Jeff Bullas

Short answer: Yes — AI can simulate funnel changes and give probability-based forecasts for A/B tests, but it’s only as good as your data and assumptions.

Here’s a clear, practical way to use AI to simulate and forecast A/B outcomes so you can make faster, smarter decisions.

What you’ll need

• Historical funnel data (traffic, step conversion rates, drop-offs by step).
• Baseline metrics (current conversion rate, sample size, variance).
• Clearly defined variants and expected changes (e.g., increase checkout conversion by 10%).
• A tool: spreadsheet + Monte Carlo add-on, or a simple Python/R notebook, or an AI platform that runs simulations.

Step-by-step (practical)

1. Map your funnel: traffic → leads → trials → purchases. Collect counts and conversion rates for each step.
2. Define the change: which step the variant affects and by how much (point estimate or distribution).
3. Choose a model: use binomial draws per step (Monte Carlo) or a Bayesian model for posterior uplift probability.
4. Run simulations: iterate 10,000 times drawing conversions for control and variant, propagate through funnel to final revenue metric.
5. Summarize results: median uplift, 95% credible/confidence interval, probability variant > control, expected revenue impact.
6. Decide rules: predefine a threshold (e.g., >80% probability of positive uplift OR >$X monthly revenue) for rollout.

Simple numerical example

• Baseline traffic: 10,000 visits, purchase rate 2% → 200 purchases.
• Variant aims +10% relative uplift → expected 2.2% → 220 purchases.
• Simulation (10,000 draws) gives a distribution — you’ll get a range, e.g., median uplift 10% with a 95% interval of -1% to +21% and a 78% chance the variant wins.
• That tells you the test is promising but not decisive; you either increase sample size or accept a measured rollout.

Common mistakes & fixes

• Ignoring seasonality — fix: use time-matched historical windows or include time trends in the model.
• Too-small samples — fix: compute required sample size or run longer tests.
• Multiple comparisons — fix: adjust thresholds or use Bayesian hierarchical models.
• Assuming perfect data — fix: audit tracking before simulating.

Copy-paste AI prompt (use as a start)

“I have historical funnel data: 10,000 weekly visits, homepage→signup 5% (500), signup→trial 20% (100), trial→paid 40% (40). I plan Variant A affecting signup→trial by +10% relative. Using a Monte Carlo simulation of 10,000 iterations, simulate control vs variant outcomes, propagate conversions to paid customers, and return: median uplift in paid customers, 95% interval, probability variant > control, and estimated monthly revenue impact if average order value is $100. List assumptions and recommend required sample size for 80% power.”

Action plan — quick wins

• Run a baseline simulation with current data today.
• If uplift probability >75%, consider a staged rollout; if 50–75%, increase sample or refine variant.
• Pre-register test rules and track the chosen metric.

Remember: AI helps quantify uncertainty and accelerate decisions, but it can’t replace clean data and clear business rules. Simulate fast, test faster, learn continuously.

Spot on: your adaptive branching idea is the missing lever. A simple yes/no or 0–2 score keeps the plan honest and the client moving. Let’s turn that into a small, repeatable system you can run from a spreadsheet in under an hour.

Big idea: Branching-in-a-Box — three tiny assets (rubric, rules, and tagged content) that let AI generate next-week plans that fit the client’s pace without heavy tech.

What you need

• One-page rubric (3 criteria, 0–2 score each) and a pass threshold.
• Tag your micro-lessons by skill, difficulty (1–3), and time (in minutes).
• Simple tracker with columns: Week, Objective, Evidence link, Score (0–2), Rule fired, Next steps, Coach notes, Due date.
• AI chat tool and two prompts: score evidence, generate next-week plan.
• Low-friction evidence capture (phone video, audio, screenshot).

