Jeff Bullas

Quick win: Use AI to generate 3-minute exit tickets that tell you what students learned — without eating your evening.

Why this works

Exit tickets and short formative assessments should be fast to make, quick to grade, and focused on one clear learning target. AI accelerates writing questions, gives instant answer keys, and helps differentiate for students who need more support.

What you’ll need

• A conversational AI tool (like ChatGPT) or other text generator.
• A clear learning objective (one sentence).
• 5–10 minutes to craft and review the AI output.
• Optional: a simple form (paper, Google Form, LMS quiz) to collect answers.

Step-by-step: create an exit ticket in 6 minutes

1. Write a one-sentence learning objective. Example: “Students will be able to identify the main idea and two supporting details of a short paragraph.”
2. Use the AI prompt below (copy-paste) and paste your objective, grade, and time limit.
3. Ask the AI to produce 4 short items: 2 quick-response (multiple choice or short answer), 1 application question, 1 reflection/self-assessment.
4. Scan and edit for clarity (2 minutes). Simplify language for younger learners and shorten questions for exit-ticket speed.
5. Decide how to collect answers (raise hands, sticky notes, short form). Collect and glance for patterns — 10 seconds per response when possible.

Copy-paste AI prompt (use as-is, replace bracketed text)

“Create a 4-question exit ticket for [grade level] about this learning objective: [paste objective]. Include: 2 quick-response questions (multiple choice or 1–2 word short answers), 1 application/problem-solving question (2–3 sentences), and 1 student reflection/self-assessment prompt. Provide answer key and a brief teacher note on what to look for in responses. Keep language simple and completion time under 5 minutes.”

Prompt variants

• Differentiate: add “also provide one simplified version for students who need support and one extension question for students who finish early.”
• Subject-specific: start with “For middle school science/elementary math/high school history…”

Example exit ticket (from the prompt)

• MC: What is the main idea of this paragraph? A/B/C/D
• Short answer: List two supporting details from the paragraph.
• Application: Rewrite the main idea in your own words and give an example from your life (1–2 sentences).
• Reflection: How confident are you with this skill? (1) Not yet (2) Somewhat (3) Confident

Mistakes teachers make — and quick fixes

• Too many questions: limit to 3–5 items. Fix: pick one target skill per exit ticket.
• Questions too complex: simplify language. Fix: ask AI for a “simplified student version.”
• No quick scoring: create a 0–2 rubric for each item so you can scan answers fast.

Action plan — do this today

1. Write one learning objective for tomorrow’s lesson (1 minute).
2. Paste it into the copy-paste prompt above and generate an exit ticket (3–5 minutes).
3. Pick a collection method (sticky note, Google Form) and use it tomorrow.

Final reminder

Start small: one clear target, one AI prompt, one quick check. Iterate based on what student answers tell you. Small, regular checks beat perfect tests every time.

Good question — and you’re on the right track. AI can absolutely help write engaging Twitter threads with strong hooks and CTAs if you guide it with clear prompts and edit like a human.

Here’s a practical, step-by-step way to get quick wins and build a repeatable process.

What you’ll need

• One clear topic or insight you care about.
• A few facts, anecdotes or links to reference (even one is enough).
• An AI writing tool (chat-based) and 15–30 minutes to refine.

Step-by-step: produce a high-converting thread

1. Define the single idea: write one-sentence thesis (what readers will gain).
2. Ask AI for a tight hook: 1 short sentence that promises value or surprises.
3. Outline 5–7 tweets: problem, evidence, steps, quick example, CTA.
4. Edit for personality: shorten, add emotion, use simple words and active verbs.
5. Add a clear CTA: save, reply, try, or click — make it obvious.
6. Schedule & test: post at a good time and track responses for improvement.

What to expect

• First drafts are fast but need human edits — AI gives structure, not soul.
• Early wins come from strong hooks and one practical takeaway per thread.
• Improve by testing variations: hook-first, story-first, or stat-first.

