Jeff Bullas

Hook: If messy notes feel like clutter, there’s a simple path to a polished post — extract the bones, ask the right AI prompt, then tidy. You’ll get publishable content faster than you think.

Why this works: Notes are raw ideas. The goal is to turn raw into structure — headline, three points, one example, and a payoff. That small scaffold makes editing easy and reduces writer’s block.

What you’ll need:

• One page of messy notes (photo or text)
• A text editor (Notes, Word, Google Doc)
• A timer (5–20 minutes)
• An AI tool you can paste prompts into (optional but powerful)

Step-by-step routine (10–25 minutes):

1. Scan (2–3 minutes): skim notes and circle the main idea. Pick one angle you care about today.
2. Extract (5 minutes): write one headline and three bullets: problem, quick fix, result for the reader.
3. AI assist (5–10 minutes): paste notes and use a focused prompt (below) to get a draft or outline.
4. Polish (5–10 minutes): edit for clarity, add one concrete example, tighten the CTA and headline.
5. Save & schedule: store as draft, decide publish date or expand later.

Example — messy note to paragraph:

Messy note: “talked about short routines, 5-min headline, use bullets, do weekly review, example: newsletter saved time”

Polished intro: Short writing routines beat long, intimidating sessions. Start with a five-minute headline and three bullets, then turn each bullet into a short paragraph. I did this for my newsletter and cut drafting time in half while keeping clarity for readers.

Best copy-paste AI prompt (use as-is):

Here are messy notes: [paste notes]. Turn these into: 1) one clear headline, 2) a 3-bullet outline (Problem / Quick fix / Benefit), and 3) a 300–350 word friendly blog post in a warm coach tone with one concrete example and one-sentence CTA. Keep reading time ~2 minutes.

Mistakes & fixes:

• Giving the AI raw noise: fix by first writing a one-line angle (helps focus output).
• Over-editing: fix by saving the first AI draft, then do one pass for clarity and one for voice.
• Too long before publishing: fix with the 20-minute limit — publish then iterate.

Quick 7-day action plan:

1. Day 1: Pick one page, follow the routine, publish a short post.
2. Days 2–4: Repeat with new notes — aim for 15–20 minutes each.
3. Days 5–7: Review drafts, pick one to expand into a longer post or newsletter.

Small, steady steps beat perfect starts. Try the five-minute extract and the AI prompt above — you’ll be surprised how quickly messy notes turn into content you can share.

Great call-out: Your note on using fixed test windows and aiming for a minimum sample before deciding is spot on. Here’s how to layer one more win on top of that: make your flow “behavior-aware” so each email reacts to what the user actually did — not just where they are in time.

Upgrade: time-based to behavior-based

• Do branch your flow by state: inactive, attempted but stuck, activated.
• Do add one-click micro-survey links (e.g., “No time”, “Confused”, “Blocked by IT”) to capture the obstacle and personalize the next email.
• Do use preheader text to preview the benefit and the one action (many teams skip this and leave free performance on the table).
• Don’t send the same Day 2 email to someone who already tried and failed — send a “fix” email instead.
• Don’t chase opens; keep your north star on activation rate and time-to-activation.

What you’ll need

• Your current setup (email tool + A/B testing + basic analytics).
• Two extra user flags: attempted_activation (yes/no) and error_reason (optional text or a short list).
• Ability to add preheader text and a monitored reply-to inbox.
• An AI assistant to draft variants for each branch and to summarize reply-to messages.