How to set it up — step-by-step

1. Write a mini-rubric (5 minutes). Choose 3 criteria that define visible progress. Example for public speaking: Clarity, Structure, Presence. Scoring: 0 = not yet, 1 = emerging, 2 = solid. Pass if total ≥ 4/6.
2. Tag your content (15 minutes). For each micro-lesson, add Skill tag (e.g., Structure), Difficulty (1–3), Time (e.g., 8 min), and Exercise link or description.
3. Define three branching rules (5 minutes).
   • Pass (≥ threshold): advance to next module; add one stretch exercise.
   • Borderline (just under): repeat focused drill + one new angle; mandatory coach review.
   • Fail (well under): short diagnostic + narrow drill; reduce scope; book 15-min coach nudge.
4. Drop the structure into your tracker (5 minutes). One row per week. Add conditional formatting: green = pass, amber = borderline, red = fail.
5. Plug AI into two moments.
   • End of week: use AI to score the evidence against the rubric (coach reviews in 2 minutes).
   • Start of next week: use AI to propose next steps based on the rule that fired and the tagged library.

Copy-paste AI prompts (use as-is)

Evidence Scoring Prompt

“You are a coach’s assistant. Score this client’s weekly evidence using ONLY this rubric and scale. Rubric criteria (0–2 each): 1) Clarity (is the message easy to follow?), 2) Structure (clear opening, middle, closing), 3) Presence (voice, pace, posture). Scoring scale: 0 = not yet, 1 = emerging, 2 = solid. Output JSON with keys: total_score (0–6), criterion_scores, 2 bullet strengths, 2 bullet improvements, pass (true/false, pass if total ≥ 4), and a one-sentence risk note. Evidence: [paste transcript or summary of what you observed].”

Next-Week Generator Prompt

“Based on this score and our branching rules, create the next week’s plan. Inputs: goal = [e.g., board-level speaking], time_available = [e.g., 3 hrs], score = [0–6], pass_threshold = 4, last_week_objective = [text]. Library (each item: skill, difficulty 1–3, time minutes, exercise description). Rules: pass → move to next skill + 1 stretch; borderline → repeat focused drill + 1 new angle; fail → diagnostic + narrow drill + coach check-in. Output: objective, 2 exercises with time, 1 stretch or drill as per rule, total time, success criteria (1 sentence), and a short coach note (risk + remediation). Keep it concise.”

Worked example (public speaking, Week 2)

• Evidence result: total 3/6 (clarity 1, structure 1, presence 1) → fail.
• AI next-week plan (fail rule): Objective: nail a 30-second opening. Exercises: (1) Script a 3-sentence opening; record 3 takes (25 min). (2) Pacing drill with metronome count-in; rerecord (20 min). Diagnostic: 2-minute self-score using rubric (10 min). Total time: 55 min + 15-min coach nudge. Success: a clean 30-second opening delivered within 35–40 seconds, no filler words in the first sentence. Coach note: risk = over-scripting; fix = rehearse, then speak from bullets.
• If borderline (4/6): Repeat opening once with story angle; add stretch: 60-second board hook using a metric.
• If pass (5–6/6): Advance to structure: craft a 2-minute message using Problem–Evidence–Action; stretch: add a credibility line.

Insider tip: freeze your rubric for 4 weeks before you tweak it. Consistency beats perfection. Adjust after you’ve seen three client cycles.

What to expect

• Setup: 45–60 minutes for rubric, tags, and prompts.
• Weekly admin: 10–15 minutes to score evidence, 5 minutes to approve AI’s next week.
• Outcomes: tighter focus, fewer stalled weeks, clearer proof of progress clients can feel and you can sell.

Common mistakes and quick fixes

• Vague rubric. Fix: use observable behaviors and a 0–2 scale; define a pass number.
• Too many branches. Fix: keep three rules; complexity kills momentum.
• Ignoring time reality. Fix: cap weekly plan to the client’s hours; trim before you add.
• Unreviewed AI scoring. Fix: coach glances at the JSON and adjusts; trust but verify.
• Content bloat. Fix: archive anything not used in two cycles; keep the library lean.

5-day mini-sprint

1. Day 1: Draft the 3-criteria rubric and pass threshold.
2. Day 2: Tag 12 micro-lessons (skill, difficulty, time).
3. Day 3: Build the tracker; paste the two prompts into your AI tool.
4. Day 4: Run one client through Week 1; collect evidence; score with AI; approve.
5. Day 5: Generate Week 2 via the branching rule; schedule a 15-minute check-in.

