Jeff Bullas

Quick win (5 minutes): Ask AI for a one-sentence outcome line plus a single “assumptions” sentence so you don’t overpromise. Paste both at the top of your proposal. It gives clarity and credibility.

Gentle refinement: Your workflow is solid. One tweak: instead of only exporting as PDF, paste a 3–4 line summary in the email body too (outcome, timeline, price range, next step). Busy buyers often decide from the email before opening attachments. Also, whenever you mention ROI, show the assumptions or a range. That small line reduces pushback and builds trust.

What you’ll need

• A short client brief (name, one KPI to move, timeline, rough budget band)
• One past winning proposal to copy structure
• Two proof points (testimonial snippets or simple metrics)
• An AI writing assistant and your editor
• Your voice sample (paste a recent email you wrote) so AI mirrors your tone
• Price tiers outline (Core / Recommended / Premium)

Step-by-step (fast, repeatable)

1. Capture the brief (10 minutes): Client name, their one KPI, key date, budget band, existing assets (website, list, ads), and any constraints.
2. Generate the decision snapshot (5 minutes): One line outcome, one line assumptions, a simple 30/60/90 milestone, and a price range.
3. Draft the proposal sections (10–15 minutes): Executive summary, Problem, Proposed solution, 30/60/90 milestones, Investment (3 tiers) with expected results, Social proof, Assumptions & boundaries, Next step.
4. Personalize (5–10 minutes): Swap in two client-specific facts (their KPI baseline or revenue) and adjust the voice to sound like you.
5. QA & risk check (5 minutes): Read aloud, verify math, and make sure “Assumptions & boundaries” is visible. Put the decision snapshot at the top.
6. Deliver & follow up: Email the 3–4 line summary in the body, attach the PDF, propose one meeting time, and ask a single question: “Which option best fits your timeline?”

Copy-paste prompts (use as-is)

1) Decision snapshot generator

“You are a senior proposal writer. Based on the brief below, write: (a) one sentence outcome line with a date, (b) one sentence of key assumptions, (c) a 30/60/90-day milestone outline, and (d) a simple investment range for three tiers (Core/Recommended/Premium). Be concise, non-technical, and outcome-first. Use ranges if data is missing and avoid absolute guarantees. Brief: [paste your brief].”

2) Pricing options with boundaries

“Using the same brief, create three pricing options (Core/Recommended/Premium). For each, list: scope in 3–5 bullets, one primary expected result with a date, one risk or dependency, and a ‘not included’ bullet to prevent scope creep. Keep it plain English and client-friendly.”

3) Voice match (make it sound like me)

“Rewrite the draft below to match the tone of my writing sample: simple, confident, and warm. Keep sentences short, avoid jargon, and keep all numbers factual. Writing sample: [paste an email you sent]. Draft to rewrite: [paste AI draft].”

4) Safe ROI line (with ranges)

“Create a single ROI sentence using conservative/base/optimistic ranges based on the inputs below. Include one assumptions clause at the end. Inputs: current monthly leads [X], close rate [Y%], average order value [Z], target uplift [%]. If a number is missing, use a clear placeholder and say ‘to be validated.’”

Worked example (how it reads)

Scenario: A local home-services company wants 25% more booked jobs in 90 days.

• Outcome: “Increase booked jobs by ~20–25% within 90 days by improving Google visibility and tightening inquiry-to-booking follow-up.”
• Assumptions: “Assumes access to Google Business Profile, call tracking, and same-day lead follow-up.”
• Milestones: 30 days: fix listings + launch local ads; 60 days: review call scripts + retargeting; 90 days: optimize for highest-converting neighborhoods.
• Investment (range): Core $X–Y; Recommended $Y–Z; Premium $Z+ (finalized after audit).

Options:

• Core: Local SEO tune-up, GBP optimization, basic ads setup. Result: visibility up; first uplift by Day 30. Not included: ongoing ad management.
• Recommended: Core + ongoing ads management + call follow-up coaching. Result: predictable lead flow by Day 60.
• Premium: Recommended + conversion tracking + monthly reporting + offer testing. Result: 20–25% booking uplift by Day 90.

