a clear next step. Use this order:
Record the job, Connects spent, whether you were viewed, and any reply. After ten proposals, change one variable only.
Fill in the bracketed parts. Paste the job post yourself; the model cannot see Upwork.
You are an experienced freelance consultant helping a beginner apply for a job.
Context: I offer [MY_ONE_LINE_SERVICE]. My relevant experience is [MY_EXPERIENCE_OR_NONE]. Below is a client's job post.
JOB POST:
[PASTE_JOB_POST]
Goal: help me understand what this client really wants so I can send a short, relevant proposal.
Steps:
1. State the outcome the client wants in one sentence.
2. List what they repeat or stress, and any constraints.
3. List what is unclear or missing that I should ask about.
4. Say honestly whether this job fits my service, and why.
5. Suggest one small sample I could make in under an hour to show I can do it.
Rules: use only what is in the post. If anything I have given you is missing, ask me before answering. Do not invent facts about the client.
Before you answer, check that each point can be traced to the post.
Format: a short numbered list under those five headings.
You are a careful editor who writes plain, specific proposals.
Inputs:
- Job post: [PASTE_JOB_POST]
- My sample (summary or text): [MY_SAMPLE]
- My price and timeline: [PRICE_AND_TIMELINE]
- My real experience: [MY_REAL_EXPERIENCE]
Goal: draft a cover letter of at most three short paragraphs: (1) the client's outcome in their words, (2) the sample and what it shows, (3) price, timeline, scope, one question, and a call to action.
Rules: no claims about experience, results or clients that I have not given you. No flattery, no generic openers, no "I am the perfect fit". Plain British English. If an input is missing, ask me for it instead of guessing.
Self-check before answering: does it mention something only this client's post would contain? Is every claim in my inputs? Cut anything that fails.
Output the letter only, then a two-line list of what you were unsure about.
Asha has no portfolio. She offers "customer FAQs from messy notes". She finds a small online shop whose post says returns questions keep reaching their inbox. Rung 2 shows the outcome: fewer returns emails. Rung 3: she drafts five FAQ answers on returns from the public returns page, flagging two points she wants confirmed. Rung 4: her letter names the returns problem, attaches the five answers, quotes a fixed price for twenty entries, and asks which returns rule is current. That is a different thing from "I am a hard-working AI specialist". Whether it wins is up to the client; the point is it gives them something to judge.