Score the proposalbefore you send it.
The book's five-criterion rubric, turned into a pre-send checkpoint. Score one proposal 1–5, get the honest verdict — SUBMIT, REVISE, DO NOT SEND, or START OVER — and the one criterion to fix first. The verdict comes from your own marks, not a hidden benchmark.
A strong-looking proposal can still be a delete.
of proposals a busy client deletes in under ten seconds — most for a structural reason the sender never diagnosed.
criteria decide whether a proposal survives: the opening, the proof, the approach, the ask, and whether it reads machine-generated.
weak criterion — or one bot tell — can sink an otherwise high-scoring proposal. The average hides it; a gate catches it.
Score a proposal. Read the verdict.
1 generic … 5 names the client's exact problem in their words
1 none … 5 a matched case study with a quantified result
1 generic process … 5 a real point of view
1 vague … 5 one specific, low-friction next step
1 clean … 5 heavy machine-generation signals (low is good)
The verdict is the book's, computed from your five marks. Scores the proposal, not any person.
The four content criteria score 1 to 5. Bot Risk is a detection mark — 1 is clean, 5 is heavy machine signals — so the Probe counts six minus your bot mark. The five contributions sum to a score out of 25, which sets the band.
Then two gates can only lower the verdict, never raise it. A single criterion below 3 holds the proposal at REVISE. A Bot Risk mark of 4 or 5 holds it at DO NOT SEND. A strong total cannot rescue a broken dimension — which is exactly the failure the rubric exists to catch.
Try the sample: every content criterion is a perfect 5 and the total is 22 of 25, deep in SUBMIT range — yet the verdict is DO NOT SEND, because the bot mark is 4.
The same verdict, offline and runnable.
The zero-dependency Python engine scores one proposal or a whole queue from a CSV and produces the identical verdict to the demo and the workbook — verified to the mark. Here it scores a proposal that looks ready and isn't:
PROPOSAL QUALITY PROBE ================================================== Opening Specificity 5/5 Proof Relevance 5/5 Approach Differentiation 5/5 Call to Action Clarity 5/5 Bot Risk Score 4/5 -------------------------------------------------- Total score 22/25 Verdict DO NOT SEND Fix this first Bot Risk Score (4/5) Gate weak criterion (Bot Risk Score) -> held to REVISE Gate bot risk 4/5 -> held to DO NOT SEND
Three rules the Probe never breaks.
The score is built only from the five criteria you enter, using the book's own thresholds. No industry multiplier, no invented lift, nothing assumed about the client.
A weak criterion or a bot tell can only lower the verdict. The Probe will never promote a proposal above what its weakest dimension earns.
It grades the proposal as a document, not the person who wrote it — and reproduces every number in the workbook so you can audit it.
A pre-send checkpoint, not a ghostwriter.
- A deterministic scorer for the book's five-criterion rubric.
- A gate that catches the one weak criterion an average hides.
- A bot-risk check that holds back machine-flagged proposals.
- A queue scorer — run a whole week of proposals at once.
- A proposal writer — you bring the proposal, it grades it.
- A predictor of win rate or revenue for any single bid.
- A scorer of people — it grades the document only.
- A replacement for the book's method; it is the companion to it.
Scope. The Probe scores a proposal as an artifact from the marks you enter; it does not score, rank, or assess any person, and it is not legal, financial, or career advice.
For the freelancer who stopped gambling.
- Freelancers and independents bidding on Upwork, by referral, or by direct outreach.
- Anyone who read The AI Proposal System and wants the rubric as a daily checkpoint.
- Solo consultants who want a second opinion on a proposal before it goes out.
- —Not for scoring or ranking people — it grades a document, not a candidate.
- —Not an enterprise RFP tool — for formal bids, see the RFP & Proposal Response Engine.
- —Not a guarantee of any win; it improves the proposal, not the odds of a given client.
Build the rest of the pipeline.
The enterprise sibling — runs a full RFP response the Shipley way when the proposal is a formal bid.
Builds the competitor and objection intel that sharpens the approach and proof sections.
Carries the proposal in your own voice — the outreach layer around the document.
Answers before you ask.
It is a deterministic scorer for one freelance or independent proposal, built on the five-criterion Proposal Quality Scoring Rubric from The AI Proposal System. You score the proposal 1–5 on Opening Specificity, Proof Relevance, Approach Differentiation, Call to Action Clarity, and Bot Risk, and the Probe returns the book's verdict — SUBMIT, REVISE, DO NOT SEND, or START OVER — and names the one criterion to fix first.
The four content criteria score 1 (weak) to 5 (strong); Bot Risk is a detection mark where 1 is clean and 5 is heavy machine-generation signals, so the Probe counts six minus your bot mark toward the total. The five contributions sum to a score out of 25, which sets the band: 20–25 SUBMIT, 15–19 REVISE, 10–14 DO NOT SEND, 5–9 START OVER. Two gates can then lower the verdict but never raise it.
The weak-criterion gate holds any proposal at REVISE if a single criterion scores below 3, even when the total would otherwise pass — a strong average cannot rescue one broken dimension. The bot-risk gate holds a proposal at DO NOT SEND if the Bot Risk mark is 4 or 5, because a proposal a trained client will flag as machine-generated should not go out, however polished the rest reads.
No. The verdict is computed only from the five marks you enter, using the book's own thresholds. There is no industry multiplier, no invented lift estimate, and nothing about the client or platform is assumed. The workbook reproduces every number so you can audit the math, and the runnable engine produces the same result offline.
You get a runnable zero-dependency Python engine that scores a single proposal or a whole queue from a CSV, a workbook that reproduces the scoring and both gates with a worked example, two playbooks (how to score honestly and how to fix the weakest criterion), and a sample queue. It runs offline; no account, no install beyond Python for the engine, and the workbook opens in Excel, Google Sheets, or Numbers.
No — it is the companion. The book teaches why each criterion matters and how to raise it; the Probe turns the book's rubric into a pre-send checkpoint you can run on every proposal and across a queue. It scores the proposal as an artifact, not any person, and it is not legal, financial, or career advice.
Stop sending on hope.
Score it first.
One purchase, lifetime access, 12 months of updates. $79, once.
Scope. The Probe scores a proposal as an artifact from the marks you enter; it does not score, rank, or assess any person, and it is not legal, financial, or career advice.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai