Jeff Bullas

One email. Three audiences. Seconds, not hours. Use AI to reshape the same message for a junior, a peer, and an executive—without losing your ask or tone.

Do / Don’t checklist

• Do start with the ask and deadline for execs (BLUF: Bottom Line Up Front).
• Do give peers concise context and next steps they can execute.
• Do give juniors clear definitions, step-by-step, and where to get help.
• Do keep numbers, dates, owners, and risks identical across versions.
• Don’t bury the lead, over-explain to execs, or use jargon with juniors.
• Don’t let AI invent facts—lock the facts in the prompt.
• Don’t mix tones—confidence for execs, collaborative for peers, supportive for juniors.

What you’ll need (5 minutes)

• An AI chat tool (any reputable one works).
• Your raw email (or bullet notes): situation, ask, deadline, impact, risks.
• Your audience “style cards” (below). Save them once—reuse forever.

Audience style cards (save these)

• Executive: 50–120 words. Start with decision or ask. Include 1–3 bullets max: impact ($, risk, timeline). Offer 1–2 options with a recommendation. No jargon.
• Peer: 80–150 words. One-line purpose, brief context, what I need from you, by when. Use bullets for tasks.
• Junior: 120–200 words. Plain language (Grade 7–8). Define terms. Step-by-step tasks with owner and due date. Link to resources or where to ask.

Step-by-step

1. Lock your core message. Use SAIN: Situation, Ask, Impact, Next step. Write it in bullets so AI can’t drift.
2. Pick the audience level(s) you need: junior, peer, executive.
3. Use the prompt template below. Paste your facts into the brackets.
4. Check the outputs for: same ask/deadline, correct numbers, tone match. If off, say “shorter,” “plainer,” or “add options.”
5. Save your best version as a reusable snippet for next time.

High-value tip: Give executives “options + recommendation” (Option A/B + your pick) and a clear cost of delay. It reduces back-and-forth and speeds decisions.

Copy-paste AI prompt (reusable)

“You are my email rewriter. Rewrite the same message for three audiences: Junior, Peer, Executive. Keep facts identical. Use the style cards below.

Facts (do not change):

– Situation: [what’s happening].- Ask: [what decision or action you need].- Deadline: [date/time].- Impact if delayed: [money/time/risk].- Key numbers: [metrics].- Owner(s): [names/roles].

Style cards:

– Executive: 50–120 words, start with the ask, include 1–3 bullets for impact ($, risk, timing). Provide 1–2 options with my recommendation. No jargon.- Peer: 80–150 words, one-line purpose, brief context, specific tasks with dates.- Junior: 120–200 words, plain language Grade 7–8, define terms, step-by-step, where to ask questions.

Return exactly three labeled versions: Junior, Peer, Executive.”

Worked example

Your raw facts (paste these into the prompt):

• Situation: Our website relaunch is 2 weeks behind because the design vendor missed the handoff.
• Ask: Approve the revised timeline or switch to in-house design.
• Deadline: Friday 5 pm.
• Impact if delayed: +$12,000 vendor overage and 3-week revenue delay.
• Key numbers: Vendor needs 3 more days; in-house can start tomorrow but drops two low-priority pages.
• Owner(s): Marketing (Anna) and Product (Leo).

Expected AI output (clean and ready to send):