Mistakes to avoid (and the fix)

• Vague ROI promises: Show ranges and assumptions. Fix with the “Safe ROI line” prompt.
• Generic intros: Start with the decision snapshot, not a history lesson.
• Scope creep: Always add a “Not included” bullet per tier.
• Too many words: Keep paragraphs under 4 lines and sentences under 18 words.
• Weak CTA: Offer one time and ask, “Which option fits your timeline?”

What to expect

• First drafts in under an hour that sound like you, not a robot.
• Clearer sales talks because the proposal leads with outcomes and boundaries.
• Less back-and-forth on price thanks to tiered options with defined results.

7-day action plan

1. Day 1: Build a one-page template with the “Assumptions & boundaries” section and the decision snapshot at the top.
2. Day 2: Collect two proof points and one past winning proposal.
3. Day 3: Run the Decision Snapshot prompt on one live opportunity.
4. Day 4: Generate three pricing options with “not included” bullets.
5. Day 5: Voice-match the draft using your email sample; cut fluff.
6. Day 6: Send the proposal. Email body = 3–4 line summary + single CTA.
7. Day 7: Review replies, note objections, refine assumptions and pricing ranges.

Today’s nudge: Use the Decision Snapshot prompt on one current lead. Paste the outcome + assumptions at the top of your proposal and in your email. You’ll look decisive and trustworthy — and that wins deals.

Turn your task list into a daily, 30-second status ritual — with zero waffle and clear next moves.

You’re close. Let’s lock in a repeatable flow that groups by person, highlights outcomes and blockers, and keeps the voice identical every day.

What you’ll need (keep this simple)

• A filtered task list with columns: title, owner, status, percent complete, due date (optional), notes.
• An AI chat box or automation where you can paste rows.
• 2–3 minutes to skim for accuracy until it’s predictable.

How to do it (5 steps)

1. Filter today’s list: keep In Progress, Done, Blocked. Cap at 10 rows per owner to avoid noise.
2. Normalize columns: ensure every row has owner, status, and percent (use 0% or blank if unknown). Add a short “next step” phrase to notes if missing.
3. Paste into the AI with the prompt below: it will group by owner, compress tiny tasks, and tag blockers so they pop in Slack/email.
4. Skim-check 3 items: owner correct, percent aligned, no invented results. If off, fix those rows and re-run just that subset.
5. Post: copy the grouped output to your standup channel. Expect 1–3 lines per person, consistent tone.

Insider trick: teach the AI your compression rules

• Merge small, related tasks under one line like “Docs + minor fixes” to keep each person under 3 lines.
• Tag blockers as [BLOCKER] so they’re easy to search or trigger alerts.
• Default next steps by status when notes are thin: In Progress → “continue and deliver next milestone”; Done → “monitor/hand off”; Blocked → “waiting on X; will proceed when unblocked.”

Copy-paste AI prompt (use as-is)

“You write concise, reliable daily standups from structured task rows. Group updates by owner. For each owner, output up to 3 short lines that cover: top achievement/result (use % or metric only if present), what’s next, and any blocker. If more than 3 tasks, merge minor related items under one line. Use active voice, no fluff. Tag blockers with [BLOCKER]. Do not invent progress or dates. If a field is unknown, omit it.

Style guide: max 30 words per line; business tone; outcome → next step → blocker. Never exceed 3 lines per owner.

Input columns: title | owner | status | percent | due_date | notes

Now produce the grouped standup for these rows:
{paste your normalized rows here}
“

Mini example (what good looks like)

• Input rows:
• Implement signup A/B | Sarah | in progress | 60% | Fri | split test running on 60% traffic
• Fix payment bug | Tom | blocked | 0% | | awaiting API key from vendor
• Blog draft | Mia | done | 100% | | published; early visits ~200

• Output:
• Sarah — A/B test live at 60%; monitoring conversion; next: review lift Friday.
• Tom — [BLOCKER] Waiting for vendor API key; next: integrate and test once received.
• Mia — Blog published; ~200 initial visits; next: promote via newsletter.

