aaron

Quick win (under 5 minutes): take two short passages from different documents, paste them into any embedding tool or run a one-line script to compute embeddings, then calculate cosine similarity. If the similarity is higher for related passages than unrelated ones, your pipeline basics work.

Good point — you’re focused on practical similarity across diverse documents, not just toy examples. That focus keeps the project deliverable and measurable.

The problem: Diverse documents (PDFs, emails, web pages, transcripts) vary in length, structure, and language. Naive search (keyword matching) fails to surface semantically relevant results.

Why it matters: Better similarity search reduces time to insight, increases user trust, and drives measurable outcomes like faster support resolution, higher conversion, or quicker research synthesis.

Experience-based lesson: The highest ROI comes from standardizing chunking + metadata, using normalized embeddings, and validating with real user queries — not from chasing the latest model.

1. What you’ll need: a small sample of each doc type (10–100), an embedding model/service, a vector store (local or cloud), and a simple query UI or script.
2. How to set up (step-by-step):
   1. Extract text from files and preserve source metadata (title, date, type).
   2. Chunk: 200–500 tokens with 20% overlap for context.
   3. Compute embeddings for each chunk and normalize vectors.
   4. Index vectors into your vector store with metadata tags.
   5. For a query: embed the query, retrieve top-N by cosine similarity, re-rank by metadata or a lightweight cross-encoder if needed.
3. What to expect: early retrieval precision ~0.6–0.8 depending on dataset; iterative tuning of chunk size and reranking improves it.

Copy-paste AI prompt (use for chunking + summaries):

“You are a document processor. Given the following text, split it into chunks of about 300 words each with ~20% overlap. For each chunk, output a JSON object with fields: id (unique), chunk_text, short_summary (1–2 sentences), primary_keywords (3–5). Also return document-level metadata: title and source_type.”

Metrics to track (start with these):

• Top-5 Precision@5 (relevance of the first 5 results)
• Mean Reciprocal Rank (MRR)
• Average query latency
• User satisfaction score (binary thumbs or 1–5 rating)

Common mistakes & fixes:

• Too-large chunks —> split more aggressively to improve recall.
• No metadata —> add source and date to filter and rerank.
• Mismatched languages —> language-detect and index separately or use multilingual model.
• Not normalizing vectors —> normalize to cosine similarity for consistent scores.

1-week action plan:

1. Day 1: Collect 10–50 representative docs and extract text + metadata.
2. Day 2: Implement chunking and run the provided prompt to produce summaries/keywords.
3. Day 3: Generate embeddings and index into a vector store.
4. Day 4: Build a simple query script/UI and run baseline queries.
5. Day 5: Measure Precision@5, MRR, latency; collect qualitative feedback from 3 users.
6. Day 6: Tune chunk size, overlap, and apply simple reranker.
7. Day 7: Re-measure and document improvements; plan next scale steps.

Your move.

5-minute win: Create a “Favorites” subfolder and drop in one prompt that already worked. Add this line at the top: “Last used: [TODAY] — Why it worked: [ONE LINE].” You’ve just started your high-confidence shelf.

The problem: Most prompt libraries collect prompts; they don’t produce consistent outcomes. What’s missing is standardization: clear variables, a fixed output format, and a simple way to test and retire duds.

Why it matters: When outputs are predictable, you cut drafting time by half, reduce rewrites, and hand off tasks with confidence. This is the difference between a folder of ideas and a dependable system.

Lesson from the field: The teams that win use “prompt shells” (repeatable templates with variables) plus an output checklist. They track winners and prune the rest. Simple, boring, effective.

What you’ll need

• One folder: Prompt Library, with two subfolders: Favorites and Sandbox.
• Filename pattern: Category — Outcome — v# — YYYYMMDD.
• A one-page “Prompt Card” template (below).