• ExecutiveDecision needed by Friday 5 pm: Approve the revised vendor timeline or switch to in-house design.- Impact if we wait: +$12,000 and ~3-week revenue slip.- Option A (Recommend): Switch to in-house tomorrow; launch core pages on time; defer two low-priority pages.- Option B: Stay with vendor; add 3 days; pay overage; full scope later.Reply with A or B.
• PeerPurpose: Keep launch viable while we address the design delay.Context: Vendor missed the handoff; they need 3 more days. If we slip again, we face a $12,000 overage and lose ~3 weeks of revenue.What I need:- Leo: Confirm in-house capacity by 2 pm today.- Anna: Prep a reduced-scope plan (defer two low-priority pages) by noon tomorrow.Decision by Friday 5 pm: Approve in-house switch or hold vendor plan.
• JuniorWe’re late on the website relaunch because our design vendor missed the delivery. That means our new pages may not be ready on time.What we’re deciding: Keep the vendor’s slower plan or move design work to our own team (in-house). If we delay again, it could cost $12,000 and push revenue by about 3 weeks.Steps for you:1) Gather the latest image files into the “Launch” folder today.2) Update the checklist and mark which two low-priority pages can wait.3) If we switch to in-house, join tomorrow’s kickoff at 10 am with Anna and Leo.If anything is unclear, reply here and tag Anna (Marketing) or Leo (Product). Decision by Friday 5 pm.

Mistakes and quick fixes

• Mistake: The AI adds new facts. Fix: Start your prompt with “Facts (do not change)” and list them as bullets.
• Mistake: Executive version is too long. Fix: Say “80 words max. Start with the ask. Include 2 bullets only.”
• Mistake: Junior version uses jargon. Fix: Add “Grade 7 reading level. Define terms in parentheses.”
• Mistake: The ask is buried. Fix: Add “First sentence must include the ask and deadline.”
• Mistake: No clear handoffs. Fix: Add “List owners by name/role and due date as bullets.”

Fast action plan (10 minutes)

1. Pick one email you’re about to send this week.
2. Write your SAIN bullets: Situation, Ask, Impact, Next step.
3. Paste the reusable prompt above into your AI tool; drop in your facts.
4. Generate all three levels; pick the one you need right now.
5. Sanity-check: Ask the AI, “In one line, what am I asking for and by when?” If wrong, regenerate.
6. Save the best phrasing as your snippet library for future emails.

Closing thought: AI won’t know your politics or priorities. You do. Feed it crisp facts and a clear ask, and it will give you audience-ready emails that move work forward today.

Quick hook: You don’t need to be a techie to use AI to plan an emergency fund and make your savings work smarter. Small steps, clear rules, immediate wins.

Why this matters: An emergency fund gives you freedom and calm. AI helps you test scenarios, find the best place for cash, and automate decisions without jargon.

What you’ll need:

• Monthly take-home pay and average monthly expenses (3–6 months is common).
• List of predictable irregular costs (car maintenance, taxes, medical).
• Access to a simple AI chat tool (like ChatGPT) or a spreadsheet.
• Bank account access to set up automatic transfers.

Step-by-step:

1. Gather numbers: income, fixed & variable expenses, debts, and current savings.
2. Decide target buffer: conservative = 6–12 months of expenses; lean = 3 months.
3. Ask AI to model timelines: how long to reach that fund given different monthly contributions.
4. Optimize placement: split funds between instant-access cash (1–2 months), high-yield savings (rest), and short-term liquid investments if you can tolerate slightly less liquidity.
5. Automate: set monthly transfers on payday. Use AI to draft rules (e.g., if balance > X move surplus to investment sweep).

Copy-paste AI prompt (use in a chat with an AI):

“You are a practical personal finance assistant. I earn $X/month (net). My average monthly expenses are $Y. I have $Z in savings and $D in debt. Recommend an emergency fund target, a monthly savings amount to reach it in 12 months, and a simple allocation between instant-access cash, high-yield savings, and short-term liquid investments. List 5 actionable steps I can automate today.”

Worked example:

If your expenses are $3,000/month and you have $4,000 saved: a 6-month fund = $18,000. You need $14,000 more. Saving $700/month reaches it in ~20 months; $1,200/month reaches it in ~12 months. Put $6,000 as instant-access cash, remainder in high-yield savings. Automate $1,200 transfer each payday.

Common mistakes & fixes:

• Do not overestimate investment returns for emergency money. Keep liquidity first.
• Do not ignore irregular costs—add a buffer for them.
• If stuck, reduce the target (short-term goal) and increase automation to build momentum.