How to do it — step by step

1. Map the “magic moment” and friction. Write the one activation action and list the top 3 blockers (time, confusion, access).
2. Set state flags. Track three states: inactive (no attempt), attempted but stuck (clicked or started but didn’t finish), activated (completed).
3. Design three branches.
   1. Inactive branch: quick benefit + “one-minute start.”
   2. Stuck branch: empathetic fix + 2–3-step mini-checklist.
   3. Activated branch: celebrate + next best action to reinforce value.
4. Personalize lightly. Use first name and plan. Add a one-click micro-survey for blockers in the stuck email; tag the click.
5. Write with an AI template. Ask for 3 subject lines, 1 preheader, and a 2–4-line body with one button for each branch. Keep a plain, human voice.
6. Implement conditional sends. Day 0 to all. Day 2 splits to inactive vs stuck. Day 5 nudges inactive again; Day 5 for stuck is a “fix” follow-up based on the micro-survey tag. Day 7 is last-chance or a “you did it” upgrade path for activated users.
7. Test in order. Subject line → CTA wording → timing → branch logic. Fixed test window (2–4 weeks) or until you hit your minimum sample per variant.
8. Review replies. Let AI summarize reply-to messages weekly into 3–5 themes; turn those themes into the next test ideas.

Worked example (SaaS trial; activation = “upload first file”)

1. Day 0 (all) — Subject: “Welcome — your fastest path to value” | Preheader: “Upload one file to unlock sharing in minutes.” Body: 2 short lines, button “Upload your first file”.
2. Day 2 (branch)
   • Inactive: “Still setting up? It takes one minute.” Preheader: “Drag, drop, done.” CTA: “Upload now”.
   • Stuck: “Looks like the upload didn’t finish.” Preheader: “3 quick fixes inside.” Body: 3 bullets: check file size, try the web uploader, or contact support. CTA: “Try the quick fix”. Include micro-survey links: “No time” | “Confused” | “Blocked at work”.
3. Day 5 (branch)
   • Inactive: Social proof + 1-liner benefit. CTA: “See how teams start in 60s”.
   • Stuck with “Blocked at work” tag: Send a plain-text style email offering an IT-friendly alt path (e.g., smaller file or approved domain). CTA: “Use the IT-safe option”.
   • Activated: Celebrate + next step (invite a teammate). CTA: “Invite one teammate”.
4. Day 7 (last chance) — Case study line + checklist. CTA: “Finish setup now”.

Insider trick: preheader + button pairing

• Use the preheader to promise the outcome in 7–10 words.
• Label the button with the exact action (“Upload your first file”), not “Get started”.
• Keep only one link per email (plus the micro-survey choices in stuck emails).

Copy-paste AI prompt (behavior-aware flow)

“You are a senior lifecycle marketer and email copywriter. Create a 3-branch onboarding flow for a product where the activation event is [ACTIVATION_EVENT]. Branches: (1) Inactive (no attempt), (2) Stuck (attempted but did not finish), (3) Activated (completed). For each branch, write 2 emails (Day 2 and Day 5) with: 3 subject lines, 1 preheader (50–80 characters), and a body of 2–4 short lines with one clear CTA button label. Use friendly, plain language and include {{first_name}}. For the Stuck branch, include a 3-bullet quick-fix checklist and add three one-click micro-survey options that I can hyperlink: “No time”, “Confused”, “Blocked at work”. End with a short plain-text style variant for low-engagement users.”

Common mistakes & fixes

• Mistake: Same content for all states. Fix: Split by inactive vs stuck; send a fix-first email to “stuck”.
• Mistake: Vague CTAs. Fix: Use the exact action as the button label.
• Mistake: Ignoring preheaders. Fix: Treat them like a second subject line — promise the outcome.
• Mistake: Testing too many things at once. Fix: Follow a simple test order and fixed window.
• Mistake: Letting replies pile up. Fix: Use AI weekly to summarize reply themes and turn them into tests.

7-day action plan

1. Day 1: Add two flags in your tool: attempted_activation and error_reason (or a simple “stuck” yes/no).
2. Day 2: Use the behavior-aware prompt above to generate copy for inactive, stuck, and activated branches.
3. Day 3: Implement Day 0/2/5/7 with conditional sends. Add preheaders and a monitored reply-to.
4. Day 4: Launch your first A/B on the subject line for the Day 2 inactive email. Fixed window: 2–4 weeks or until your minimum sample is reached.
5. Day 5: Set up micro-survey links in the stuck email and tag clicks to “No time / Confused / Blocked”.
6. Day 6: Create a plain-text style fallback for low-engagement users (one link, no images).
7. Day 7: Review early signals: activation rate, time-to-activation, and reply themes. Choose the next single test (CTA wording or timing).