Prompt Card (copy-paste template)

Purpose: [What this produces, e.g., 120-word newsletter intro]Audience: [Who it’s for]Voice DNA (5 bullets): [e.g., Warm, direct, no jargon, short sentences, one CTA]Variables: [Topic], [Offer/CTA], [Length], [Deadline/Timeframe]Output Checklist: [Word count], [3 bullets of benefits], [1 example], [Single CTA], [Plain text]Status: Draft/Approved/Retired — Last used: [Date] — Owner notes: [Why it works]

Steps to build a library that performs

1. Create 5 prompt shells tied to outcomesPick your highest-frequency tasks: Newsletter intro, Client update email, Social post, Report summary, Blog intro. Each gets a Prompt Card using the template above.
2. Lock the output before the proseAdd a tight Output Checklist to every shell (counts, bullets, CTA). Structure beats tone for consistency.
3. Embed your Voice DNA onceWrite five bullet cues that define your brand voice. Paste the same five into every shell to maintain consistency across tasks.
4. Test fast: A/B two small variationsDuplicate the shell, change one thing (tone or structure), run both, pick the clearer draft. Move the winner to Favorites. Archive the loser in Sandbox with a one-line reason.
5. Version with intentIncrement versions only when you change structure or variables. Cosmetic tweaks don’t get a new version.
6. Add a 30-day review tagAt the top of every Prompt Card, include “Review by: [Date+30].” Anything not used by then is retired or merged.

Copy-paste prompt (Newsletter intro shell)

Act as a senior marketing writer. Use the following constraints and produce one clean draft.Goal: Write a [LENGTH]-word newsletter intro that previews three practical tips about [TOPIC].Voice DNA: Warm, direct, no jargon, short sentences, confident, one CTA.Structure: 1) Hook (1–2 sentences). 2) Three bullets of benefits (no fluff). 3) One example (1 sentence). 4) Clear CTA that starts with “Try this this week: [SIMPLE ACTION]”.Rules: Plain text, no headings, keep within [LENGTH] words ± 10%. Avoid clichés. Zero emojis.Now ask me any missing variables before writing.

Copy-paste refinement prompt (turn any rough draft into final)

Improve the draft below to match the Output Checklist exactly: [Word count], [3 bullets of benefits], [1 example], [Single CTA], [Plain text]. Keep the Voice DNA: warm, direct, short sentences, confident. Remove filler and redundant phrases. Output only the final text.

What to expect: Once shells are in place, first drafts arrive in minutes and need light edits. Favorites become your “autopilot” for recurring work. Sandbox stays messy by design — but contained.

Metrics to track (weekly, simple)

• Time to first usable draft (minutes) — target: under 5 for common tasks.
• Edit ratio — number of edits per draft; aim to reduce 30–50% in 30 days.
• Reuse rate — % of tasks completed with a Favorite shell; goal: 70%+.
• Approved shells — count of shells marked “Approved”; goal: 8–12 within a month.
• Cycle time — start to send/publish; aim for a 40% reduction.

Common mistakes and quick fixes

• Too many categories — fix: cap at 5–8; merge anything overlapping.
• Vague output — fix: add counts and checklist; structure first, style second.
• No variables — fix: bracketed slots [TOPIC], [LENGTH], [CTA] in every shell.
• Hoarding drafts — fix: Draft/Approved/Retired status; prune monthly.
• Inconsistent voice — fix: the same 5-line Voice DNA pasted into every prompt.
• One-off brilliance lost — fix: save winning outputs as Example Output inside the Prompt Card.

1-week action plan

1. Day 1 (20 min): Create folders (Prompt Library, Favorites, Sandbox). Paste the Prompt Card template into a new note.
2. Day 2 (20 min): Build 5 shells tied to outcomes. Add Voice DNA and Output Checklist to each.
3. Day 3 (20 min): Run one real task through two shells (A/B). Move the winner to Favorites with a one-line note.
4. Day 4 (15 min): Standardize filenames and add Review by: [Date+30] to each card.
5. Day 5 (15 min): Add one Example Output to each shell (the best real result so far).
6. Day 6 (15 min): Use a Favorite for a real deliverable. Measure time to first draft and edit ratio.
7. Day 7 (10 min): Retire one stale prompt, promote one shell to Approved, and update metrics.