7-day action plan:

1. Day 1: Gather income/expense numbers.
2. Day 2: Pick target months (3/6/12).
3. Day 3: Use the AI prompt above.
4. Day 4: Open a high-yield savings if needed.
5. Day 5: Set up automatic transfers.
6. Day 6: Check allocation and adjust.
7. Day 7: Review and schedule quarterly check-ins with the AI.

Small, steady actions beat perfect plans. Use AI to test scenarios, then act. Start with one automated transfer this week.

Quick win: Use AI to build short, practical voice and style checklists your team can actually follow — in under an hour.

Why this matters: consistent voice saves time, avoids edits, and ensures every piece of writing sounds like your brand. AI can draft checklists fast; you refine them with your team.

What you’ll need

• 3–5 real writing samples from your team (emails, posts, articles).
• A short list of your brand goals (e.g., friendly, expert, concise).
• An AI chat tool (ChatGPT or similar) or a teammate who can paste prompts.
• 15–45 minutes for the first draft, then short tests.

Step-by-step (do this once, then iterate)

1. Define the top 3 voice traits. Pick 3 words: e.g., friendly, confident, helpful.
2. Gather examples. Copy 3 short passages that match and 3 that don’t.
3. Ask AI to draft a one-page checklist. Use the prompt below (copy-paste).
4. Trim to 8–12 actionable items. Make each item a single sentence and include examples.
5. Test on 2 real drafts. Apply the checklist, time edits, gather feedback.
6. Refine weekly. Keep the checklist on a shared doc the team can access.

Copy-paste AI prompt

“You are an expert brand voice editor. Given these brand traits: friendly, confident, helpful. Using the three sample good sentences: ‘Thanks for reaching out — we’ll handle this.’ ‘Here’s the short plan to get started.’ ‘Let me know if you want examples.’ And three poor examples: ‘Please be informed that…’ ‘Attached is the document for your convenience.’ ‘We regret to inform you…’ Create a concise voice and style checklist of 10 items. Each item must be one line, include a short example of ‘good’ and ‘bad’ (one sentence each), and finish with a suggested replacement phrase if applicable. Keep language simple and practical.”

Example checklist items

• Use active voice. Good: “We’ll send the report.” Bad: “The report will be sent.” Replace with: “We’ll send the report.”
• Keep sentences under 20 words. Good: “Here’s the plan to start.” Bad: “This is a plan which outlines multiple steps to be completed.”
• Open friendly, close helpful. Good opener: “Thanks for reaching out.” Bad opener: “Please be informed.”

Mistakes & fixes

• Too vague checklist: make items actionable (replace vague word with exact alternative).
• Too long: aim for 8–12 items so people actually use it.
• No examples: always add a one-line good/bad example.

7-day action plan

1. Day 1: Define traits and collect samples.
2. Day 2: Generate checklist with AI.
3. Day 3: Trim to essentials.
4. Day 4: Test on two drafts.
5. Day 5: Refine with feedback.
6. Day 6: Publish shared checklist.
7. Day 7: Train the team (10–15 minute session).

Start small, ship fast, improve weekly. The first checklist won’t be perfect — it will get far more useful once your team uses it.

Good point — wanting practical, ready-to-use guidance is exactly the right approach. Below is a simple, step-by-step way you can use AI to draft a Professional Services Agreement (PSA) and proposal terms quickly, safely and with confidence.

Quick win (try in under 5 minutes): Paste the AI prompt below into your AI tool with just the client name and project title. You’ll get a first-draft PSA outline that you can refine.

What you need before you start

• Basic project facts: client name, your company name, scope summary, price or rate, timeline.
• Decisions on key risk items: payment terms (deposit, milestones), liability cap, IP ownership preference.
• A copy of your current or ideal template (if you have one) to speed edits.