Expectations and quick wins

• Consistency: same voice daily, 1–3 lines per person.
• Clarity: blockers tagged, next steps explicit, fewer follow-up questions.
• Speed: after day one, generation plus skim-check in ~2 minutes.

Premium templates you can reuse

• Developer-tight version: “Owner — Result (%/metric if present); next step; [BLOCKER if any]. ≤25 words.”
• Manager scan version: “Owner — Top win; top risk/blocker; next milestone date if present. ≤3 lines.”
• Risk-first version: “Owner — [BLOCKER]/risk first; impact; next move; escalation path only if named in notes.”

Common mistakes and fast fixes

• AI adds progress you didn’t do: Include a percent column and the “do not invent progress” rule. Spot-check 3 items.
• Too many lines per person: Add “merge minor related tasks” and “max 3 lines per owner.”
• Vague next steps: Add a “next step (one short phrase)” column to the task list. The AI simply mirrors it.
• Owner mix-ups: Ensure a single owner per row; filter out subtasks without owners.

1-week action plan

1. Day 1: Create the five columns; export 5–10 rows; run the prompt; adjust tone to your team.
2. Day 2: Expand to full team; enforce “≤3 lines per owner”; start tagging blockers.
3. Day 3: Make a one-click export view (pre-filtered). Save your prompt as a reusable template.
4. Day 4: Add the “next step” column to tasks; require it on new work items.
5. Day 5: Measure time saved; ask for feedback on clarity and tone; tweak limits if needed.
6. Day 6–7: Spot-check accuracy (>95%); lock the routine; consider a simple automation to paste and post.

Bonus: two tiny upgrades

• Priority tags: add P1/P2 in notes; ask the AI to prefix “P1” when present so leaders spot urgency.
• Decision log: if “decision:” appears in notes, add a short “Decision — …” line under that owner for easy search later.

Last nudge: Run one trial today with five rows. If the output is tight and accurate, make it tomorrow’s routine. Small habit, big clarity.

Nice call-out: redacting account details and using an exported CSV/PDF is the single safety step most people skip. Good people don’t hand over passwords — they share a safe file instead.

Here’s a short, practical playbook you can do this week. Quick wins first, then a simple system so you keep what you find.

What you’ll need

• One recent month of transactions (CSV or PDF) with personal IDs redacted
• A phone or computer and a spreadsheet app (or pen and paper)
• A separate savings account for automated transfers

Step-by-step (do this in under 90 minutes)

1. Open the CSV or list out transactions. If paper, take photos and type the merchant + amount into a quick list.
2. Sort by merchant and amount. Flag recurring names and any charges over $50.
3. Label each item: recurring subscription, grocery, dining, transport, utility, one-time large, other.
4. Pick 3 concrete targets: one subscription to cancel/downgrade, one habit to cut (e.g., takeout), one bill to negotiate or switch.
5. Estimate monthly savings and add them up. For one-time changes, divide annual savings by 12 to compare monthly impact.
6. Automate: set an automatic transfer of the expected savings to your separate account on payday.
7. Track for 30 days. Verify cancellations and repeat charges are gone; adjust the next month if needed.

Worked example — real, simple

• Streaming: $12/mo — cancel = $12
• Takeout: $300/mo — target 50% = $150
• Phone plan: $60 → switch to $40 = $20

Immediate monthly savings = $182 → $2,184/year. Automate $150/month and use $32 to cover short-term buffer.

How AI helps

• Quickly categorizes CSV rows, finds duplicates, spots subscriptions you forgot, and estimates savings.
• It doesn’t access accounts — you feed it the CSV. Never paste whole statements with IDs into public chat.

Mistakes & fixes

• Mistake: Cancelling without checking contract terms — Fix: ask for retention offers; record cancellation confirmation numbers.
• Mistake: Moving savings back to checking — Fix: automate to a separate savings account and label it clearly.