Bottom line: Stop collecting prompts. Build shells with variables, lock the output, test quickly, and promote only the winners. Your library becomes a production system, not a scrapbook.

Your move.

Nice: that five-minute quick win is exactly right — starting beats perfection.

The problem: vague aims (“understand”, “know”) lead to fuzzy lessons and impossible-to-measure results. You need clear objectives and success criteria that map directly to evidence — fast.

Why this matters: measurable objectives make assessment simple, save prep time, and give learners confidence. They also let you prove improvement quickly to parents and administrators.

Quick lesson from practice: teams who draft objectives with AI then run a 5-minute exit ticket for two classes get actionable data the same day. That lets them adjust teaching the next lesson — not next term.

Step-by-step (what you’ll need, how to do it, what to expect)

1. What you’ll need: one vague aim (one line), learner level (age/grade), time limit for tasks (5–15 minutes), and your AI chat tool.
2. How to do it:
   1. Paste this prompt (below) into AI. Get back: a SMART objective, 3 “I can” success criteria, a 5‑question exit ticket with timings, and lower/higher cognitive variants.
   2. Pick the objective, tweak one verb or time limit, convert success criteria into a checklist for the class.
   3. Run the exit ticket at lesson end, collect scores, and mark who met each criterion.
3. What to expect: a ready-to-use objective, student-friendly criteria, and a 5-minute assessment you can use immediately.

Copy-paste AI prompt (use this)

Convert this learning aim into: 1) a single measurable objective (one sentence, SMART), 2) three student-facing “I can” success criteria, 3) one lower-cognitive and one higher-cognitive version of the objective, and 4) a 5-question exit ticket (2 short answers, 2 quick tasks, 1 multiple choice) with suggested timing per question. Learner level: [insert age/grade]. Aim: “[insert vague aim]”. Keep language simple and actionable.

Metrics to track (start here)

• % of students meeting each success criterion (per class, per lesson).
• Average exit-ticket score.
• Time to produce a usable objective (goal: <10 minutes).
• Change in scores after one instructional tweak (target: +10–20%).

Common mistakes & fixes

• Using non-observable verbs (“understand”) — fix: replace with describe/explain/create/compare.
• Success criteria too vague — fix: add quantity or time (“list three”, “in 10 minutes”).
• No immediate assessment — fix: always attach a 3–5 question exit ticket.

1-week action plan (clear next steps)

1. Day 1: Pick three vague aims you use regularly.
2. Day 2: Run the prompt for each aim and select objectives.
3. Day 3: Create success-criteria checklists and the exit tickets.
4. Day 4–5: Teach one lesson and run the exit ticket; collect scores.
5. Day 6: Review metrics, tweak one verb/time, retest with next class.
6. Day 7: Save the best templates and schedule reuse for next unit.

Your move.

Quick win (under 5 minutes): Open your AI chat and paste the prompt below to get a printable safety card and image briefs for “Skittles Rainbow.” You’ll have a ready-to-use checklist before the kettle boils.

Hook: Turn messy internet instructions into a standard, repeatable one-pager with clear visuals and a no-guesswork safety checklist.

The problem: Home experiments are inconsistent — missing quantities, vague hazards, and cluttered pictures. That slows you down and increases risk.

Why it matters: Clear, age-appropriate visuals plus explicit “do/don’t” items cut confusion and reduce incidents. It also makes supervision easier to delegate.

What I’ve learned: Constraint your prompts. Ask for single-action diagrams, IF–THEN emergency steps, and a “Do Not Use” list. Cap steps at 7, force quantities, and set a reading level. This produces reliable, repeatable outputs.