Example thread (6 tweets)

• Hook: Most people post threads that no one reads — here’s a better way.
• Problem: You write long, unfocused threads that lose attention after tweet two.
• Why it happens: No clear promise, too many ideas, no rhythm.
• Quick fix: Pick one promise, give 3 practical steps, add a mini example.
• Mini example: “Try this today: write your hook, then list 3 actionable tips.”
• CTA: If you want a ready-to-post thread on this topic, reply with your niche.

Mistakes & fixes

• Mistake: Long, vague hooks. Fix: Make the benefit specific and urgent.
• Mistake: Too many ideas. Fix: Stick to one promise and one CTA.
• Mistake: No edit pass. Fix: Read aloud and trim 30% of words.

Copy-paste AI prompt (use as-is)

Write a Twitter thread of 6 tweets for an audience of small business owners. Start with a strong, curiosity-driven hook. Each tweet should be short, clear and actionable. Include a practical example and finish with a single, specific CTA asking readers to reply if they want a free template. Tone: friendly, confident, simple.

Variant prompts

• Change audience to “freelancers over 40” and CTA to “save this thread.”
• Ask for 8 tweets and include one statistic in tweet 3.

Action plan (next 30 minutes)

1. Pick one topic and write a one-line promise.
2. Use the copy-paste prompt above to generate a draft.
3. Edit for brevity and personality, add CTA, then post.

Pragmatic optimism: let the AI do the heavy lifting on structure — you bring the voice and the final edit. Try one thread today and iterate from the response.

Great question. You’re asking if AI can help you create packaging designs with proper dielines. Smart move—this is one of the fastest, most practical uses of AI for non‑designers.

Here’s the simple truth: AI can get you from idea to a printer-ready carton faster, as long as you stick to standard box styles and double-check with your printer. You don’t need to be technical—just follow the steps.

What you’ll need

• An AI assistant (for text + code) to generate SVG dielines
• A vector editor: Adobe Illustrator (best) or Inkscape (free)
• An image model for concept art (Midjourney, DALL·E, Firefly—any is fine)
• Basic measurements of your product (width, depth, height)
• Your printer’s specs: bleed (usually 3 mm or 1/8”), color naming for cut/crease lines, file format (often PDF/X-1a or X-4)

How it works (end-to-end)

1. Pick a box style (quick wins): Reverse Tuck End (RTE) carton, Straight Tuck End (STE), or Crash-Lock (auto bottom). If unsure, choose RTE for most small products.
2. Measure the product in millimeters: Width (W), Depth (D), Height (H). Add a few mm clearance so it slides in easily (typically +2–3 mm on W and D, +3–5 mm on H).
3. Brainstorm the look with an image model. Generate style boards before you design. See prompt below.
4. Generate the dieline (SVG) with your AI assistant. Paste the SVG into Illustrator or Inkscape. Scale check is critical—1 mm must equal 1 mm.
5. Add brand artwork on a separate “Artwork” layer. Keep text vector, 100% K black for small black text, and images at 300 dpi effective resolution.
6. Set production lines: Spot color named exactly what your printer wants (common: “CutContour” for cut; “Crease” for fold). Cut = magenta 100% stroke; Crease = cyan 100% stroke. No fill. Set these to overprint stroke.
7. Bleed and safe areas: Add 3 mm bleed outside the cut line; keep essential elements 3–5 mm inside the cut line (safe area).
8. Soft-proof: Print to scale on your office printer, cut and fold the mockup. Check fit, flap clearances, barcode visibility, and glue area overlap.
9. Export as PDF/X-1a or X-4 with layers intact. Include crop marks if requested. Send to your printer for a preflight check before full production.