1-week action plan

1. Day 1: Export/redact transactions.
2. Day 2: Scan and flag 6–8 items.
3. Day 3: Choose 3 targets and calculate impact.
4. Day 4: Cancel/downgrade and call providers (use retention script).
5. Day 5: Set automated transfer of expected savings.
6. Day 6–7: Verify changes and record first saved amount.

Copy-paste AI prompt (redact personal IDs first)

“I have a one-month CSV of transactions with columns: date, merchant, amount. Please categorize each line into: recurring subscription, utility/insurance, grocery, dining/takeout, transport, one-time large, other. Then identify the top 3 monthly changes that will save the most, estimate monthly and annual savings, and provide exact next steps to cancel/downgrade or cut the habit (include scripts for calling providers).”

Small, consistent moves beat one big overhaul. Do the scan, pick three targets, automate the savings — and watch momentum build.

Great point about focusing on learning goals first — framing an AI simulation around clear objectives keeps the activity useful, safe and memorable. I’ll add a practical, step-by-step approach you can use right away.

Why this works: AI can play a historical figure to spark curiosity, practice conversation and test understanding. When done with clear learning goals and simple guardrails, it’s a low-cost, high-engagement activity for classrooms and hobby groups.

What you’ll need

• A device with a browser and internet access.
• An AI chatbot or model (a web-based chat tool is easiest).
• Short, reliable source notes about the historical figure (biography highlights, quotes, timeline).
• Basic rules: accuracy focus, respectful tone, and a debrief plan for students.

Step-by-step setup

1. Define the learning objective (e.g., understand decisions of the figure, practice historical empathy).
2. Prepare a 1-page fact sheet: 6–10 verified points (dates, key events, known quotes).
3. Open your AI chat tool and set a clear system prompt (see example below).
4. Run a short test: ask the AI three factual questions and one open-ended question to check tone and accuracy.
5. Run the activity with learners: 10–15 minute dialogue rounds, then a 10-minute debrief comparing AI answers to your fact sheet.

Example system prompt (copy-paste)

You are role-playing as [Name of historical figure]. Speak in first person, remain consistent with verified historical facts only, and keep answers concise (1–3 short paragraphs). When unsure, say “I am not certain” and ask the facilitator for clarification. Begin by introducing yourself in one sentence and ask the learner one question about why they want to speak with you.

Common mistakes & fixes

• AI adds myths or modern language: Fix by instructing “avoid anachronisms; cite or reference the source of facts.”
• AI becomes preachy or off-topic: Fix by limiting response length and adding “ask a question back to the learner.”
• Learners take AI statements as gospel: Fix by always debriefing and comparing to the fact sheet.

7-day quick action plan

1. Day 1: Pick figure and objective. Create fact sheet.
2. Day 2: Draft system prompt and test with 3 questions.
3. Day 3: Revise prompt for tone/accuracy. Prepare activity script.
4. Day 4: Run pilot with a colleague or friend.
5. Day 5: Adjust based on pilot feedback.
6. Day 6: Run with learners; record common issues.
7. Day 7: Debrief, iterate, and scale up.

Closing reminder: Start small, focus on clear learning goals, and use a short fact sheet plus a debrief to keep things accurate and educational. This gives you fast wins and a path to improve.

Here’s a faster, safer ad-variation system you can copy today. It locks your variables, keeps language within Meta policy, and produces paste-ready blocks you can test this week.

Why this version works: it combines Angle × Tone × Format with strict character ranges, policy-safe phrasing, and visual direction. That means clearer signals in 48–72 hours and less wasted spend.

• What you’ll need
  • A short sheet with: Product/Service, Audience, Primary Benefit, Proof Point (1 fact), Preferred CTA, Brand Tone (1–2 words), Price/Offer (if any), Compliance Notes (words to avoid).
  • Meta ad specs handy: Headline 25–40 chars, Primary Text 90–125 chars, Description 30–40 chars.
  • An AI tool that returns plain text.