Copy-paste prompt (master template)

“Turn the following home science activity into a one-page Safety + Steps card. Audience: children aged [AGE RANGE] with [SUPERVISION LEVEL]. Output sections: 1) Objective (1 sentence, plain words), 2) Materials (numbered, exact quantities, common household substitutes), 3) Procedure (5–7 numbered steps, one action per step, short sentences, parent voice), 4) Hazards (bullet list with risk level low/medium/high + one-line mitigation), 5) Required PPE, 6) Emergency actions (IF–THEN in two steps), 7) Do Not Use list (items to avoid for home use), 8) Cleanup instructions, 9) Image briefs (3, each: composition, top-down or close-up, 2–3 labels max, single-action line diagram, bright high-contrast, large labels). Constraints: keep language at grade [4–6], avoid jargon, include a final ‘Stop here before next activity’. Here is the activity: [PASTE YOUR CURRENT STEPS AND MATERIALS].”

Bonus prompt (image generator brief, copy-paste)

“Single-action line diagram, top-down flat lay of [WORKSPACE OR ACTION]. High contrast, thick outlines, 80% empty space. 2–3 large labels: [LABEL1], [LABEL2], [LABEL3]. One bold arrow to show the motion if needed. Plain white background. Avoid clutter, avoid extra props, avoid small text.”

3-minute example (Skittles Rainbow)

• Paste the master template and set: ages 6–10, close adult supervision.
• Materials to include: plate, warm water (1 cup), Skittles (20), paper towels. Substitutes: any colored sugar-coated candy.
• Expect: 6 steps, hazards like “choking (low) → keep candy away from mouths during setup,” PPE = none required, emergency = IF candy is swallowed → encourage coughing; IF choking persists → call emergency services.

Step-by-step (build once, reuse forever)

1. Standardize inputs: Write one-sentence objective; list materials with quantities and safe substitutes; set age and supervision level.
2. Generate text: Run the master prompt. Ask the AI to keep each step to one short sentence and add cleanup.
3. Create visuals: Use the three image briefs to produce: a top-down workspace layout, a single-action close-up of the trickiest step, and a labeled materials layout.
4. Assemble the one-pager: Title, age/supervision, materials (left), steps (center), safety + emergency (right). Keep white space. Large labels.
5. Test with a fresh pair of eyes: Ask a helper to follow the page without help. Record every question or hesitation.
6. Revise and version: Fix unclear steps, add missing hazards, tighten language. Save as v1.1 with date.

Insider tricks that raise quality

• Single-action only: One motion per image. Ask explicitly for “no more than 3 labels.”
• IF–THEN safety: Replace vague warnings with triggers and actions (fast to follow under stress).
• Substitute-first materials: Parents complete more experiments when a substitute is listed next to each item.
• Reading level lock: Specify “grade 4–6” for ages 8–12. It’s the difference between confident and confused.
• Stop points: Insert “Stop here before next activity” to prevent rushing.

QA prompt (use after the first draft)

“Act as a safety and clarity reviewer for a home science one-pager. Score 0–5 on: Clarity of steps, Hazard coverage, Age fit, Visual specificity, Emergency readiness, Readability. For any score under 4, propose 1–2 precise edits. Confirm if a first-time adult could supervise without extra questions.”

Metrics to track (results and targets)

• Time to usable draft: target < 20 minutes.
• Test-user questions: target 0–2 on first run.
• Readability: grade 4–6 for ages 8–12.
• Hazard coverage: every identified hazard has a mitigation and IF–THEN action.
• Incident rate: 0 incidents across 3 supervised trials before sharing.
• Print legibility: all labels readable at arm’s length.

Common mistakes and fast fixes

• Overloaded images: If an image shows two actions, split into two. Ask AI: “single-action diagrams only.”
• Missing quantities: Force numbers. Add “materials must include exact quantities and common substitutes.”
• Vague emergencies: Swap “be careful” for IF–THEN steps with two clear actions.
• Too many steps: Cap at 7. Ask for “one short sentence per step.”
• Kid-unfriendly language: Lock reading level and request “parent voice, plain words.”