Why this works

• It blends human judgment with AI speed.
• It keeps personalization simple, visible, and measurable.
• It creates artifacts (scores, evidence, notes) you can show in progress reviews and proposals.

Bottom line: start with the three-rule branch, one rubric, and a lean library. Keep weeks short, criteria clear, and decisions binary. You’ll get faster wins, fewer stalls, and a coaching practice that scales without losing the human touch.

Love the “who it’s for” microline and your 3×3 diagnostic. That’s the right foundation. Let’s add one more lever most teams miss: use AI to rewrite your pricing page in your customers’ words (not yours) and match that message to each traffic source — without changing prices first.

Why this works: People buy when they see their problem, their language, and a safe next step. AI can mine that language from your reviews, support tickets, and call notes in minutes. Then you test the message and the anchor before you touch price. Cleaner signals, faster wins.

What you’ll need

• Export of 100–300 recent support tickets, reviews, or call notes (CSV or copy/paste is fine).
• Your current pricing page copy (desktop + mobile screenshots help).
• Testing tool or CMS split test; analytics with revenue events.
• Basic traffic source labels (paid search, social, organic, email).

Step-by-step: Voice-of-Customer (VoC) pricing makeover

1. Mine the customer language with AICopy 100–300 lines of reviews/tickets into AI and run this:

Copy-paste prompt:“You are a conversion strategist. Analyze this customer text: [PASTE 100–300 LINES OF REVIEWS/TICKETS/CALL NOTES]. Extract: 1) top 10 outcomes customers want in their words, 2) top 10 anxieties or objections about price/commitment, 3) phrases that signal urgency or value, 4) common decision triggers (trial length, guarantees, team size). Turn this into: a) three 10-word outcome bullets per plan, b) a 6-word ‘who it’s for’ line per plan, c) five concise objection–answer pairs for an inline FAQ near the CTA. Keep copy plain, specific, and non-hype.”

1. Upgrade your pricing page blocks (no price change yet)
   • Add the VoC outcome bullets and the “who it’s for” line under each plan.
   • Place a 3–5 question micro-FAQ under the primary CTA using the objection–answer pairs (e.g., “What happens after my trial?” “You keep your data; we don’t auto-charge.” Only if true).
   • Keep your “Most popular” badge and add a credible anchor (premium reference tier or annual savings phrasing) as you outlined.
2. Match message to traffic source (simple, high-ROI)
   • Create 3 headline/subhead sets from the VoC output: one for paid search (problem-first), one for organic (benefit-first), one for email/returning (time-to-value).
   • Use your testing tool to swap just the headline/subhead by UTM/source. Prices stay constant. KPI is RPV; secondaries are plan mix and annual attach.
3. Design a clean decoy anchor (ethical)
   • Add a higher-reference tier with 1–2 unmistakable differentiators (e.g., SSO, priority support). Mark mid-tier “Most popular.”
   • Show annual savings as a specific number (“Save $168/year”), not a vague percentage.
4. Run the test
   • Primary KPI: revenue per visitor (RPV). Guardrails: refund rate, support tickets from pricing/checkout.
   • Duration: until your calculator shows ~80% power or 2–4 weeks. Low traffic: rotate variants by day-of-week (switchback) for two cycles and compare average daily RPV.
5. Let AI analyze results and plan the follow-up

Copy-paste prompt:“Act as a test analyst. Here are results by variant and source/device: [PASTE TABLE]. KPIs: RPV (primary), conversion, AOV/ARPU, plan mix, annual attach, refund rate. Tell me: 1) overall and segment winners, 2) which element likely drove lift (headline match, anchor, FAQ), 3) any guardrail issues, 4) two follow-up tests to isolate the driver, 5) rollout plan (global vs segment) with risks.”