1. How to run it (step-by-step)
   1. Pick one row from your sheet.
   2. Paste the “12-Pack Prompt” below. It generates 3 tones × 4 angles = 12 variations with guardrails.
   3. Skim for brand voice and claims. Trim any lines by a few characters if your preview shows truncation.
   4. Name ads clearly (Angle_Tone_Format) and upload as equal-budget A/B groups.
   5. Run 3–7 days. Pause losers after 48–72 hours; shift budget to winners. Keep imagery constant for the first test wave.

Copy-paste AI prompt (12-Pack, policy-safe, paste-ready)

Act as a senior performance copywriter for Facebook/Instagram. Generate 12 ad variations as numbered blocks 1–12. Each block must include EXACTLY these fields, one per line:
Angle: [Outcome | Problem Relief | Social Proof | Offer/Value]
Format: [Benefit-led or Problem-led]
Tone: [Confident | Friendly | Urgent]
Headline (<=40 chars): …
Primary Text (90–125 chars): …
Description (30–40 chars): …
CTA: [Shop Now | Learn More | Book Now | Sign Up]
Visual Direction (1 sentence): …
Overlay Text (3–5 words): …
Aspect Ratio: [1:1 or 4:5]
Policy Check: [Safe | Needs Rewrite] + reason if not safe

Rules:
– Keep language simple and measurable. Include the benefit and one proof point.
– Avoid personal attributes (e.g., “you’ve got back pain”), sensitive traits, before/after claims, or guarantees. Use community phrasing (“many people”, “busy professionals”).
– No health or financial promises. No second-person diagnoses.
– Do not exceed character limits.

Use these inputs:
Product: [INSERT]
Audience: [INSERT]
Primary benefit: [INSERT]
Proof point (1 fact/stat): [INSERT]
Price/Offer (if any): [INSERT or N/A]
Preferred CTA: [INSERT one from list]
Brand tone hint: [INSERT e.g., calm, upbeat]
Compliance notes (words to avoid): [INSERT or N/A]

Output style:
– Numbered 1–12.
– One field per line in the order above.
– No extra commentary except the fields.
At the end, add a short section: “Top 3 to test first” with IDs and one-line rationale.

Variant prompt (CSV-style for sheets)

Return the same 12 variations as 12 CSV rows using this header once:
Angle,Format,Tone,Headline,Primary Text,Description,CTA,Visual Direction,Overlay Text,Aspect Ratio,Policy Check
Follow the same inputs, rules, and character limits. Use commas inside quotes if needed. No extra text before or after the table.

Policy-safe rescue prompt (use if any line is flagged)

Rewrite the following ad copy to be Meta policy-safe while preserving meaning and character limits. Avoid personal attributes, medical/financial promises, and before/after claims. Return Headline (<=40), Primary Text (90–125), Description (30–40), CTA, and a 1-sentence Visual Direction. Text to fix: [PASTE COPY]

Quick example (2 of 12 shown)

• Inputs: Product: Weekly Meal Planner App. Audience: Busy families. Benefit: Saves 3–5 hours each week. Proof: 4.7★ average from 2,100 reviews. Offer: 14-day free trial. CTA: Sign Up. Tone: friendly.

• 1) Angle: Outcome | Format: Benefit-led | Tone: FriendlyHeadline: Dinners, planned in minutesPrimary Text: Cut meal chaos. Get a weekly plan in 5 minutes and save 3–5 hours each week. Rated 4.7★ by 2,100 families.Description: Make weeknights easierCTA: Sign UpVisual Direction: Overhead of simple family dinner with app on phone beside ingredients.Overlay Text: Plan in 5 minutesAspect Ratio: 4:5Policy Check: Safe
• 2) Angle: Social Proof | Format: Benefit-led | Tone: ConfidentHeadline: 2,100 reviews. 4.7★Primary Text: Join thousands using the planner that saves hours and reduces takeout. Start a 14‑day free trial today.Description: Trusted by busy familiesCTA: Sign UpVisual Direction: Carousel showing app screens + star rating badge.Overlay Text: 4.7★ RatedAspect Ratio: 1:1Policy Check: Safe