One-week action plan with KPIs

1. Day 1: Pick 3 safe experiments (no heat, no glass). Draft objective, materials with substitutes, and age/supervision. KPI: 3 complete inputs.
2. Day 2: Run the master prompt for all three. KPI: 3 first-draft one-pagers.
3. Day 3: Generate three visuals per experiment using the image prompt. KPI: 9 clear diagrams.
4. Day 4: Assemble printables and print. KPI: 3 pages legible at arm’s length.
5. Day 5: Test with a helper who hasn’t seen them. Log questions and any hesitation. KPI: ≤2 questions per page.
6. Day 6: Revise using the QA prompt; tighten hazards, IF–THEN, and steps. KPI: All QA scores ≥4/5.
7. Day 7: Supervised trial with children. Record incidents (target 0) and time from start to finish. Finalize v1.1 PDFs.

What to expect: Clear, printable one-pagers in a day; reliably safer, repeatable experiments in a week; confidence to hand off supervision without hand-holding.

Your move.

Turn that 3-bullet outline into a newsletter your audience actually reads — fast.

The problem: You have an outline, but not the time or confidence to craft tight copy, a compelling subject line, and a clear CTA. AI can do the heavy lifting — if you prompt it correctly.

Why this matters: A polished newsletter increases opens, reads, and clicks. For a small business or solo professional, that equals leads, bookings, and revenue. Poor execution wastes the list and your time.

What I’ve learned: AI won’t replace your judgment. Treat it like a skilled editor: give context, constraints, and the outcome you want. You’ll get useful copy in minutes if your prompt is precise.

What you’ll need:

• Rough outline (3–8 bullets)
• Audience note (who, age 40+, non-technical)
• Desired length (e.g., 300–450 words)
• Tone (warm, confident, actionable)
• One primary CTA

1. Prepare the outline: Clean bullets into 3–5 points with any examples or stats you want included.
2. Use a structured AI prompt: Paste the outline and ask for subject lines, preview text, full newsletter, and a 1-line CTA.
3. Edit for voice: Read and tweak wording to match your brand — shorten sentences, swap jargon, add a personal sentence if appropriate.
4. Optimize subject lines: Pick 2–3 variants; A/B test the top two on a small slice of your list.
5. Schedule and send: Use your email tool’s scheduling and track opens, clicks, and replies.

Copy-paste AI prompt (use as-is):

“You are an expert newsletter editor. Transform this rough outline into a 350-word newsletter for a non-technical audience aged 40+. Deliver: 1) three subject-line options (short, benefit-driven), 2) 1-line preview text, 3) ~350-word newsletter with a 2-sentence intro, three short sections (each 2–3 sentences), and a clear one-line CTA at the end. Keep tone warm, confident, and actionable. Here is the outline: [PASTE YOUR OUTLINE].”

Metrics to track:

• Open rate (benchmark: 20–35% depending on list quality)
• Click-through rate (CTR) — goal: 2–8%
• Conversion (CTA clicks to signups/sales)
• Time-to-publish (minutes saved using AI)
• Reply rate / qualitative feedback

Common mistakes & fixes:

• Too generic output — fix: add audience detail and desired examples in the prompt.
• No clear CTA — fix: specify desired action and format (button text, link, or reply).
• Overlong paragraphs — fix: request short sections and bullet highlights.
• Tone mismatch — fix: give 2–3 sample sentences that show your voice.

1-week action plan:

1. Day 1: Clean outline and audience note; run the AI prompt above.
2. Day 2: Edit the draft for voice; pick two subject lines.
3. Day 3: Send A/B test to 10–20% of list.
4. Day 4: Review metrics; pick best subject line and finalize content.
5. Day 5: Send to full list; monitor opens and clicks.
6. Day 6–7: Follow up to non-openers with a different subject line or resend to engaged segment with stronger CTA.

Your move.