Concrete example (use and adapt)

• Who it’s for (under plan names): “For solo pros,” “For 3–20 person teams,” “For regulated workflows.”
• Outcome bullets (mid-tier): “Consolidate 5 tools into one,” “Onboard a teammate in 10 minutes,” “Monthly reporting done in 1 click.”
• Micro-FAQ near CTA: “Do I need a credit card?” “No. Start free, upgrade anytime.” “Can I cancel?” “Yes, anytime. No fees.” (Only if accurate.)
• Anchor: Add “Advanced — from $149 — SSO, audit logs, priority support.” Mid-tier labeled “Most popular.” Annual toggle shows “Save $168/year.”

Insider trick: message-first price elasticity check

• After a messaging win (RPV +5% with constant price), run a short follow-up with only the mid-tier +10% for new traffic. If conversion holds within ~3% and RPV increases, you’ve got headroom. If not, keep the messaging win and revert price.

Common mistakes and fast fixes

• Too many moving parts: Lock everything except the variables you’re testing. Document the change and dates.
• Mobile truncation: On small screens, plan cards often hide your best bullet. Manually check and trim to 8–10 words per bullet.
• Weak anchors: Premium tier must have obvious, high-value differences. If visitors can’t spot them in 3 seconds, the anchor won’t work.
• Counting clicks, not dollars: Judge by RPV, plan mix, and annual attach. Button CTR can mislead.

7-day action plan

1. Day 1: Export reviews/tickets. Capture baselines (RPV, conversion, AOV/ARPU, plan mix, annual attach, refund rate).
2. Day 2: Run the VoC prompt. Produce outcome bullets, “who it’s for,” and micro-FAQ.
3. Day 3: Build two variants: A) Anchor + VoC copy, B) Annual savings + micro-FAQ. Keep price points constant.
4. Day 4: Set up message-by-source headlines (paid, organic, returning). QA events and revenue tracking.
5. Day 5: Launch. Freeze other page changes.
6. Day 6: Health check only. Confirm even traffic split and clean data.
7. Day 7: Use the analysis prompt with early segment data. If guardrails hold, continue to full duration; if broken, pause and fix.

What to expect: Faster clarity with fewer variants. Most teams see early lifts from message match and credible anchoring before any price moves. Segment wins usually appear first (mobile or paid search). Roll out to those segments, then expand.

Keep it simple, keep it honest, and let the numbers — not opinions — make the call.

Onwards,Jeff

Two-minute upgrade: Add a 60-word cap. When you paste your three bullets into the AI, say “keep it under 60 words, one clear action.” That single guardrail keeps the draft tight and human.

Why this works: People scan. A short subject and a 2–3 sentence body get read in under 10 seconds. The AI does the phrasing; you keep the context and the decision.

What you’ll need

• Any AI writing assistant (in your email or browser).
• Your three bullets: purpose, one fact/date, one action.
• One minute to read aloud and tweak one phrase.

Step-by-step (tighten your draft)

1. Write the three bullets (one line each).
2. Paste into your AI with the 60-word cap and “plain English.”
3. Ask for a 3–5 word subject and one direct CTA.
4. Read it out loud once. Swap any stiff phrase for your words.
5. Send. If no reply in 48 hours, nudge with one sentence.

Copy-paste prompt (general)

“I have three bullets: 1) [purpose], 2) [one fact/date], 3) [single action I want]. Write a concise email with a 3–5 word subject and a 2–3 sentence body. Cap at 60 words. Use plain English and sound like a helpful colleague. Include one clear CTA (reply/confirm/click) and nothing extra.”

Insider trick: build a quick voice snapshot (90 seconds)

1. Copy two emails you’ve sent that sound like you.
2. Tell the AI: “Learn my tone from these two emails: [paste]. When you write for me, match this tone: friendly, direct, no buzzwords, short sentences.”
3. Then run the general prompt. Expect drafts that feel more “you” immediately.

Mini template library (use as-is)

• Request (scheduling): “Subject: Quick time this week. Body: Hi [Name], can we pick a 15‑minute slot? I’m free Tue 2–4 or Wed 9–11. Please reply with the time that suits you.”
• Update (no action): “Subject: Short update. Body: Quick heads-up: [one-line update]. No action needed today; I’ll share the next step on [date].”
• Decision (one choice): “Subject: Need a call? Body: We can solve [issue] by A (fast, low cost) or B (thorough, higher cost). Please reply A or B by [date].”
• Follow-up (polite nudge): “Subject: Gentle reminder. Body: Checking in on [topic]. Could you reply yes/no by [date]? Happy to adjust if timing is tight.”