Copy‑paste prompts you can use today

• Concept art / style board (use your image model of choice):“Create a clean packaging style board for a [Reverse Tuck End box] for [product: e.g., herbal tea sachets], target audience [health-conscious adults 40+], design vibe [calm, premium, minimal], dominant colors [sage green, warm white], include front panel hero, side panel info, and back panel story. Show 3 variations with different typography and pattern treatments.”
• SVG dieline generator (use your AI assistant for code):“You are a packaging engineer. Generate an dieline for a Reverse Tuck End carton sized W=70 mm, D=30 mm, H=120 mm. Requirements: 3 mm bleed; 5 mm safe area; glue flap 18 mm; bottom and top tuck flaps with dust flaps; include hanger tab optional (separate group I can delete). Create separate groups/layers: ‘CUT’, ‘CREASE’, ‘BLEED’, ‘SAFE’, ‘ARTWORK_GUIDE’. Set spot colors: CUT = 100% magenta named ‘CutContour’; CREASE = 100% cyan named ‘Crease’. Strokes 0.25 pt, no fills on production lines. Include small registration marks outside the bleed. Units in millimeters. Return valid, well-scaled SVG (1 SVG unit = 1 mm). Do not include explanations—SVG only.”

Example outcome to expect

• An SVG file that opens in Illustrator at correct real-world size with clearly labeled groups and colored cut/crease lines.
• A flat layout showing panels: front, back, sides, top and bottom flaps, dust flaps, and a glue flap. Bleed and safe guides visible.
• After placing artwork, a printer‑ready PDF with layers: Artwork, Cut, Crease, Bleed/Guides.

Insider tips that save reprints

• Name spot colors exactly as your printer specifies. Some finishing tables trigger on “CutContour.” Get it right.
• Overprint the cut/crease strokes so they don’t knock out your artwork.
• Barcodes: 80–200% magnification, quiet zone clear, black on light background, vector if possible.
• Glue area: Keep ink/tape-free within the glue flap unless your printer says otherwise.
• Bleed: 3 mm (EU) / 0.125 in (US) standard. Patterns and backgrounds must extend into bleed.
• Color: Work in CMYK. Keep small text single-color (100K) to avoid registration fuzz.

Common mistakes and quick fixes

• Wrong scale: If the SVG imports at the wrong size, check the viewBox and unit assumptions. Ask AI to regenerate with “1 SVG unit = 1 mm.”
• Missing layers: If everything’s one layer, regroup in Illustrator and rename layers to your printer’s spec.
• Flaps colliding: Build a paper mockup. If dust flaps fight the product, increase depth by 1–2 mm or shorten dust flaps.
• Artwork on fold: Move key text/logos 3–5 mm away from crease lines.
• Low-res images: Check Effective PPI in your vector editor. Aim for 300 PPI at placed size.
• RGB export: Export as CMYK PDF/X and embed/outline fonts.

Two quick paths

• Fast DIY: RTE box, AI SVG dieline, simple brand color + logo, one hero image, back panel details. One evening.
• Polished: AI concepts, refine typography, subtle pattern, spot UV or foil mockup, 3D preview (use a 3D packaging mockup tool), printer preflight, then run.

Action plan for this week

1. Pick the box style and capture W/D/H.
2. Use the concept prompt to pick a visual direction.
3. Generate the SVG dieline with the code prompt; open in Illustrator and verify size.
4. Add bleed/safe, set Cut/Crease spot colors, place artwork.
5. Print a paper mockup; adjust if needed.
6. Export PDF/X and send to your printer for a quick preflight before you order.

Reality check: AI handles standard cartons very well. Complex shapes, curved cuts, and multi-part displays may still need a packaging CAD tool or your printer’s CAD team. Start simple, iterate fast, and involve your printer early—they’re your best partner.

Quick win: Try this now — change “The report was written by Maria.” to active voice: “Maria wrote the report.” You’ve just done it in under 30 seconds.

Great point in your thread title — focusing on simple examples and beginner tips is exactly the right approach. Below is a practical, step-by-step way to spot passive voice and turn it into strong, active sentences using plain rules and an AI prompt you can copy-paste.