Insider tips that lift CTR fast

• Hook library: try numbers (Save 3–5 hours), time (In 5 minutes), risk-removal (Cancel anytime), and social proof (4.7★).
• Repeat the CTA once in Primary Text. Meta prefers clarity; users click more when told what’s next.
• Overlay Text should echo the headline, not introduce a new idea. Consistency boosts scan-ability.
• Keep imagery constant on your first test wave. Change only copy to isolate winners.

Common mistakes & fast fixes

1. Soft, vague benefits → Add a number or time frame.
2. Policy red flags → Swap “you” + trait claims for community phrasing and outcomes.
3. Too many variables → Lock image and CTA; rotate headlines/primary only.
4. Messy exports → Use the CSV variant prompt; import straight to your sheet or Ads Manager template.

5-day action plan

1. Day 1: Fill the sheet for 3–5 offers. Add one proof stat each.
2. Day 2: Run the 12-Pack Prompt per product. Pick the best 3–4 per offer.
3. Day 3: Upload to Ads Manager. Name by Angle_Tone_Format. Even budgets.
4. Day 4: Pause any ad with CTR below your baseline (or CPA 20–30% above target).
5. Day 5: Shift budget to winners. Spin 4–6 new headlines using the Policy-safe rescue prompt for the winning angle.

What to expect: 12 clean, test-ready variations per product. Early CTR signal by 48–72 hours. CVR/CPA stabilizing by day 5–7. Your job becomes picking angles, not wordsmithing from scratch.

Start with one product today. Use the 12-Pack, keep imagery fixed, and let the numbers tell you what to scale.

Hook: A one‑line habit protects trust. Do it before you publish.

Why this matters

Readers care who checked the facts and accepted responsibility. A short, clear AI disclosure + a quick provenance note keeps clients calm, compliance happy, and you out of post‑publication headaches.

What you’ll need

• One document (Word/Google Doc/PDF).
• A simple editor and a project folder.
• A provenance file (text, spreadsheet) and a disclosure template snippet.

Step‑by‑step (do this now)

1. Choose level: minimal (internal), contextual (client), or formal (regulated).
2. Add disclosure: put a one‑line or short paragraph in header, footer or cover page. Example lines you can copy:
   • Minimal: “Drafted with assistance from an AI tool; reviewed by [Author Name].”
   • Contextual: “This draft used AI for structure and wording. All data and recommendations were verified and approved by [Author Name].”
   • Formal: “Portions of this document were produced using an AI tool. [Author Name] reviewed, validated, and accepts responsibility for the final content. See project provenance log.”
3. Human review (mandatory): fact‑check figures, confirm names/dates, remove placeholders, check tone. Initial or log the review.
4. Log provenance: add one line to your project log: date, AI tool, scope (drafting/editing), reviewer initials.
5. Template it: save disclosure as a template snippet with a reviewer field to force the pause.

Worked example

Two‑page client briefing: add the contextual line on the cover. In the project folder add: “2025‑11‑22 — AI tool: [name] — Scope: drafting — Reviewer: AB.” During review check numbers, confirm client names and remove any tool placeholders. Save as a client briefing template.

Common mistakes & quick fixes

• Mistake: No disclosure. Fix: Add the one‑line and retroactively flag recent client docs.
• Mistake: Too much tech detail. Fix: Keep wording plain; keep technical notes internal.
• Mistake: Skipping review. Fix: Make reviewer initials required in the template.

Quick 5‑day action plan

1. Day 1: Add one‑line disclosure to next outgoing doc and save as template.
2. Day 2: Create a client paragraph disclosure and add to client templates.
3. Day 3: Start a provenance log in the project folder.
4. Day 4: Audit last 10 docs and add disclosures where appropriate.
5. Day 5: Track one metric (follow‑ups or revision count) to see impact.