Copy-paste prompt (subject line supercharger)

“Based on these bullets: [paste], generate 5 subject lines, 3–5 words each, that are clear and non-salesy. Examples to match: ‘Quick check-in,’ ‘Short update,’ ‘Scheduling quick call,’ ‘Decision by Wed?’ Stick to plain English.”

Copy-paste prompt (one-sentence follow-up)

“Write a single-sentence follow-up for this email: [paste original email]. Tone: polite, direct. Ask for a simple yes/no or a date. Keep to 18–25 words.”

Concrete examples (before → after)

• Before: “Circling back on the Q3 numbers and potential resourcing options given timelines.”After: “Can we review Q3 numbers for 15 minutes? I’m free Wed 10–12. Please reply with a time that works.”
• Before: “Following up to determine next steps on the proposal we sent last week.”After: “Did the proposal hit the mark? If yes, I’ll send the kickoff plan. If not, reply with one change you’d like.”
• Before: “Wanted to touch base regarding scheduling the training.”After: “Can we lock a training date? Options: Sep 12 or Sep 19. Please reply with your pick.”

High-value habits that compound

• One email, one ask: If you have two asks, send two emails. It doubles clarity.
• Word budget: Tell the AI “max 60 words.” You’ll get shorter, sharper drafts.
• Numbers beat adjectives: Replace “soon” with a date, “quick” with minutes.
• Plain English swap: Instead of “circling back,” use “checking in.”

Common mistakes and easy fixes

• Hidden or weak CTA → End with a verb: “Please reply with A or B by Thu.”
• Too formal/stuffy → Ask the AI: “Sound like a colleague. Short sentences.”
• Overflowing details → Move extras to a second email or attach after they say yes.
• No time options → Offer two windows. People pick faster.

1‑week action plan

1. Today (10 minutes): Create your voice snapshot. Save the general prompt and the follow-up prompt in a notes app.
2. Days 1–3: Send 3 AI-assisted emails a day. Cap at 60 words. Track time-to-send and replies in 48 hours.
3. Day 4: Build a 12‑item “phrase bank” from your best lines (subjects, CTAs, time options).
4. Days 5–6: Test two subject styles (e.g., “Quick question” vs “Need your input”). Keep everything else the same.
5. Day 7: Review metrics: reply rate, time-to-decision, average word count. Keep the winners, retire the rest.

What to expect: drafts in under a minute, emails under 60 words, clearer CTAs, faster replies within 24–48 hours. You’ll sound like you—just sharper.

Keep the three bullets. Add the 60‑word cap. One ask per email. That’s the system.

Nice summary, Aaron — I like that you focus on revenue per visitor (RPV) not just clicks. That’s the practical shift most teams miss.

Here’s a compact, practical plan you can run this week to get real, measurable lift from AI-powered pricing tests — even if your site traffic is modest.

What you’ll need

• Analytics access (GA4, Mixpanel or similar) with revenue events tracked.
• A/B testing tool or CMS split-testing feature.
• Baseline metrics: conversion rate, ARPU/AOV, monthly visitors.
• List of customer segments you care about (SMB, enterprise, referral sources).

Step-by-step (do-first mindset)

1. Pick one clear KPI: Revenue per visitor (RPV). This keeps price moves honest.
2. Generate 6 prioritized hypotheses with AI (use the prompt below). Pick the top 2 you can build fast.
3. Build 2 variants only: a pricing-copy variant and a structural variant (e.g., add an anchor or show annual savings).
4. Estimate sample size in your testing tool. If traffic is low, run a sequential test: smaller batches and review by segment after 2 weeks.
5. Launch, monitor for technical issues, and let it run to reach at least ~80% power or a pre-agreed time window (2–4 weeks for mid-traffic sites).
6. Feed raw results back into AI for interpretation and next-step tests.