What you’ll need

• A short sentence or paragraph you want to improve.
• An editor (Word, Google Docs) or an AI chat box where you can paste text and a prompt.
• Five minutes.

Step-by-step: How to convert passive to active

1. Find the passive clue: look for a form of “to be” (is, was, were, been) + a past participle (written, made, called) and often a “by” phrase.
2. Ask: who performed the action? That person or thing becomes the subject.
3. Move that subject to the front, use a simple past or present verb, and keep the object after the verb.
4. Read the active sentence aloud — it should feel clearer and more direct.

Examples

• Passive: “The meeting was scheduled by the team.” → Active: “The team scheduled the meeting.”
• Passive: “A decision was made.” → Active options: “We decided.” or “Management decided.” (Choose the correct agent.)

Common mistakes & fixes

• Mistake: Dropping the agent and changing meaning. Fix: If the agent matters, keep it (e.g., “by the board”).
• Over-correcting to awkward subject choices. Fix: Use natural subjects like “we,” a department, or a named person.
• Losing formality or tone. Fix: Keep formal wording where needed (e.g., “The committee approved the policy.”)

Try this AI prompt — copy and paste exactly as written

Turn the following passive sentences into active voice. For each sentence, give the active version and one brief note on what changed (who becomes the subject, and how the verb changed):

1) The report was written by Maria.
2) The meeting was scheduled by the team.
3) A decision was made.

If the agent is unclear, suggest two possible active versions (one formal, one casual).

Action plan (5 minutes)

1. Pick three passive sentences from your writing.
2. Use the prompt above in an AI chat or do the steps manually.
3. Choose the active option that best keeps your meaning and tone.

Reminder: Active voice usually reads faster and feels stronger. Use it where clarity and action matter — but keep passive when you need to emphasize the result or preserve neutrality.

Hook: Want faster wins from your creatives? Use AI to generate and test variations quickly so you learn what works — not what you hope works.

Why this matters

If you’re over 40 and not a techie, think of AI as a creative assistant that drafts many believable options in minutes. Pair that with small, rapid A/B tests and you’ll find winners fast, save ad spend, and scale what works.

What you’ll need

• Baseline creative (current ad, email, landing page)
• Simple A/B testing tool or ad platform split-testing (Facebook, Google, email provider)
• Spreadsheet to track results
• AI copy tool (Chat-style or headline generator) and an image generator or template tool
• Clear metrics: CTR, conversion rate, cost per conversion

Step-by-step: a 7-day rapid A/B sprint

1. Day 1 — Define your test: pick 1 goal (e.g., increase CTR) and 1-2 variables: headline and image.
2. Day 2 — Create variations: use AI to produce 3 headlines and 3 image concepts. Keep length and tone rules simple.
3. Day 3 — Build 4–6 ad variants (combine headlines and images). Keep sample sizes achievable (small budgets but equal splits).
4. Days 4–6 — Run test: let each variant get meaningful impressions (aim for at least a few hundred clicks or 1,000+ impressions depending on platform).
5. Day 7 — Analyze and pick the winner by comparing CTR and conversion. Pause losers; scale winners and repeat with a new variable.

Example

Goal: raise landing page CTR. Variables: headline and hero image. AI generates 3 headlines and 2 image styles. Create 4 ad combos, run for 5 days with equal budget. Variant B shows 30% higher CTR — use that headline across channels and test a new image.

Common mistakes & fixes

• Mistake: Testing too many variables at once. Fix: Test 1–2 variables per sprint.
• Mistake: Stopping too early. Fix: Run until you have meaningful clicks/impressions, not just 24 hours.
• Mistake: Changing audience mid-test. Fix: Keep audience constant.
• Mistake: Ignoring creative fatigue. Fix: Rotate or refresh creatives after 1–2 weeks.