AI prompt you can copy‑paste

“Rewrite this disclosure for a professional audience: ‘This document used AI assistance.’ Produce 3 polished options: a one‑sentence internal memo line, a concise paragraph for client reports, and a formal footnote for regulated documents. For each option, suggest placement (header, cover note, endnote) and add a one‑line rationale. Also output a one‑line provenance template I can copy into a project log.”

Closing reminder: Small habit, big payoff. Add the line, do a quick human check, and log it. That three‑step routine protects credibility and saves time later.

Spot on: your 20–30 word, consistent captions are the highest-leverage move. Let’s add two simple upgrades that make your LoRA snap into brand: a Style Passport and a Calibration Grid.

5-minute upgrade (do this now)

• Create a one-screen Style Passport you’ll paste into every workflow:
• Palette: 3–5 hex codes (e.g., #FFDAB9, #7BA7A1, #F4F1EC)
• Mood words: calm, confident, understated
• Lighting: soft directional, flat shadows
• Composition: centered close-up, generous negative space
• Texture/finish: matte, minimal grain
• Allowed motifs: hands, simple props
• Banned elements: busy patterns, text overlays, watermarks, neon colors
• Trigger token: a unique word for training, e.g., brndx (not a real word)

Why this helps: captions teach “what,” the passport teaches “how.” Together, they remove guesswork so the LoRA learns one voice.

What you’ll need

• Your curated 80–120 images and draft captions.
• The finished Style Passport (above).
• A unique trigger token (e.g., brndx) you’ll put in every caption.
• Access to LoRA training (local or managed) and 2–3 short runs to iterate.

Step-by-step — lock in consistency

1. Finalize the Style Passport: keep it short and use it everywhere (captioning, prompts, QA).
2. Standardize captions: append the trigger token to every caption (start or end). Example: “Centered close-up of serum bottle on warm neutral background, soft directional light, calm mood — brndx.”
3. Use controlled tags: choose 6–8 tags only from your passport words (e.g., minimal, matte, soft-lighting, warm-neutrals, centered, close-up).
4. Prep your negatives: one house negative list you’ll use in every generation: “no text, no watermark, no neon, no clutter, no harsh shadows.”
5. Train small, check fast: short first run, then extend only if quality improves. If outputs look too generic, slightly increase training steps; if they clone images, reduce.
6. Calibrate strength: test LoRA strength at 0.6, 0.8, 1.0, 1.2 with the same prompts. Pick the strength that best fits the passport.
7. Score and iterate: rate 30–50 samples on three things: palette, lighting, composition. Fix the worst offender (often captions) and rerun.

Copy-paste prompt — caption standardizer

“You will be given an existing caption and the Style Passport below. Rewrite the caption to 20–30 words that clearly state subject, dominant colors (use hex codes), composition, and mood. Add 6–8 keywords chosen only from the passport vocabulary. Append the trigger token exactly once at the end. Output format: Caption: [text] || Keywords: [kw1, kw2, …]. Style Passport: [paste your passport here]. Trigger token: brndx.”

Copy-paste prompt — calibration grid (use to test your LoRA)

“Generate a set of images in the style of brndx (apply LoRA at strength: [0.6|0.8|1.0|1.2]). Subject: person holding product; composition: centered close-up with negative space; lighting: soft directional, flat shadows; palette: #FFDAB9, #7BA7A1, #F4F1EC; mood: calm, confident; texture: matte. Negative: no text, no watermark, no neon, no clutter, no harsh shadows. Produce one image per listed strength with identical seed for fair comparison.”

Insider tricks that move the needle

• Unique trigger: use a made-up token (brndx) so the model doesn’t confuse your style with public words.
• Hex codes in prompts: including exact colors reduces palette drift more than generic “warm” language.
• Split by sub-style: if you have product and lifestyle looks, train two small LoRAs instead of one mixed model.
• Fixed prompt deck: keep 3–5 standard prompts you always test. That becomes your reliable scoreboard across runs.
• House negatives: paste the same negative list into every generation to prevent recurring flaws.