Try this now (5 minutes): Paste your current hero copy into an AI and ask it to rewrite with one dominant benefit, one primary CTA, and a single trust line using only your real proof. Publish as Variant B. Measure hero CTA clicks for a week. That single swap often moves the needle fastest.

The problem: Pretty words don’t convert if they promise too much, split attention across multiple CTAs, or bury proof. AI can draft structure; you supply truth and priorities.

Why it matters: Conversions come from clarity, credibility, and one action. Fix the above-the-fold and you impact CTR and sign-ups immediately.

Field lesson: Most lift comes from three things — a clear headline, one CTA, one proof element above the fold. Start there before tinkering with secondary sections.

What you’ll need

• One-paragraph product summary (who, what, outcome).
• Top 3 benefits written as outcomes (not features).
• One verifiable proof item (testimonial, metric, guarantee).
• Primary CTA (single action word + outcome).
• Tone and length constraints (e.g., friendly, hero ≤120 words).

How to do it (step-by-step)

1. Draft the brief: 4–6 sentences + 3 outcome benefits + one proof + your primary CTA.
2. Run the prompt below to produce three hero options, claim flags, and a mobile-short hero.
3. Redline for truth: Remove anything the AI flagged as [CLAIM] if you can’t verify it today. Replace with plain benefits.
4. Pick one CTA: Keep only one primary CTA above the fold. If needed, move secondary actions to footer.
5. Publish A/B: Variant A (current) vs Variant B (AI hero). Split traffic 50/50. Do not change anything else.
6. Watch the right numbers: CTR on hero CTA and sign-up rate. Iterate from data, not opinions.

Copy‑paste AI prompt

“Act as a senior conversion copywriter. Using the brief below, deliver: (A) Hero v1 clear, v2 emotional, v3 credibility-led. Each hero must include: H1 (6–10 words), subhead (18–28 words), 3 benefit bullets (outcome-first), 1 social-proof line using only provided proof, 1 primary CTA (verb + outcome ≤4 words). Also provide: (B) Mobile hero (≤60 characters, H1 + 1 benefit + 1 CTA), (C) Objection block: 5 FAQs with concise answers using only the brief, (D) Compliance flags: mark any hard claims with [CLAIM], (E) Voice note: write at Grade 7 reading level. Do not invent metrics, brands, or guarantees. If proof is insufficient, insert [ADD PROOF HERE]. Brief: [PASTE ONE-PARAGRAPH SUMMARY]. Benefits: [LIST 3 OUTCOME BENEFITS]. Proof: [PASTE VERIFIABLE TESTIMONIAL/METRIC]. Primary CTA goal: [E.G., Start free trial / Book demo]. Tone: [E.G., Friendly, professional]. Return sections clearly labeled.”

Insider tricks

• Proof-first headline: Feed the proof first, then ask the AI to generate three headlines derived from the proof. Credibility beats cleverness.
• Readability pass: Ask the AI to rewrite to Grade 7 and remove jargon. Simpler copy gets more clicks.
• Objection mining: Paste a few support emails or call notes (anonymized). Prompt: “Extract the top 5 buying objections and write one-sentence answers using only this content.” Add that block under the hero.

What to expect

• Speed: You’ll have three clean hero options in minutes.
• Quality: Structure will be strong; you must police claims and brand tone.
• Impact: Testing headline and CTA typically yields faster wins than full rewrites.

Metrics to track

• Hero CTR: Hero CTA clicks / hero views (target a lift vs. your current baseline).
• Landing conversion rate: Sign-ups / visits (track by variant).
• Scroll to proof: Percent of visitors who reach the first proof element (simple proxy for engagement).
• Time to first click: Shorter is usually better for clarity.

Common mistakes & fixes

• Invented stats — Fix: delete or replace with “Customers say…” and only use verified quotes/ratings.
• Multiple CTAs — Fix: one primary CTA above the fold; demote everything else.
• Feature-speak — Fix: rewrite bullets as outcomes (“Save an hour weekly,” not “Automation tools”).
• Wall of text — Fix: shorten subhead, keep bullets to 5–7 words each.
• Voice mismatch — Fix: paste two examples of your brand copy into the prompt and say “match this tone.”