Concrete example (copy-and-build)

• Control: current pricing page.
• Variant A (anchor): Add a premium $149/mo tier (crossed as “advanced”), label $79/mo as “Most popular” with a green badge and monthly vs annual toggle showing 20% savings.
• Variant B (value-focus): Keep prices but replace long feature list with three clear outcome bullets + customer logo carousel and CTA “Start saving in 24 hours”.
• Track: conversions, revenue per visitor, and trial-to-paid conversion by referral.

Common mistakes & fixes

• Testing price + messaging at once — fix: change only one levers per test.
• Stopping when results look good — fix: use pre-set stopping rules or 80% power as your guide.
• Ignoring segments — fix: always segment by source/device/company size before declaring a winner.

1-week action plan

1. Day 1: Pull baseline metrics and choose RPV.
2. Day 2: Use the AI prompt below to generate hypotheses and test setups.
3. Day 3–4: Build the two variants and QA tracking.
4. Day 5: Launch test.
5. Day 6–14: Monitor for issues; don’t stop early. Review segment-level early signals after 7 days.

AI prompt (copy-paste)

“You are a conversion optimization expert. Analyze this pricing page: [PASTE PAGE COPY OR URL]. Current metrics: conversion rate X%, average order value $Y, traffic Z visitors/month. Generate 6 prioritized hypotheses to increase revenue per visitor. For each hypothesis provide: 1) exact copy and layout changes, 2) suggested price points or bundles, 3) expected impact (low/medium/high) with rationale, 4) one A/B test setup (control vs variant) including which segments to monitor, and 5) estimated sample size or run duration given traffic Z. Also provide two headline variants and two CTA texts to test.”

Start small, learn fast, iterate. The quickest wins are clarity, anchor effects, and highlighting real savings.

You nailed the timebox and the one-human-tweak rule. Let’s turn your five‑minute routine into a dependable mini‑system that boosts replies without losing trust. Two upgrades make the difference: a quick evidence check before you send, and a simple three-touch sequence you can run on autopilot.

Goal: fast, specific intros that feel human, stay accurate, and convert to short conversations.

What you’ll need:

• Lead list with fields: name, role, company, public touchpoint, opening line, CTA, date sent, follow-up dates, outcome.
• Prospect’s public LinkedIn/profile post or company news page.
• An AI assistant you trust for quick summarizing (browser-based is fine).
• A 5-minute timer and a two-line message template.

The 5-minute run (keep it tight):

1. Find one signal (90 seconds): Open their latest public post or company news. Grab a single concrete hook: event, quote, product update, or role change.
2. Evidence gate (30 seconds): Ask yourself: Is this fact visible on their public profile or post? If not visible, don’t reference it. No assumptions.
3. Draft with AI (60 seconds): Use the prompt below to get two 2‑sentence intros and one bump message. Keep under 40 words for the opener.
4. Humanize (60 seconds): Edit one line to add a real detail (shared city, one sentence on what you learned, or a specific compliment). Remove any fluff.
5. Log and tag (30 seconds): Save the final opener, CTA, and the touchpoint in your CRM notes. Set follow-ups for Day 3 and Day 7.
6. Send (30 seconds): 2 sentences + soft CTA. Done.

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

“You are my concise LinkedIn outreach assistant. Using only the public content I paste after this, do the following: 1) List 2–3 specific facts you can verify from the text (no guesses); 2) Propose two different two-sentence openings that reference one fact each (under 40 words, warm and professional); 3) Write one short follow-up bump for 3 days later (10–20 words, no pressure); 4) Flag any uncertainty or missing context in one line. Do not invent details. Keep the language clear and human.”

Insider trick: RATER cue for fast personalization

• Role: their title or team.
• Activity: post, talk, or project.
• Trigger: news, launch, hiring, milestone.
• Evidence: the public proof you saw.
• Relevance: why your note matters now.