Common mistakes & fixes

• Trigger missing in captions → Add the token to every caption; retrain short.
• Color drift → Put hex codes in captions and prompts; remove images with off-brand lighting.
• Busy backgrounds → Add “clean background, generous negative space” to captions; strengthen the negative list.
• Overfitting → Reduce epochs/steps, add mild augmentation, and prune near-duplicate images.
• Underfitting (generic look) → Tighten captions, ensure trigger is present, and add 10–20 more on-brand examples.

What to expect

• After 2–3 quick runs, target a 70%+ acceptance rate for internal drafts.
• Calibrating LoRA strength is often the fastest path from “close” to “on-brand.”

5-day action plan

1. Build your Style Passport and negative list (15 minutes). Pick a unique trigger token.
2. Standardize 50–100 captions using the caption prompt. Append the trigger to each.
3. Run a short training pass. Save the checkpoint even if imperfect.
4. Use the calibration grid prompt at four strengths; score palette, lighting, composition 1–5.
5. Fix the top issue (usually captions or outliers) and retrain short. Deploy the best strength into a small campaign test.

Closing thought: your captions teach clarity; the Style Passport and Calibration Grid teach consistency. Stack those three and your LoRA becomes a dependable, on-brand creative assistant fast.

Nice quick win — that one-line disclosure is exactly the low-friction step that protects trust. Here’s a short, practical playbook to make that habit reliable and repeatable.

Why add more than a one-liner? A single sentence preserves trust quickly. Adding a tiny provenance step and a mandatory human review keeps you out of trouble and speeds stakeholder acceptance.

What you’ll need

• A current document (Word, Google Doc, PDF).
• A basic editor and access to your project folder.
• A one-line disclosure template and a provenance log (simple text file or spreadsheet).

Step-by-step (do this now)

1. Add the one-line disclosure to header/footer or cover page: “Drafted with assistance from an AI tool; final content reviewed and approved by [Author Name].”
2. Create a provenance entry: date, AI tool, scope (drafting/editing), reviewer initials. Example: “2025-11-22 — AI tool: [name] — Scope: drafting — Reviewer: AB”.
3. Run a quick human review: fact-check numbers, verify names, correct tone, remove sensitive details.
4. Save the disclosure as a template field in your doc system so it auto-populates next time.
5. Log the provenance entry in your project folder for traceability.

Simple disclosure options (copy-and-use)

• Minimal (internal): “Drafted with assistance from an AI tool; reviewed by [Author Name].”
• Contextual (client): “This draft was generated with the assistance of an AI writing tool for structure and language. All data and recommendations were verified and approved by [Author Name].”
• Formal (regulated): “Portions of this document were produced using an AI tool. [Author Name] reviewed, validated, and accepts responsibility for the final content. See project provenance log for details.”

Common mistakes & quick fixes

• Mistake: No disclosure. Fix: Add the one-line immediately and add provenance for recent deliverables.
• Mistake: Too much tech detail. Fix: Use plain language; keep technical notes in internal policy documents.
• Mistake: Skipping human review. Fix: Make a reviewer initial required in the template before finalizing.

Quick 5-day action plan

1. Day 1: Add one-line disclosure to next outgoing doc and save as template.
2. Day 2: Create a one-paragraph client disclosure and add to client templates.
3. Day 3: Start a provenance log in project folder.
4. Day 4: Audit last 10 documents; add disclosures where needed.
5. Day 5: Track one metric (follow-up queries or revision count) to measure impact.

Copy-paste AI prompt (use to generate tailored disclosure options)

“Rewrite this disclosure for a professional audience: ‘This document used AI assistance.’ Produce 3 options: a one-sentence internal memo line, a short paragraph for client reports, and a formal footnote for regulated documents. For each option, suggest placement (header, cover note, endnote) and add a one-line rationale.”

Small practice, big payoff: Add the line, log the tool, and confirm you’ve reviewed the facts. That three-step routine saves credibility and time.