Closing thought: AI makes writing fast; behavior-aware logic makes it effective. Start with one branch split this week (inactive vs stuck), pair it with a clear preheader and action-labeled button, and watch activation move.

Let’s lock in an inbox that sorts itself — quick wins today, smarter automation by the weekend. We’ll blend simple rules (fast) with a focused AI classifier (smart) and add guardrails so nothing critical goes missing.

Where AI fits: rules catch the obvious; AI handles the fuzzy messages that waste your attention. Start in “suggestion” mode for 2–3 days, measure, then let high-confidence items move automatically. Expect rapid improvement as you correct edge cases.

What you need

• Email admin access (Gmail or Outlook).
• 6–8 labels with clear intent (Action-High, Action-Low, Finance, Clients, Newsletters, Internal, Optional: Uncertain).
• 20–50 example emails per label (kept in folders for testing).
• Optional: Zapier, Make, or Power Automate; an AI API key if using a third-party model.
• A simple privacy plan (see notes below).

Build the foundation (15–30 minutes)

1. Define labels + response rulesKeep labels purpose-driven. Add a rule per label (e.g., Action-High: reply within 4 hours; Finance: review daily).
2. Create native filters (fast wins)Gmail: Settings > See all settings > Filters and Blocked Addresses > Create a new filter. Use From (domains), Subject (keywords), and Has the words (invoice, unsubscribe, receipt). Apply label + auto-archive if appropriate (e.g., Newsletters).Outlook: Rules > Manage Rules & Alerts > New Rule. Use sender/domain, subject phrases, and add “assign to category” and “move to folder.”
3. Collect examplesCreate temporary folders like TRAIN-Finance, TRAIN-Clients. Drag 20–50 real emails into each. These power your AI prompt testing.

Smart layer: AI classifier flow (Zapier/Make/Power Automate)

1. Trigger: New email (exclude anything your native rules already filed).
2. Sanitize (privacy optional but wise): replace email addresses, names, and numbers with tokens before sending to AI (e.g., [EMAIL], [NAME], [AMOUNT]).
3. AI step: send the email text + the prompt below. Get back JSON with label and confidence.
4. Decision: if confidence >= 80 and label ≠ Uncertain, apply label and move; else add label Uncertain or prefix subject with [SUGGESTED: Finance].
5. Log: save predictions and your corrections to a simple spreadsheet for weekly tuning.

Insider template: Routing policy + robust AI prompt (copy-paste and customize)

“You are my email routing assistant. Use this policy and return ONLY JSON.

Labels and purpose:• Action-High: time-sensitive requests, commitments, clients with deadlines.• Action-Low: non-urgent tasks, FYIs that require action later.• Finance: invoices, receipts, payments, contracts with amounts/dates.• Clients: client comms that are not primarily finance or urgent action.• Newsletters: marketing/news updates, bulk mail, promotions.• Internal: team, HR, ops from our domains.• Uncertain: use only when confidence < 75 or content conflicts.

Guardrails:• VIP senders always Action-High: [ADD VIP EMAILS OR DOMAINS].• If email contains invoice/receipt/payment with amounts or PO numbers → Finance unless it’s clearly a newsletter.• If multiple labels apply, choose the single most urgent/purposeful label (prefer Action-High > Finance > Clients > Internal > Action-Low > Newsletters).• If unsubscribe/footer-heavy and no request → Newsletters.• If the message is a calendar invite, treat as Action-Low unless urgent language indicates Action-High.

Output ONLY this JSON schema:{“label”:”one label from the list”,”confidence”:0-100,”reason”:”short phrase”,”summary”:”one-line”,”escalate”:true|false}

Set escalate=true if label is Action-High or Uncertain. Do not include any extra text.