Message templates (fill the brackets):

• Initial – professional: “Hi [First Name], your note on [specific point] from [post/event] was useful — especially [small insight]. I help [role/company type] with [relevant outcome]. Worth a quick 15 minutes to compare notes on [topic]?”
• Initial – conversational: “Hey [First Name] — loved your take on [specific]. We’re exploring similar work with [peer/company type]. Open to a quick chat to swap what’s working on [topic]?”
• Bump (Day 3): “Looping back on the [topic] note — open to a quick compare?”
• Value drop (Day 7): “Sharing a 2‑line takeaway we’ve seen for [role]: [insight]. If useful, happy to trade notes for 15 mins.”

Persona hook examples (steal these):

• Head of Sales: “curious how you’re handling ramp time with the new segment — one tweak cut ours by 18%.”
• Ops/COO: “saw the rollout note — what surprised you most in week 1? We learned a simple pre‑mortem saved rework.”
• Product Lead: “your launch post on [feature] — how are you validating the adoption signal? We’ve used a 3-question micro-survey with good signal.”

Quality gate (30 seconds before you send):

• Specificity score (0–3): 0 = generic; 1 = vague; 2 = mentions a real fact; 3 = cites exact quote/event and why it matters. Aim for 2–3.
• Factual check: every claim is visible on their public post/profile?
• Friction check: one ask only; 15 minutes or one question.

What to expect:

• 3–5x faster drafting, with a dependable tone.
• Reply lift when the opener references one clear fact and a single, low-friction ask.
• Occasional small slip-ups — your evidence gate protects your credibility.

Common mistakes and quick fixes:

• Fake familiarity (acting like friends) — use respectful, neutral warmth.
• Data dumping — two sentences only; your calendar link can wait.
• Vague hooks (“love your content”) — cite a line, event, or metric.
• Private/sensitive inputs — stick to public posts and company announcements.
• Sending without a follow-up — schedule Day 3 and Day 7 when you log the first message.

Example (filled):

• Touchpoint: “Spoke at CleanTech Summit on grid storage; new pilot with regional utility.”
• Opener: “Hi Maya, your CleanTech Summit point on storage ROI vs reliability was sharp. We’re mapping utility–storage pilots — open to 15 minutes to compare what’s worked in early stages?”
• Bump (Day 3): “Quick nudge on the pilot compare — open to a short swap?”
• Value drop (Day 7): “A pattern we’re seeing: early pilots improve ROI when ops owns the success metric, not BD. Helpful?”

7-day action plan:

1. Day 1: Load 30 leads. Add a notes column for “public touchpoint.”
2. Day 2: Run the 5-minute routine on 20 leads. Track specificity score and time.
3. Day 3: Send bumps to non-responders. A/B two openers (professional vs conversational).
4. Day 4: Review replies. Keep the higher-performing tone; tweak CTA words.
5. Day 5: Process 40 more leads with the winning tone. Keep the evidence gate.
6. Day 6: Add one persona hook line for your top 3 roles.
7. Day 7: Send value drops. Summarize metrics: reply rate, meeting rate, time per lead, accuracy errors.

Final nudge: Keep it simple, keep it specific, and let the AI do the heavy lifting while you supply the judgment. One real fact + one clear ask beats any long pitch — every time.

Love your plan: starting with a tiny AI motion spec and pasting one-line, per-layer notes into your Figma plugin is the fastest way to see if timing feels right. Let’s add a few pro moves so you can go from idea to a clean, lightweight Lottie in minutes — reliably.

Try this now (under 5 minutes)

• In Figma, draw a 24px circle behind any icon. Name layers: icon and ripple.
• Open your animation/Lottie plugin. Set FPS to 24 or 30.
• Animate ripple: at 0ms scale 0.2, opacity 0; at 120ms scale 1.08, opacity 0.18; at 240ms scale 1.0, opacity 0. Fade out by 300ms if needed. Easing: ease-out.
• Export as Lottie. Expect <10 KB and smooth playback.

What you’ll need

• Figma (or After Effects) and a Lottie-capable plugin (e.g., LottieFiles/Figmotion, or Bodymovin for AE).
• Simple SVGs (1–3 layers). Avoid blur, shadows, and blend modes for export safety.
• An AI assistant to generate a short motion spec with exact numbers.
• A phone or browser preview to check smoothness and size.