Copy-paste AI prompt (use in Chat or an AI copy tool)

“Write 6 short headlines (6–10 words each) for a paid ad promoting a simple online course on improving LinkedIn profiles for mid-career professionals. Tone: confident, friendly, action-oriented. Include 2 with a question, 2 with numbers, and 2 with a direct benefit + CTA. Keep language simple and avoid jargon.”

Quick action plan (this afternoon)

1. Pick your baseline creative and metric.
2. Run the AI prompt above and pick 3 headlines.
3. Create 2 image options (simple photo vs. illustrated). Combine into 4 ads.
4. Start a 5-day split test with equal budget and track results in a sheet.

Closing reminder

Start small, measure, then scale. AI speeds creativity — your judgment steers it. Try one sprint this week and you’ll learn more than weeks of guesswork.

Hook: Want faster, higher-response renewal and expansion emails that still feel human? Use AI as your drafting partner — then personalize and send.

Quick correction: AI won’t magically know your customer. It’s a drafting tool — you must add specifics (metrics, timelines, tone) before sending. Treat AI output as a first draft to speed you up, not the final legal or relationship copy.

Why this works: AI saves time on structure, subject lines, and multiple variations. You keep the relationship context and final sign-off. That combo wins renewals and upsells.

What you’ll need:

• Customer context: contract end date, usage, recent wins, pain points.
• Clear goal: renew, expand seats, propose add-on, or set a meeting.
• AI tool or chat window (any GPT-style assistant).

Step-by-step (do this now):

1. Gather 3 facts: renewal date, one success metric, one risk or objection.
2. Choose tone: professional, friendly, urgent, or consultative.
3. Use the prompt below (copy-paste) to generate 3 email variants: short, detailed, and meeting request.
4. Quick-edit each variant: add personal line referencing a recent call or result.
5. Pick subject line from the AI suggestions, A/B test 2 versions to a small segment.
6. Send follow-up reminders at +3 days and +7 days with concise CTAs.

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

“Write three renewal email variants for a SaaS customer whose contract expires in 30 days. Facts: customer ‘Acme Co’, renewal date 30 days from now, current plan: Professional (10 seats), usage: 85% seat utilization, success: reduced onboarding time by 20%, concern: budget review. Goals: 1) encourage renewal on same plan, 2) propose expansion to 15 seats, 3) ask for a 20-minute meeting. Provide subject lines, short preview text, and a one-sentence personalized opening. Keep tone consultative and friendly.”

Example output (short variant):

Subject: Quick renewal + idea for expanding Acme’s seats
Preview: 30 days until renewal — one idea to save time and support growth
Hi [Name], congratulations on cutting onboarding time by 20%. With usage at 85% of 10 seats, you may hit capacity soon. Can we lock in your renewal and discuss expanding to 15 seats to avoid interruptions? Are you available for 20 minutes next week?

Mistakes to avoid & fixes:

• Mass send without personalization — Fix: add one specific success metric or call note per email.
• Being vague on next step — Fix: offer a specific date/time or a calendar link (or ask them to suggest one).
• Overloading with features — Fix: focus on the one benefit tied to their business goal.

Action plan (next 30 minutes):

1. Pull the 3 facts for one priority customer.
2. Run the copy-paste prompt above in your AI tool.
3. Edit the best draft for personalization and send an A/B test to a small group.

Closing reminder: Use AI to draft, your knowledge to close. Small personalization beats perfect automation every time.

Try this now (under 5 minutes)

Copy, paste, and fill the brackets to get a tight founder intro for your About page:

Prompt: “You are a brand copywriter. Write a 120–150 word founder intro in first person that feels warm, confident, and human. Audience: [target audience]. Industry: [industry]. Tone: [choose 3 words: e.g., honest, energetic, practical]. Include: my turning point ([one moment]), 1 proof point ([metric or media mention]), and a simple invite to act ([book a call / join newsletter]). Keep sentences short, avoid clichés, no jargon, and make the first line a hook that shows the problem I solve.”

Why this matters

Your founder story and About page do three jobs: build trust, clarify what you do, and invite the next step. Done right, they turn browsers into believers without sounding salesy.