Email to classify: [PASTE EMAIL TEXT HERE]”

Pro tip: include 3–5 short examples in the prompt (few-shot learning) once you have them. Accuracy usually bumps meaningfully without code changes.

Example (what good looks like)

Input email: “Hi — please approve the attached invoice for $3,200 for Client X by Friday.”

Expected JSON:{“label”:”Finance”,”confidence”:92,”reason”:”Invoice approval with amount and deadline”,”summary”:”Approval request for $3,200 invoice by Friday”,”escalate”:true}

Rollout and expectations

• Day 1–2: Rules will catch 40–60% immediately.
• Day 3–5: AI in suggestion mode will route most of the fuzzy 30–40% at 70–85% confidence.
• Week 2: With corrections and 5–10 examples added to the prompt, high-confidence auto-move often reaches 80–90% precision for your core labels. Your results will vary; keep a light review cadence.

Mistakes to avoid (and the fix)

• Label sprawl: more labels = lower accuracy. Fix: stay under 8 and merge low-volume categories.
• Sender-only rules: forwards and aliases break them. Fix: add content keywords and negative keywords (e.g., exclude “unsubscribe” for Clients).
• No safety net: everything moves even when uncertain. Fix: use Uncertain + escalate=true for low confidence.
• One-shot prompts: never updated. Fix: add 3–5 real examples and one hard edge case per label after week 1.
• Privacy blind spots: sharing PII. Fix: tokenize sensitive data or use provider-native AI when required.

90-minute sprint

1. List your 6–8 labels and write one sentence of purpose for each.
2. Build 6–10 native filters for the obvious senders/keywords; auto-file Newsletters.
3. Create TRAIN- folders; drop 20 emails into each.
4. Set up the AI flow in suggestion mode; apply [SUGGESTED: Label] prefix.
5. Review 25 suggestions; correct and note patterns to add to guardrails.

Week 2 tune-up

• Raise auto-move threshold to 80–85% for low-risk labels (Newsletters, Internal).
• Add VIP overrides and 5 few-shot examples to the prompt.
• Track: % auto-labeled, false positives, manual corrections, time-to-action on Action-High.

Closing thought: keep this lightweight. Rules do the heavy lifting, AI cleans up the messy middle, and your weekly tweaks make it feel like magic — without trusting your inbox to chance.

Agreed on your two smart points: quick A/B with unique QR codes beats more design tweaks, and character counts are only a guide because fonts behave differently. Great instincts. Let’s add one layer that locks tone, fits the frame, and speeds approvals.

Try this in 5 minutes: paste one sentence you’ve written that feels perfectly “you” into the prompt below. It builds a tiny voice snapshot you can reuse in every flyer prompt.

Copy-paste prompt: 60‑Second Voice Snapshot

“Analyze this sentence for brand tone: [paste your sentence]. In 80 words or less, produce a micro-brief with 3 Do’s, 3 Don’ts, 5 signature phrases I naturally use, and 5 banned words/phrases I avoid. Keep it plain and practical. Ask me for one tweak, then finalize the micro-brief in a single paragraph I can paste into future prompts.”

Why this matters: AI matches voice when it sees concrete rules and examples. Add one benefit + one proof, and you get copy that sounds like you and converts.