Step-by-step — how to do it

1. Gather the facts above (10–15 minutes).
2. Open your AI tool and paste the prompt below. Start with a short request for a concise draft (5–10 clauses).
3. Review AI output for clarity: check scope, deliverables, fees, dates, and termination terms.
4. Edit language to match your tone (plain English). Remove vague phrases like “reasonable efforts.”
5. Send to a lawyer for a quick review (ask for a limited-scope review focusing on liability and IP).

Copy-paste AI prompt (use as-is)

“Draft a concise Professional Services Agreement outline (5–10 short clauses) for [My Company: ACME Consulting] providing [Project: Website UX redesign] for [Client: Example Co]. Include: scope of services, deliverables, timeline, fees and payment schedule, acceptance criteria, intellectual property ownership, confidentiality, limitation of liability with a reasonable cap, termination rights, and governing law. Keep language plain, and mark any items that should be reviewed by a lawyer. Output as short numbered clauses suitable for converting into a contract.”

Example snippet you might get

1. Scope: ACME Consulting will deliver a website UX redesign including discovery, wireframes, and two rounds of revisions. Deliverables: discovery report, wireframes (desktop & mobile), final design files.

Common mistakes and fixes

• Vague scope — Fix: add specific deliverables and acceptance criteria (what “done” looks like).
• No payment milestones — Fix: require deposit + milestone payments tied to deliverables.
• Over-reliance on AI wording — Fix: simplify legalese and flag key risk items for lawyer review.

Simple action plan (next 48 hours)

1. Gather project facts now (15 minutes).
2. Run the AI prompt and get a draft (5 minutes).
3. Refine and send to counsel for a quick check (ask for 24–48 hour turnaround).

Reminder: AI speeds drafting and produces a professional starting point — but it’s not a lawyer. Use it to do the heavy lifting, then add human review. Start small, iterate, and you’ll have a reliable template in no time.

Short answer: Yes. With the right inputs, AI can auto-draft a clean, consistent brand style guide from your existing materials in under two hours. Think of it as an 80–90% first edition you can refine.

How it works (in plain English)

• AI reads your logo, screenshots, PDFs, and posts.
• It extracts your colors, fonts, voice, imagery cues, and layout habits.
• It assembles these into a one-pager plus a fuller guide: rules, examples, and ready-to-use templates.

What you’ll need

• 5–10 example assets: logo (SVG/PNG), 3–5 recent posts or emails (text), 2–3 social posts, a landing page or brochure (PDF/image), 3–5 photos that feel “on-brand.”
• Any known brand elements (hex colors, font names) if you have them.
• An AI chat tool that accepts file uploads and images.

Step-by-step (the fast path)

1. Assemble your inputs
   • Create a folder named “Brand Inputs.”
   • Rename files clearly: logo-primary.png, post-1.txt, landingpage.jpg, photo-mood-1.jpg.
   • If fonts are unknown, include screenshots of your website headings and body text.
2. Pick your assistant
   • Use an AI chat that can read images and PDFs. You’ll upload your materials in one chat thread so it learns the pattern.
3. Run the extraction prompt
   • Upload the assets and paste the prompt below. Ask for a one-page summary and a full guide.
4. Color sanity check
   • Ask for 5–7 colors with exact hex codes, roles (primary, secondary, accent, background, text), and usage percentages.
   • Request contrast checks for body text on light/dark backgrounds to meet accessibility (AA) where practical.
5. Font pairing
   • Ask for a heading and body font with web-safe or widely available fallbacks, plus size/line-height guidance.
   • Confirm license availability for commercial use. If unsure, pick Google Fonts equivalents.
6. Voice & tone rules
   • Have AI extract your voice from your emails/posts: 3–5 voice pillars, “say this/not that,” and 2 sample paragraphs in your tone.
7. Logo usage
   • Generate rules: safe area, minimum size, background do/don’t, monochrome variant, and misuse examples.
8. Imagery & icon style
   • Define look: lighting, color temperature, subjects, crop, and negative list (what to avoid).
   • Ask for 5 example prompts you can reuse in image tools for consistency.
9. Templates
   • Request 2–3 layouts: social post, email signature, presentation cover. Include spacing, hierarchy, and call-to-action styling.
10. Export
    • Ask AI to output a one-page quick reference and a detailed guide you can paste into your doc tool of choice.