What you’ll need (10–15 minutes to gather)

• 3 facts: who you help, what you sell, the result you create.
• 1 turning point: the moment that made you start.
• 1 proof point: number, milestone, or client quote.
• Your voice in 3 words (e.g., grounded, candid, optimistic).
• A simple call to action (what you want readers to do next).

The structure that works (and why)

• Origin: the itch you had to scratch.
• Obstacle: the messy middle or hard lesson.
• Outcome: what you learned and the result you deliver.
• Mission: the simple promise today.
• Invite: the next step for the reader.

Step-by-step: draft your founder story

1. Gather inputs (use the list above).
2. Generate your draft using this prompt:

Prompt: “Act as a senior brand storyteller. Create a 450–600 word founder story in first person using the Origin–Obstacle–Outcome–Mission–Invite structure. Audience: [describe]. Industry: [industry]. Voice: [3 words]. Include 1 short anecdote with sensory detail, 1 proof point ([number or milestone]), and 1 line that shows vulnerability without oversharing. Keep paragraphs short. Avoid buzzwords. End with a clear invite to [CTA].”

1. Tighten: Ask the AI to cut 15% fluff and make the first sentence a hook.
2. Add rhythm: Request 10% sentence variety (mix of short and medium lines).
3. Sanity check: Read it aloud. If it doesn’t sound like you, adjust the tone words.

Step-by-step: build the About page (sections)

1. Hero: One-liner that names the problem and promise.
2. Credibility: 3–5 proofs (numbers, logos, milestones, awards).
3. Founder story: The polished version from above.
4. Values in action: 3 bullets showing how you behave, not slogans.
5. Approach: How you work in 3–4 steps.
6. Social proof: 2–3 short testimonials with outcomes.
7. Call to action: One clear next step.

Prompt: “You are a conversion copywriter. Draft an About page that includes: Hero one-liner, Credibility bullets, Founder story (summary, not full), Values in action (3 bullets), Our approach (4 steps), Social proof (3 short quotes), and a single Call to action. Audience: [describe]. Offer: [what you sell]. Voice: [3 words]. Keep it scannable with short paragraphs and bullet points. No jargon, no clichés, and avoid repeating the same claim twice.”

Premium insider tricks

• Two-Voice Pass: First, ask for a “journal-style” draft to capture honesty. Second, ask for a “website-ready” pass for clarity and flow. Merge the best lines.
• Proof Power-Up: Add one number (clients served, years, success rate) and one named method (“The 3-Step Reset”) to anchor authority.
• Voice DNA: Feed the AI 2–3 small writing samples of yours, then say: “Mirror this tone.” It reduces generic phrases dramatically.

Example (what a finished founder story can feel like)

“Ten years ago, I almost closed my first business. Great product, no clear message. I spent nights rewriting the site while my coffee went cold and my confidence followed. The breakthrough wasn’t a hack—it was telling the truth: who I help, why it matters, and what happens next.

Today I help service founders turn long, tangled bios into simple stories that win trust. I’ve worked with 200+ small businesses and watched conversions rise when the copy finally sounded like a person, not a brochure.

Here’s what I believe: your story isn’t about you—it’s about the reader seeing themselves in your journey. If you want clear words that feel like a handshake, not a pitch, let’s start there.”

Common mistakes (and quick fixes)

• Too much timeline: Readers don’t need every year. Fix: keep one turning point and one result.
• Vague mission: “Empower” means nothing. Fix: say the outcome in plain words.
• No proof: Claims without evidence erode trust. Fix: add one number or specific client outcome.
• Weak invite: “Learn more” is mushy. Fix: offer one clear action with what happens next.
• Corporate tone: Jargon kills warmth. Fix: read aloud and swap buzzwords for everyday language.