What you’ll need

• Your logo, 1–2 fonts, and up to 3 hex colors
• One approved sentence that is definitely “you” and three voice words (e.g., warm, practical, confident)
• Event facts: who, what, when, where
• One-line benefit and one credible proof (e.g., “200+ attendees last year” or “free templates included”)
• QR/short link generator to make unique, trackable codes

The simple system (repeatable)

1. Make your voice micro-brief. Use the snapshot prompt above. Keep it nearby and paste it into every copy request.
2. Build a message map. Three tiny lines: Benefit, Proof, Next step (CTA). This keeps the flyer persuasive, not fluffy.
3. Generate layout-safe copy. Ask for tight ranges (headline 3–6 words, body 25–35 words) and plain English. Include your micro-brief and sample sentence.
4. Self‑critique loop. Have the AI score its own draft against your three voice words (1–5), then revise to improve the lowest score.
5. Fit‑to‑frame rewrite. After you test the design visually, give the AI your actual line limits (“headline must fit in 26–30 characters in [font name]”) and ask it to preserve meaning while shaving characters.
6. Test with QR variants. Two headlines or two CTAs, each with its own QR. Check scans at 48 hours, then scale the winner.

Copy-paste prompt: Message‑Map to Flyer Copy

“Create flyer copy that fits a simple three‑zone layout. Use my micro-brief: [paste your finalized micro-brief]. Voice words: [3 words]. Sample sentence: [paste yours]. Audience: [age/role]. Event: [what/when/where]. Benefit: [one line]. Proof: [one line]. Produce: 3 headlines (3–6 words), 2 subheads (6–12 words), 1 body paragraph (25–35 words), and 2 CTA lines (3–5 words). Rules: plain language, light warmth, no clichés, avoid my banned words. Label each line clearly.”

Copy-paste prompt: Self‑Critique and Fit‑to‑Frame

“Score the draft below against my voice words [list] from 1–5 each and explain why in one short line. Rewrite once to raise the lowest score. Then provide a second version that fits these visual limits in [font name]: Headline max ~[X] characters on one line; Subhead ~[Y] characters; Body 25–35 words; CTA 3–5 words. Keep meaning and tone.”

Worked example (so you can see it)

• Voice words: calm, plainspoken, optimistic
• Benefit: Sleep better in 7 nights
• Proof: Doctor-led tips you can use tonight
• Event: Sleep Better Workshop — Tue, May 28, 6–7:30pm, Oak Center

• Headline options: Sleep You Can Keep • Rest Starts Here • Better Nights, Finally
• Subhead: Practical steps for stress‑free sleep
• Body (31 words): Learn simple routines, bedroom tweaks, and wind‑down habits that actually work. Doctor‑led session with time for questions. May 28, 6–7:30pm, Oak Center. Bring a friend and compare notes.
• CTA: Save your seat

Insider tips that save time

• Use a “fallback” headline. Keep one pre‑approved, shorter headline in your template for tight spaces or narrow posters.
• Set a reading level. Ask for “Grade 6–8 reading level, no jargon.” It increases scan‑and‑understand.
• Control rhythm. Request: “Body = 2 short sentences, 1 medium sentence.” It keeps the copy airy and legible.
• Proof helps persuasion. A tiny proof line (social proof, guarantee, inclusion of snacks/materials) can lift RSVPs without adding clutter.

Common mistakes and quick fixes

• Looks perfect on screen, wraps in print. Fix: print at 100% on office paper and check line breaks before you run the job.
• Busy background steals contrast. Fix: add a translucent overlay behind text or choose a simpler image crop.
• QR scans slowly. Fix: increase size, ensure a white quiet zone, and test from the typical viewing distance.
• Tone drifts after several edits. Fix: paste the micro‑brief and sample sentence into every new prompt and request a self‑critique pass.
• Generic benefits. Fix: force one specific outcome (“meet two new contacts”) and one proof (“sponsor demo, 20 mins”).

Your 48‑hour plan

1. Today (20–30 min): Create your voice micro‑brief. Draft a Benefit/Proof and build or open your locked template.
2. Today (20–30 min): Run the Message‑Map prompt to get 3 headlines, 2 subheads, body, and CTAs. Drop the top pick into the template and test print.
3. Tomorrow (20–30 min): Make Variant B (change headline or CTA), generate two unique QR codes/links, and distribute both to the same list or locations.
4. Tomorrow (5 min): Check scans after 48 hours. Keep the winner, archive the micro‑brief, and reuse the system for your next event.