Copy-paste prompt (core extractor)

“I’ve uploaded my brand materials. Create two deliverables: (1) a one-page Brand Quick Reference and (2) a 6–10 section Brand Style Guide. Extract what exists and propose sensible defaults where gaps appear. Include: Brand essence (1–2 sentences), Voice pillars (3–5) with ‘Do/Don’t’ and sample paragraph; Color palette with hex codes, roles, and usage percentages; Accessibility contrast notes; Typography pairings with sizes, weights, line-heights, and safe fallbacks; Logo usage (safe area, min size, backgrounds, misuse); Imagery & icon style with keywords and a negative list; Layout system (grid, spacing scale); Common templates (social post, email signature, slide cover) with specs. Present the one-pager first, then the full guide. Keep it plain text so I can paste into a document.”

High-value twist (insider trick)

• Add three “anchor brands” you admire and a “forbidden list.” Ask AI to bias recommendations accordingly.
• Ask for “rules + reasons + examples” so your team understands the why, not just the what.
• Lock voice with a simple matrix: e.g., Warm 80%, Clear 90%, Bold 40%.

Copy-paste prompt (refinement pass)

“Refine the guide using these constraints: Anchor brands: [Brand A], [Brand B], [Brand C]. Forbidden list: [no neon gradients, no stocky handshake photos, no jargon]. Keep contrast readable (AA for body), simplify palette to 5 colors max plus neutrals, use Google Fonts only, and give usage examples for each rule. Return a ‘Minimal One-Pager’ followed by the ‘Full Guide v2.’”

What an AI-built guide typically includes

• Essence: Practical, human-first, optimistic.
• Colors: Primary #0B5FFF (60%), Secondary #111827 (20%), Accent #FF7A59 (10%), Background #F8FAFC (10%), Text #111827.
• Typography: Headings: Playfair Display → fallback Georgia/serif; Body: Source Sans 16/24 → fallback Arial/sans-serif. Size rules and spacing scale (4/8/16/24/32).
• Voice: Clear verbs, short sentences, avoid buzzwords. Example paragraph included.
• Logo: Safe area = 0.5× mark height, min width 120px, no overlays on photography without a solid background.
• Imagery: Natural light, real people, neutral backgrounds, subdued color temperature. Avoid staged stock.
• Templates: Social post grid, email signature with hierarchy, slide cover with headline + subhead + accent bar.

Common mistakes and fast fixes

• Feeding weak inputs: If you upload random or outdated materials, AI will mirror that. Fix: Choose your best 12 months.
• Too many colors: Over 7 colors causes chaos. Fix: Cap at 5 plus neutrals.
• Illegible text: Light gray on white looks chic but fails. Fix: Ask for contrast notes and darker text.
• Unlicensed fonts: AI might suggest paid fonts. Fix: Request Google Fonts or verify licenses before publishing.
• No “don’t” examples: Teams drift without guardrails. Fix: Ask AI to generate 5 misuse examples for logo, colors, and tone.
• Skipping a test run: A guide untested is theory. Fix: Produce one social post and one email using the new rules, then adjust.

90-minute action plan

1. Gather assets (15 min): logo files, 5 text pieces, 3–5 images, 1–2 layout screenshots.
2. Upload + run core extractor (20 min): use the prompt above.
3. Color/type/voice refinements (20 min): request accessibility, Google Fonts, usage percentages, and do/don’t lists.
4. Templates (15 min): ask for social, email signature, slide cover with exact sizes and spacing.
5. Test drive (10 min): have AI draft a LinkedIn post and a short email using the guide.
6. Final tidy (10 min): ask for a one-pager and a full guide ready to paste into your doc.

