aaron

Try this now (under 5 minutes): Open your AI tool and paste the prompt below. You’ll get a one‑screen influencer brief, a DM you can send to creators, and a one‑line reporting template. That’s your pilot ready today.

Copy‑paste AI prompt:

“You are an expert influencer campaign producer. My goal is: [STATE ONE KPI, e.g., ‘500 link clicks in 3 weeks’]. Audience: [TARGET]. Product: [WHAT IT IS + KEY BENEFIT]. Offer: [DISCOUNT OR CTA]. Create: 1) A one‑page brief with goal, deliverables (post/story), must‑mentions, timeline, and tracking instruction using a unique short link or discount code. Keep it one screen. 2) A DM I can paste to creators inviting them to participate. 3) A one‑line reporting template creators can paste back: ‘URL | Impressions | Clicks | Spend (if any)’ and the due day each week. 4) Three caption options (short/medium/long), five relevant hashtags, two image/story directions, and two CTA lines (direct + soft). Keep language natural and easy to adapt to the creator’s voice.”

The problem: Non‑technical teams drown in details — too many metrics, scattered briefs, inconsistent tracking. Decisions slow down and budgets drift.

Why it matters: One KPI, one link/code per creator, one weekly check forces focus. It turns influencer marketing into a controllable acquisition channel, not a guessing game.

What I’ve learned: After hundreds of pilots, the winners share three traits — creator‑friendly briefs, ruthless weekly reallocation, and clean attribution (UTMs or codes). Everything else is optional.

Step‑by‑step (do this in order)

1. Pick the KPI and threshold. Example: Clicks with a target CPC ≤ $1.50 for the pilot. Write it at the top of the brief.
2. Create unique tracking for each creator. Use one short link or one discount code per creator. If using UTMs, use a consistent pattern: utm_source=influencer, utm_medium=social, utm_campaign=[CampaignName], utm_content=[CreatorName_Format].
3. Generate your brief with the AI prompt. Edit for clarity; keep it one screen. Include the reporting line and deadline (e.g., every Monday by 10am).
4. Send the DM + assets. Attach logo, one product image, key message, and their unique link/code. Offer a small bonus for on‑time reporting.
5. Collect URLs + numbers weekly. Use a shared sheet or simple form with columns: Creator, Post URL, Impressions, Clicks (or your KPI), Spend, Date, Notes.
6. Summarize with AI and reallocate. Paste the week’s rows into your AI tool to produce a one‑paragraph summary and a keep/tweak/pause decision per creator (prompt below). Move budget toward winners within 24 hours.
7. Rinse weekly for 2–4 weeks. Lock in the top performers and brief 2–3 similar creators for scale.

AI prompt for weekly summary and decisions

“You are my performance analyst. Here is our KPI: [e.g., Clicks, CPC ≤ $1.50]. Here are this week’s results (paste table rows: Creator | Impressions | Clicks | Spend | Notes). 1) Calculate CTR (Clicks/Impressions), CPC (Spend/Clicks), and rank creators by CPC then CTR. 2) Write a 120‑word executive summary in plain English with totals vs. goal and the single biggest driver. 3) Give per‑creator actions: KEEP (scale), TWEAK (new caption/time), or PAUSE, each with one sentence reason. 4) Suggest one test for next week (message angle, format, or timing). Keep it decisive and non‑technical.”

Metrics to track (and simple formulas)

• Clicks (primary KPI) — the number that moves decisions.
• CTR = Clicks / Impressions. Use to judge message/creative resonance.
• CPC = Spend / Clicks. Your efficiency gate for reallocating budget.
• Compliance rate = Creators who reported on time / Total creators.
• Optional: Conversions and CPA if you have codes/checkout tracking.

Insider tricks that lift results fast

• Pin the CTA in comments and ask creators to mention “link in bio or pinned comment.” Small change, big click lift.
• Story follow‑ups within 24 hours using the same link/code often reclaim 10–30% missed clicks.
• Force‑rank creators by CPC then CTR. Only scale the top third; move the bottom third to “tweak or pause.”

Mistakes & fixes

• Mixing KPIs mid‑pilot. Fix: Freeze the KPI for 2–4 weeks; change only after the pilot ends.
• Messy UTMs or duplicate links. Fix: Maintain a single mapping row per creator with their exact link/code. Test each link before posting.
• Over‑scripted briefs. Fix: Provide angles and must‑mentions, not word‑for‑word scripts.
• Poor reporting compliance. Fix: One‑line template, weekly reminder, and a small on‑time bonus.

What to expect: Week 1 is noisy — results vary widely. By Week 2, your CPC and CTR stabilize. If your CPC is above target, switch message angle or format; if below, double down and bring in look‑alike creators.

1‑week action plan

1. Today: Pick KPI + threshold. Build one unique link/code per creator. Run the brief prompt and finalize the one‑pager.
2. Day 2: Send DMs + assets + reporting line. Set a calendar reminder for weekly reporting.
3. Day 3–4: Posts go live. Capture URLs. Check links/codes function.
4. Day 5: Mid‑week pulse: ask for early numbers from anyone underperforming to course‑correct.
5. Day 6: Paste data into your sheet. Run the weekly AI summary prompt. Decide KEEP/TWEAK/PAUSE.
6. Day 7: Reallocate budget to winners. Update briefs for tweaks. Queue next week’s posts.