Polish prompts (copy-paste)

• “Rewrite this in first person with shorter sentences. Keep my voice [3 words]. Remove buzzwords. Keep the key facts. Make the first line a hook.”
• “Cut 20% without losing meaning. Replace any vague word with a concrete one. Highlight one sentence I can use as a tagline.”
• “Turn this into an About page section list: Hero, Credibility, Founder story (100 words), Values (3 bullets), Approach (4 steps), Social proof (3 quotes), CTA. Keep it skimmable.”

What to expect

• Draft 1: honest but messy. Good.
• Draft 2: tighter with proof.
• Draft 3: website-ready. You’ll have a clean story and an About page outline you can publish.

30-minute action plan

1. Collect your inputs (facts, turning point, proof, voice, CTA).
2. Run the 120–150 word intro prompt and paste it at the top of your About page.
3. Run the full founder story prompt; trim to fit.
4. Use the About page prompt to build the remaining sections.
5. Do the Two-Voice Pass and one read-aloud edit.
6. Publish a v1. Improve later with real testimonials.

Final nudge

Your story doesn’t need to be epic. It needs to be clear, specific, and human. Start with the quick intro, stack in a proof point, and invite the next step. Momentum beats perfection.

Nice, simple question — exactly the kind that gets fast wins. The idea that AI can both analyse cohort retention and suggest lifecycle nudges is practical and ready to use.

Why this matters: cohort analysis shows when people stop coming back. AI helps turn those patterns into specific, timed nudges you can test quickly — without complex math or a data scientist on every call.

What you’ll need

• Basic cohort data: user_id, signup_date, event_date (or week/month number), and an engagement flag (1/0).
• A tool to compute cohorts: spreadsheet or a simple SQL query. No-code analytics (Mixpanel/Amplitude) or Google Sheets work fine.
• An AI assistant (ChatGPT-like) for idea generation and message drafting.
• A/B test capability in your email/CRM or in-app messaging system.

Step-by-step: from data to nudges

1. Collect 6–12 weeks of user-event data. Clean obvious duplicates or bots.
2. Define cohorts (by week or month of signup) and calculate retention per period (percent active).
3. Spot the biggest drop-offs — e.g., week 1→2 or month 1→2.
4. Feed the retention snapshot into the AI with context (product type, main value prop, channels available).
5. Ask the AI for 3 concrete nudges per cohort: timing, channel, short message, and one metric to test.
6. Prototype the top nudge, run an A/B test for 2–4 weeks, measure lift on the retention window.

Copy-paste AI prompt (use as-is)

“You are a product growth analyst. I have cohort retention data in CSV format (columns: cohort_week, week_number, retention_rate). Here is a small sample:ncohort_week,week_number,retention_raten2025-09-01,1,0.60n2025-09-01,2,0.35n2025-09-08,1,0.58n2025-09-08,2,0.32nProduct: an online course platform. Main value: fast, practical lessons. Channels: email, in-app, push. Provide 3 actionable lifecycle nudges targeted at the week 1→2 drop, each with timing, channel, message (<=140 chars), a success metric, and a simple A/B test design.”

Worked example

• Data shows week1→2 drops from ~60% to ~33%. AI suggests: Day 3 onboarding tip (email), Day 7 micro-challenge (in-app), Day 10 social proof + offer (push/email).
• Example message: “Start Lesson 2 — 7 minutes to a skill you can use today.” Metric: % returning in week 2. Test: 50/50 sample, measure lift after 14 days.

Common mistakes & fixes

• Don’t blast everyone. Fix: segment by behavior or intent.
• Don’t trust noisy short windows. Fix: use 4–8 weeks of data and smooth spikes.
• Don’t confuse correlation with cause. Fix: validate with A/B tests.

2-week action plan

1. Export cohort data and compute retention table.
2. Run the AI prompt above and pick 2 nudges.
3. Build messages and set up A/B tests for one cohort.
4. Run tests 2–4 weeks, review results, iterate.

Small experiments give fast learning. Start with one cohort, one nudge, one clear metric — then scale what works.