Bonus prompt (quick test drive)

“Using the Brand Quick Reference you just created, write a 120-word LinkedIn post and a 75-word email intro announcing our new offer. Keep voice and formatting on-brand. Then explain in 3 bullets how you applied the style rules.”

Reality check

• AI can condense, label, and standardize what you already do—and propose smart defaults where you haven’t decided.
• You make the final calls on taste, licensing, and compliance.
• Expect to iterate once or twice; the first output is the draft you improve.

Start small: get your one-page quick reference today. That single sheet will immediately make your posts, emails, and slides look and feel like one brand.

You’ve already nailed the right target: clear, kind, and specific. Those three words keep report card comments useful for families and motivating for students.

Why this matters

Parents want to know two things: what their child can do and what to do next. AI can help you say both quickly, warmly, and with evidence—without sounding robotic.

What you’ll need

• 3–5 bullet points of evidence per student (work habits, one concrete example, one growth area).
• Your school’s tone and grading policy (e.g., no comparative language, use strengths-first).
• Time box: 20–30 minutes for a first set of 10 comments.

The simple structure (3-bead necklace)

1. Strength with one precise example.
2. Growth with one precise example.
3. Next step with a supportive action (what family and student can do).

How to do it (step-by-step)

1. Collect quick evidence: copy your gradebook notes and one quote, score, or observable behavior per student.
2. Choose tone: warm–professional, plain language, grade-6 reading level.
3. Use the Master Prompt below to draft 2–4 sentence comments.
4. Refine: ask AI to shorten, soften, or add a specific example if it’s vague.
5. Batch: paste 5–10 students at a time using the Batch Prompt.
6. Final pass: read aloud once (the “ear-to-parent” test) and paste into your system.

Master Prompt (copy–paste)

Paste this into your AI tool and fill the brackets:

“You are a K–12 teacher-writer. Write a 2–4 sentence report card comment that is clear, kind, and specific using this structure: 1) strength + concrete evidence, 2) growth + concrete evidence, 3) next step + supportive tone. Constraints: grade-6 reading level, no jargon, avoid repeated phrases, positive-first, 80–120 words. Include one numeric or observable detail. Subject: [subject]. Grade level: [grade]. Student: [first name or “the student”]. Evidence bullets: [3–5 bullets with examples, e.g., ‘scored 18/20 on fractions quiz,’ ‘reads aloud with expression,’ ‘needs reminders to start tasks’]. What matters most to family: [e.g., confidence, independence, organization]. School tone notes: [e.g., do not compare to peers; keep growth mindset]. Pronouns: [she/he/they]. Output one comment only.”

Quick Fix Prompt (when AI is vague)

“Revise the comment to add one specific piece of evidence (score, quote, or observed behavior) and one concrete next step the family can try at home in one sentence. Keep it warm and under 110 words.”

Batch Prompt (for 5–10 students)

“Create separate report card comments using the structure strength–growth–next step. Each should be 2–4 sentences, kind in tone, and include one concrete example. Use grade-6 reading level. For each student below, write one comment labeled with their name. School tone: [notes].
– Student: [Name 1]; Subject: [Subject]; Evidence: [3–5 bullets].
– Student: [Name 2]; Subject: [Subject]; Evidence: [3–5 bullets].
– Student: [Name 3]; Subject: [Subject]; Evidence: [3–5 bullets].”

High-value insider tricks

• Progress delta sentence: “Since September, [student] moved from [starting point] to [current point] by [specific action].” It shouts growth without comparisons.
• One-number rule: Add exactly one number (score, frequency, pages read). Enough to be concrete, not cold.
• Family-friendly swap: Replace jargon (“executive function”) with plain phrases (“planning and organizing schoolwork”).
• Tone dial: Ask AI: “Make the tone 10% warmer but still professional.” Micro-adjustments beat rewriting.