One-week action plan

1. Day 1: Build the brief. Run the prompt. Select two heroes (clear vs credibility-led). Produce the mobile-short hero.
2. Day 2: Redline [CLAIM] items. Insert verified proof or [ADD PROOF HERE] placeholders to fill later. Finalize one primary CTA.
3. Day 3: Launch A/B (50/50 traffic). Instrument hero CTR and sign-up by variant.
4. Days 4–6: Monitor daily. Do not change traffic sources. Aim for at least 500 visits/variant or 20 conversions/variant before deciding.
5. Day 7: If Variant B beats A by ≥10% relative on hero CTR and holds on sign-ups, ship B. Next test: CTA text or objection block placement.

Bonus prompt (microcopy tune-up)

“Rewrite these 3 CTAs as single-action, ≤3 words each, matching this tone: [TONE]. Provide 5 options that emphasize outcome, not effort: [PASTE CURRENT CTAS].”

Data wins. Keep the hero simple, proof visible, and the CTA singular. Then iterate.

Your move.

Hook: Want proposals that close faster and at higher prices? AI will help — but only if you use it to sharpen outcomes, not replace strategy.

A quick note on context: There wasn’t a prior point to respond to, so I’ll assume your priority is measurable results and clear KPIs. Here’s a focused, step-by-step approach to use AI to write client proposals that win more deals.

The problem: Most proposals are generic, long, and fail to tie recommendations to tangible business outcomes.

Why it matters: A clear, outcome-driven proposal reduces sales cycles, increases win rates and supports higher prices because clients understand ROI.

My experience / lesson: When teams use AI to create tailored executive summaries, data-backed outcomes and modular pricing options, win rates improve and proposal prep time drops 50% or more — but only when prompts and templates are disciplined.

1. Prepare what you’ll need
   • Client brief (pain, goals, timeline, budget range)
   • One reusable proposal template (sections only)
   • Performance examples / case studies with metrics
   • Access to an AI model (chat interface or API)
2. Create a winning structure
   1. Executive summary (impact & ROI)
   2. Challenge & proposed solution
   3. Deliverables, timeline, milestones
   4. Pricing options tied to outcomes
   5. Risk mitigation and next steps
3. Use AI to draft and tighten
   1. Feed client brief + template to AI
   2. Ask for one-page executive summary first
   3. Generate 2 pricing options (standard/premium) with clear deliverables
4. Human polish — edit for voice, confirm numbers, insert case-study proof.

Copy-paste AI prompt (use as-is):

“Generate a one-page executive summary for a proposal to [Client Name] in the [industry]. State the client problem, the proposed solution, expected outcomes with measurable KPIs (percent increases or time savings), a 6-month timeline with milestones, and 2 pricing options (Standard — outcomes X, Y; Premium — outcomes X, Y, Z). Tone: confident, clear, non-technical. Length: ~200 words.”

Metrics to track

• Proposal win rate (%)
• Average time to create a proposal (hours)
• Proposal-to-contract conversion speed (days)
• Average deal value
• Number of proposal revisions per deal

Common mistakes & fixes

1. Generic language — Fix: customize executive summary to client KPIs.
2. Too many options — Fix: offer 2 clear packages with ROI differences.
3. Late delivery — Fix: set a 48-hour AI-first draft SLA.
4. No proof — Fix: attach 1 relevant case study focused on metrics.

1-week action plan

1. Day 1: Gather 5 client briefs and one template.
2. Day 2: Build 3 prompt templates (exec summary, pricing, timeline).
3. Day 3: Generate drafts for two live prospects; refine.
4. Day 4: Insert metrics, case studies, finalize pricing.
5. Day 5: Internal review and A/B test two executive summaries.
6. Day 6: Send proposals to prospects.
7. Day 7: Measure responses; update prompts based on feedback.

Your move.