What to expect

• 1–2 revisions to nail tone the first time. Faster thereafter.
• Cleaner layouts because copy arrives with limits and rhythm baked in.
• Measurable lift from small A/B changes, especially in the headline or CTA.

Bottom line: You’ve got the testing loop right. Pair it with a tiny voice micro‑brief, a message map, and a self‑critique pass, and AI will deliver flyers and posters that sound like you, fit your layout, and earn the RSVP.

Totally agree: your quick quality gates and the small node template are the secret to repeatable results. Let’s add a few pro-level tweaks that make hair highlights behave, fabric look touchable, and your AI swatches drop straight into production with less fiddling.

What you’ll add to your kit

• A “raking light” HDRI or a simple key light at a low angle for fast tests.
• A checker/ruler plane in-scene (1 cm grid) for instant scale sanity checks.
• An extra micro-noise texture (tiny 256–512 px tiling noise) to break repeats.

Step-by-step: elevate the fabric and hair workflow

1. Prompt with measurable detail: Ask for fiber size and density, not just style. This helps scale look right from the first render and makes tiling safer.
   • Specify weave type (plain weave, twill, knit) and thread thickness (e.g., 0.2–0.4 mm).
   • Ask for slight imperfections (fuzz, pilling, flyaway strands) but in neutral lighting.
2. Split your detail into two normals: make a macro normal (weave or clump shapes) and a micro normal (fine fibers). Combine them in your shader. Macro sells form; micro sells realism.
   • Macro: use the height from your high-pass at a larger blur radius; Normal strength ~0.4–0.7.
   • Micro: duplicate the height, higher contrast, smaller blur; Normal strength ~0.1–0.3.
3. Roughness that reads correctly: remember bright = rough, dark = shiny.
   • Fabric: Mostly bright (0.6–0.9) with slightly darker streaks where thread is polished or pressed.
   • Hair: Darker along the fiber tangent for strong anisotropic streaks; keep roots slightly rougher.
4. Fabric shader finishing:
   • Use a small sheen value (0.2–0.4) instead of pushing specular. Cloth loves sheen.
   • Add a subtle backscatter/subsurface if the fabric is thin (0.05–0.15) for a soft edge glow.
   • Mix in a low-strength micro-noise normal (scale 2–4x the base) to kill pattern repetition.
5. Hair shader finishing:
   • Base: use your color swatch plus a root-to-tip gradient (5–10% darker at the root).
   • Anisotropy: start at 0.7–0.9; rotate the tangent map to match strand direction.
   • Longitudinal roughness low (0.15–0.3), azimuthal roughness higher (0.6–0.8) for pleasing streaks.
   • Add a clump mask (soft, low-contrast) into roughness so highlights bunch like real hair.
6. Scale once, reuse forever: place your swatch on the checker plane. Count threads per cm. Adjust until the knit/weave matches reality. Save that scale inside your node template for future drops.
7. Fast quality gate: one 10–15 second render under raking light and one under your go-to HDRI. If highlights sparkle or smear, fix in this order: micro-normal strength → anisotropy rotation (hair) → roughness contrast → macro-normal strength.

Copy-paste AI prompts (production-safe)

• Fabric, refined: “Create a seamless, tileable 4096×4096 close-up fabric texture: tight plain weave cotton, thread diameter 0.3 mm, 28–32 threads per cm, subtle microfiber fuzz, neutral flat top-down lighting, natural warm gray, no shadows, no logos, repeat-safe, clean edges, high-frequency microdetail suitable for PBR.”
• Hair, refined: “Generate a seamless 4096×4096 hair fiber map: straight medium-density human hair, natural dark brown with gentle melanin variation, visible strand direction left-to-right, slightly darker roots, a few flyaway strands, neutral lighting, no background, no text, tileable and repeat-safe.”