Example outputs (what “good” looks like)

• Math, Grade 5: “Student shows steady problem-solving and checks work carefully; on the fractions quiz, they scored 18/20 after practicing equivalent parts. They now explain steps aloud and spot mistakes more quickly. Next, we’ll keep building confidence with multi-step problems by underlining key words and solving one part at a time; a few practice questions at home each week will help lock this in.”
• Reading, Grade 3: “Student reads with lively expression and chooses a wider range of books; they finished two chapter books this month and shared favorite parts in discussion. Stamina is growing, and they use picture clues to decode tough words. Next, we’ll practice pausing at commas and summarizing a page in one sentence; try a 10-minute family read-aloud to model this rhythm.”
• Science, Grade 8: “Student asks thoughtful questions and tests ideas carefully; their lab report on chemical reactions included clear tables and a correct conclusion. They now cite evidence more consistently, though they sometimes rush data entry. Next, we’ll slow down with a checklist before submitting; at home, a quick re-read to confirm units and labels will boost accuracy.”

Common mistakes and quick fixes

• Vague praise: “Works hard.” → Add evidence: “Completed all weekly writing drafts and improved sentence variety.”
• Future-only comments: “Needs to improve.” → Add growth: “Moved from 6/10 to 8/10 on inference questions by using text clues.”
• Too long: Cap at 2–4 sentences; ask AI: “Reduce to 95 words; keep the examples.”
• Copy-paste feel: Vary sentence openings: Ask AI for 3 phrasings and pick one.
• Jargon: Swap with plain language; prompt: “Replace jargon with family-friendly terms.”

Fast action plan (30 minutes)

1. Open your gradebook and copy 3–5 evidence bullets per student.
2. Paste the Master Prompt for your first subject and generate 5 comments.
3. Run the Quick Fix Prompt to add one number and one home strategy.
4. Read aloud, tweak pronouns and names, paste into your system.
5. Repeat for the next 5 students.

What to expect

• First batch may need a few tweaks as the AI learns your tone; by batch two, it will be faster and more on-voice.
• You’ll keep your professional judgment—AI drafts, you decide.
• Parents will hear a warm, specific story of progress and a simple next step they can support.

Closing thought

Clear, kind, specific comments come from small, concrete details. Feed AI the details, use the 3-bead structure, and you’ll produce comments that inform, encourage, and move learning forward—without the late-night slog.

Want to scale AI without losing human judgment? Smart — that balance is what makes AI useful, not just fast.

Good point to focus on practical, repeatable workflows. Below is a clear, hands-on plan you can try this week.

What you’ll need

• Defined task (e.g., content moderation, support triage, document summarization)
• AI model or service (starter: a reliable LLM or classification API)
• Human reviewers (part-time or full-time) with clear guidelines
• A routing system (simple queue or workflow tool)
• Metrics: accuracy, time-to-decision, % escalated

Step-by-step workflow

1. Map the decision points. Break the task into: auto-handle, auto-flag, escalate to human.
2. Create clear, short guidelines for humans (examples of accept/reject/modify).
3. Build a first-pass AI layer that: predicts label, confidence score, and a short rationale.
4. Route low-confidence or high-risk items to humans. Keep a random sample of high-confidence items for spot-checks.
5. Capture human edits and store them as training labels. Retrain or fine-tune periodically.
6. Monitor KPIs weekly and refine thresholds for auto-handle vs. escalate.

Example — content moderation for a blog

1. AI auto-rejects obvious spam (confidence >95%).
2. AI flags potential policy-violations with rationale; if confidence 60–95% send to reviewer.
3. Randomly sample 5% of auto-rejects and auto-accepts for human review.
4. Review decisions feed back weekly for model retraining and guideline tweaks.