Step-by-step (repeatable workflow)

1. Define one sentence and a limit. Example: “Idle → press → success pop.” Target size: <30 KB.
2. Ask AI for a micro-spec with exact values. You want times in ms, easing names, and per-layer transform values at 0, mid, end.
3. Prep layers for Lottie. Name layers clearly (bg, icon, check). Flatten boolean ops, convert strokes to fills where possible, center anchors.
4. Enter keyframes in the plugin. Use only translate/scale/opacity for version 1. Play, then tweak a single number (usually duration or scale).
5. Export as Lottie. Set FPS 24–30. Disable hidden layers. In AE/Bodymovin, prefer “convert text to shapes,” and avoid effects.
6. Preview on-device. Check timing, stutter, and file size. If stutters, stagger motions and simplify paths.

Copy-paste prompt (robust)

“You are an animation engineer for UI microinteractions. Create a Lottie-safe motion spec for a 3-state button: idle → press → success. Constraints: 24–30 FPS, under 30 KB, transform-only (translate/scale/opacity), no blurs or blend modes. Output three parts: 1) concise spec (durations in ms, easing names), 2) per-layer keyframes at 0%, 50%, 100% with exact numbers (scale in %, opacity 0–1, position in px relative to anchor), 3) export notes (layer names, FPS, expected file size). Layers: bg, icon, check. Keep transitions 150–350ms each. Be explicit and minimal.”

Example spec you can drop in now (icon press → checkmark pop)

• States & duration: Idle (0ms) → Press (0–160ms) → Success (160–420ms). Total ≈ 420ms.
• Easing: Press = ease-out; Success pop = back-out; Fade = linear.

• bg
  • 0ms: scale 100%, opacity 1.0
  • 160ms: scale 96%, opacity 1.0
  • 420ms: scale 100%, opacity 1.0
• icon
  • 0ms: scale 100%, opacity 1.0
  • 160ms: scale 92%, opacity 1.0 (ease-out)
  • 260ms: scale 100%, opacity 0.0 (fade out 200–260ms)
• check
  • 160ms: scale 70%, opacity 0.0
  • 260ms: scale 108%, opacity 1.0 (back-out)
  • 420ms: scale 100%, opacity 1.0

Export notes: center anchors on each layer; no masks; FPS 24; trim hidden layers; expect <20–30 KB with simple shapes.

Insider tricks that keep Lottie small and smooth

• Three-keyframe rule: For each layer, use 0%, 50–70%, 100%. It’s enough for natural motion without bloating JSON.
• Path diet: Simplify SVG paths before animating. Every extra node becomes weight.
• Transforms over morphs: Prefer scale/translate/opacity. Morphs can balloon size and cause stutter.
• Anchor sanity: Place anchors where movement pivots. Center for pops; bottom for slide-ins.
• Whole pixels: Round positions to whole numbers to avoid shimmer on low-DPI screens.

What to expect

• Most 1–3 layer microinteractions land under 40 KB on first export; 20–30 KB after a quick path cleanup.
• Playback should be smooth at 24–30 FPS. If not, stagger layer animations by 40–80ms and remove simultaneous heavy moves.
• Timeboxing works: one 15-minute loop usually gets you a shippable animation.

Mistakes and fast fixes

• Using effects (blur, shadows). Fix: replace with a subtle opacity fade or a second, lighter fill layer.
• Too many layers. Fix: merge decorative shapes; keep only the moving parts separate.
• “Soft” motion that feels laggy. Fix: shorten the press to 120–160ms; use ease-out instead of linear.
• File size spikes. Fix: simplify paths, reduce keyframes, drop FPS to 24.

Action plan

1. Today (20 minutes): Pick one microinteraction, run the prompt above, build it with transform-only keys, export, and preview on your phone.
2. Tomorrow (20 minutes): Optimize: simplify paths, tidy layer names, set FPS 24–30, re-export and log file size.
3. This week: Ship 2–3 Lotties into staging, measure tap-to-feedback time and any change in task completion or click-through.

Closing thought

Keep it tiny, timed, and testable. Ask AI for exact numbers, paste them once, and you’ll have crisp microinteractions you can export as Lottie on a reliable loop — no heavy software required.