Example setups that just work

• Denim: Macro normal 0.6; micro normal 0.2; roughness 0.7 base with 0.55 streaks on raised threads; sheen 0.25; tiny blue tint in albedo; optional displacement 0.1–0.2 mm for hero shots only.
• Brown straight hair: Anisotropy 0.85; longitudinal roughness 0.2; azimuthal 0.7; micro normal 0.15; macro normal 0.35; root-to-tip gradient -8% value; subtle clump mask multiplied into roughness.

Insider trick: build a 30-second “gloss ramp test.” Create a plane with five roughness stripes (0.2, 0.4, 0.6, 0.7, 0.85). Drop your roughness map over it as a multiply. Render under raking light. If the map collapses into one tone, increase contrast slightly; if it looks glittery, blur the micro areas by 1–2 px.

Common mistakes and fast fixes

• Over-sharpened normals → Halos and sparkle. Fix: reduce micro-normal strength first; add a tiny blur to the height before conversion.
• Tiling too obvious → Break it with a low-contrast micro-noise overlay and 2–3 color variations you randomly mix per object.
• Plastic-looking fabric → Lower specular, raise sheen slightly, push roughness brighter overall.
• Flat hair highlights → Increase anisotropy and rotate the tangent/flow map to match strand direction; darken roughness a touch along the fiber.
• Scale drift between assets → Lock a standard: “plain weave cotton = 30 threads/cm,” and store that in your node group note.

4-day action plan

1. Day 1: Generate 3 fabric and 3 hair swatches using the prompts. Tile-test and clean seams. Save at 4096 if possible.
2. Day 2: Build macro and micro height/normal pairs; create roughness maps; assemble two node templates (fabric + hair) with saved scale values.
3. Day 3: Run the gloss ramp and raking light tests. Tweak only three knobs: micro-normal, anisotropy rotation, roughness contrast. Save presets.
4. Day 4: Apply to a real scene. Add micro-noise to kill repeats, finalize color tint in-shader, and export the package (albedo, macro/micro normals, roughness, optional displacement, gradient/clump for hair).

Closing thought: Keep it boring and repeatable. Two normals (macro/micro), a smart roughness, and one quick raking-light check will get you to “real” fast. Template it once, and every new AI swatch becomes a 15-minute drop-in rather than a half-day build.

Quick win: you can move from chaotic inbox to predictable system in a few hours — not weeks — by combining native filters with a simple AI classifier.

Below is a clear, non-technical plan, a ready-to-use AI prompt, examples, and traps to avoid. Follow this and you’ll see measurable time saved within days.

What you’ll need

• Email account admin access (your Gmail or Outlook settings).
• 6–10 meaningful labels (Action-High, Action-Low, Finance, Clients, Newsletters, Internal is a good starter list).
• 20–50 sample emails per label (move to a folder or export copies).
• Optional: automation tool account (Zapier, Make, or native scripts) and an AI API key if you use a third-party model.

Step-by-step (do this)

1. Define labels and a simple SLA per label (e.g., Action-High — reply within 4 hours).
2. Create native filters first: sender rules, domain rules, and subject keywords to catch obvious messages (~40–60%).
3. Collect and tag 20–50 sample emails per label — store them for training/testing.
4. Use the AI prompt below to classify ambiguous emails. Run it in suggestion mode (add a prefix like [SUGGESTED] to subject or add a temporary draft) for 48–72 hours.
5. Review suggestions daily, correct mistakes (these corrections improve prompt / rules). Track % auto-labeled and manual corrections.
6. When stable, switch high-confidence predictions (confidence > 80) to auto-label and auto-move.

Ready-to-copy AI prompt (single-label JSON output)

Use exactly as-is. Paste the full email text where prompted.

“You are an email triage assistant. Labels: Action-High, Action-Low, Finance, Clients, Newsletters, Internal. Read the email below and return ONLY a JSON object with these keys: label (one of the labels), confidence (0-100), reason (one short sentence). Do not add any other text. Email: “[PASTE EMAIL TEXT HERE]””

Variant — multi-label with summary

“You are an email triage assistant. Return ONLY JSON with keys: labels (array of labels), confidence (0-100), summary (one-line), reason (one short sentence). Include multiple labels only if confidence > 70 for each.”