Do / Do not checklist

• Do: Start small; use confidence thresholds and sampling.
• Do: Make human guidelines short, example-led, and revisable.
• Do: Log decisions and build a feedback loop.
• Do not: Assume the model is right without random audits.
• Do not: Flood humans with every decision—prioritize escalations.

Common mistakes & fixes

• Too many false positives: raise confidence threshold, improve examples for model.
• Humans lack consistency: run calibration sessions and use checklists.
• No feedback loop: tag reviewed items and retrain monthly.

Copy-paste AI prompt (use as the first-pass processor)

Prompt: “You are an assistant that reviews user-submitted content. Provide: 1) a 2-sentence neutral summary; 2) classification tag(s) from [safe, spam, hate, adult, other]; 3) a confidence score (0-100); 4) a one-line rationale explaining the decision. Use simple, factual language.”

30/60/90 day action plan

1. 30 days: Pilot with one workflow, set thresholds, begin sampling audits.
2. 60 days: Implement feedback loop, run weekly KPI reviews, adjust routing rules.
3. 90 days: Retrain model with labeled data, expand to next workflow.

Keep it iterative: deploy small, measure, fix, and scale. Human-in-the-loop at scale is about rules, sampling, and continuous learning — not perfection from day one.

Good question — you’re asking the exact practical question many founders over 40 should ask: can AI actually write a pitch and build a targeted media list that works? Short answer: yes — if you give AI the right inputs and do a little human polishing.

Why this matters: an AI draft gets you out of blank-page paralysis and produces many tailored ideas fast. Your job is to guide it, refine the voice, and respect journalist preferences.

What you’ll need

• Clear niche description (one sentence)
• Target audience and angle (who cares and why)
• Key facts: data points, launch date, quotes, assets (images, one-pager)
• Examples of coverage you like (3 outlets or article titles)

Step-by-step (do-first, quick wins)

1. Prepare inputs: write your one-sentence niche and three selling points.
2. Use the AI prompt below to generate 3 short pitch variants and a media list draft.
3. Pick the best pitch, tighten the subject line to 6–9 words, and add a single, compelling data point in the first paragraph.
4. Review the media list: verify each journalist’s recent work, beat, and contact method (email or social).
5. Personalize before sending: reference a recent story or the reporter’s beat in one sentence.

Copy-paste AI prompt (use this exactly)

“You are a PR specialist. I run [one-sentence niche]. My target audience is [audience]. Key selling points: 1) [point], 2) [point], 3) [point]. Provide: A) three short email pitches (subject line + 3-sentence body each) tailored to tech/business/lifestyle reporters; B) a targeted media list of 12 outlets/journalists with beat, why they’d care, and suggested angle. Keep the tone professional but warm. Include a 20-word boilerplate about the company and one suggested quote from the founder.”

Example output (short)

• Subject: New AI tool cuts freelance taxes by 40% — story?
• Pitch: Hi [Name], our AI helps freelancers reduce tax time by automating expense tracking and claiming credits. We’ve tested it with 200 users and cut average tax bills by 40%. Founder available for interview and to share user stories.
• Media list entry: TechToday — Sarah Jones — freelance economy/fintech — wrote about gig-worker benefits last month — angle: data-driven savings for independent workers.

Mistakes & fixes

• Too generic: fix by adding one specific metric or customer example.
• Long subject lines: fix to 6–9 words and include the hook.
• Mass-send without personalization: fix by referencing a recent story or beat.

7-day action plan

1. Day 1: Gather inputs and run the AI prompt.
2. Day 2: Select and refine pitch variants.
3. Day 3–4: Vet and personalize the top 30 journalists.
4. Day 5: Send 5 highly personalized pitches.
5. Day 6–7: Follow up and track responses.

Reminder: Use AI to accelerate work, not replace judgment. Test, measure opens/replies, iterate fast. Small, polished outreach beats bulk blasting every time.