Example

Sample email: “Hi Sarah — please approve the attached invoice for $3,200 for Client X. Need confirmation by Friday so we can pay.”

Expected JSON output (single-label prompt):

{“label”:”Finance”,”confidence”:92,”reason”:”Invoice approval requested with payment deadline.”}

Common mistakes & fixes

• Over-reliance on sender-only rules — add content checks so a forwarded thread doesn’t misroute.
• Too many labels — consolidate to purpose-driven categories for faster decisions.
• Privacy concerns — anonymize or use provider-native automation if you can’t send email bodies to third-party AI.

7-day action plan (fast)

1. Day 1: Define labels and SLAs.
2. Day 2: Create native filters for obvious cases.
3. Day 3: Gather sample emails per label.
4. Day 4: Run AI prompt in suggestion mode on new mail.
5. Day 5: Review, correct, refine prompt and filters.
6. Day 6: Enable auto-label for >80% confidence items.
7. Day 7: Measure % auto-labeled and adjust.

Start small, iterate fast. Do the native filters first, add AI for the fuzzy cases, and move to auto-mode when confidence is consistent. That’s where the real time-savings live.

Good focus on meal prep and batch cooking — that’s the smart place to start. AI makes the boring planning work disappear so you can cook less and live more.

Why this helps

AI can turn your food preferences, fridge contents and time limits into a practical, repeatable plan: recipes that scale, a grouped shopping list, a clock-by-clock prep schedule and storage/reheat tips so meals stay tasty.

What you’ll need

1. Phone or computer with an AI chat tool (any chatbot will do).
2. Quick notes: dietary needs, favorite ingredients, serving sizes, available fridge/freezer space.
3. Basic kitchen tools for batch cooking (large pots, baking trays, airtight containers).
4. 30–90 minutes to set up the first plan.

Step-by-step plan

1. Define goals: How many meals? How many people? Any calorie or diet limits?
2. Inventory: List what you already have (proteins, grains, veggies, spices).
3. Ask AI for a weekly plan: Request recipes that batch-cook, portion instructions and a 2–3 hour prep day schedule.
4. Get a grouped shopping list: AI outputs items by store section (produce, dairy, pantry) to save time.
5. Prep day script: Follow the AI’s timed steps so dishes finish around the same time.
6. Label & store: Use dated labels and simple reheat instructions the AI provides.

Example

If you want 10 lunches + 6 dinners for two people: AI suggests three protein bases (roast chicken, lentil chili, baked salmon), two grain bases (rice, quinoa), and three vegetable sides. It gives a 3-hour prep schedule: roast chicken (60 min), chili (45 min overlap), grain batch (30 min), veg sheet pan (25 min). It creates fridge/freezer rules.

Copy-paste AI prompt (use as-is)

“I need a one-week meal prep plan for 2 adults: 10 lunches and 6 dinners. We are omnivores, prefer Mediterranean-style flavors, one gluten-free person, no dairy. I have 2 hours Saturday to batch-cook. Create 3 main recipes that scale, a detailed prep schedule broken into 15-minute steps, a shopping list grouped by store section, storage/reheat instructions, and suggested portions per meal.”

Common mistakes & quick fixes

• Too many recipes — stick to 3–4 bases. Fix: ask AI to simplify.
• Ignoring fridge space — measure before shopping. Fix: tell AI your container count.
• Not labeling — always date and name. Fix: use the AI’s simple label text.

Action plan — do this today

1. Note dietary needs and how many meals you want.
2. Open an AI chat and paste the prompt above.
3. Review the plan, tweak one recipe, and approve the shopping list.
4. Do a single 2-hour prep this week to test the system.
5. Adjust portions and storage after tasting day one.

Start small, measure what worked, then scale. AI speeds the planning — you get the time back in your week.