Decision rules (print these)

• If CPC ≤ target and CTR is rising: scale budget by +25–50% next week.
• If CPC > target but CTR is average: tweak message/format and try one more post.
• If CPC stays above target after two tries: pause and replace with a similar creator to your top performer.

Simple system, fast feedback, focused spend. Your move.

— Aaron

Jeff’s quick win is right: small batches + one format + fast review = speed and control. Now let’s turn that into a repeatable system that delivers consistent quality, measurable progress, and less rework.

The issue: ad‑hoc prompts produce uneven problems and unclear steps. You waste time fixing avoidable errors.

Why this matters: a simple “practice set factory” gives you reliable output, traceable difficulty, and a clear audit trail. That’s how you scale beyond one good set.

What you’ll need

• An AI chat.
• A one-page checklist (format fields, difficulty scale, review criteria).
• A simple spreadsheet or document to log sets, errors, and learner results.

The playbook (do this in order)

1. Define the target: topic, sub-skills, level, and difficulty scale (1–5). Example (fractions): addition, subtraction, multiplication, division, simplifying.
2. Lock the format: every item must include Problem, Hint, Steps (numbered), Final Answer, Self‑check, Difficulty (1–5), Est. Time (sec), Skill tag, and a targeted misconception.
3. Generate with constraints using the prompt below. Require a coverage summary so you can verify distribution at a glance.
4. Audit separately: run an “auditor” prompt that recomputes each item, flags PASS/FAIL, and fixes only the failures. This keeps quality high without regenerating everything.
5. Calibrate difficulty: compare Est. Time vs. actual time a learner needs. Adjust difficulty scale rules in your prompt (+/− 1 level) next run.
6. Version and reuse: save “Form A” and auto-create a “Form B” variant (same skills, new numbers) for spaced practice.
7. Log outcomes: record accuracy, review time, and learner scores. Tune prompts based on what the data says.

Copy-paste prompt: Practice Set Generator (robust)

“Create 10 practice problems in [SUBJECT], focusing on [SKILL_TAGS], for a [LEVEL] learner. Mix problem types: [PROBLEM_TYPES]. For each item, output the following labels exactly, one per line:

ID: [1..10]Skill: [one of SKILL_TAGS]Problem: [clear, standalone prompt]Hint: [1 concise sentence]Steps: [numbered steps, show each arithmetic transformation clearly]Final Answer: [single value or expression]Self-check: [quick verification or substitution that confirms the result]Difficulty: [1–5]Est. Time (sec): [integer 20–120]Common Misconception: [specific error to watch for]

Constraints:- Steps must be 4–8 lines, readable by a non-expert.- Use friendly but non-trivial numbers; avoid repeating numbers across items.- Distribute skills roughly evenly and vary contexts.- No images; text only.

After the 10 items, add:Summary: coverage by Skill (%), average Difficulty, average Est. Time (sec).”

Optional quick auditor prompt

“Act as a solution auditor. For each item above: independently recompute; mark PASS or FAIL; if FAIL, provide Corrected Steps, Corrected Final Answer, and the Root Cause. Confirm that Self-check actually validates the result. Output only items that FAIL plus a 1-line summary of total PASS/FAIL.”

What to expect

• First run: 1–3 items may need fixes. The auditor prompt closes most gaps in under 3 minutes.
• After two iterations: accuracy ≥ 95% and review time ≈ 60–90 seconds per problem.
• Variant generation becomes one command (“Form B”), maintaining difficulty and coverage.

Metrics to track (weekly)

• Content accuracy after audit (%). Target: ≥ 95%.
• Average review time per problem (sec). Target: ≤ 90.
• Readability score (your judgment, 1–5). Target: ≥ 4.
• Distribution coverage by skill (%). Target: within ±10% of plan.
• Learner correct rate on Form A vs. Form B (%). Target: +10–20% on second exposure.
• Rework rate (items needing regeneration). Target: ≤ 10%.

Common mistakes and fast fixes

• Inconsistent format → Force labeled fields exactly; your prompt should state “labels exactly as written.”
• Woolly steps → Cap to 4–8 steps and require each arithmetic move on its own line.
• Drifting difficulty → Include Est. Time and recalibrate next run based on real timings.
• Weak self-checks → Demand a concrete verification (substitute back or recompute a different way).
• Overfitting to one pattern → Require varied contexts and rotate number ranges.

7‑day plan (light lift)

1. Day 1: Set your skills list, difficulty rules, and the fixed format (copy the labels above into a doc).
2. Day 2: Run the Generator for fractions (10 items). Log coverage and timing.
3. Day 3: Run the Auditor; fix fails only. Aim for ≥ 95% accuracy.
4. Day 4: Create Form B variants (same skills/difficulty, new numbers). Save both sets.
5. Day 5: Pilot with one learner; record time per problem and correct rate.
6. Day 6: Adjust difficulty rules and hint style based on data; regenerate 5 targeted items.
7. Day 7: Roll the template to a second topic (decimals or ratios) using the same system.

Insider tip: add “Common Misconception” per item. It forces the model to design for typical errors, which lifts learning value and reduces vague steps.

Your move.

Hook: Use AI to extract ethical, actionable insights — but treat outputs as hypotheses, not facts. Protect people first; act on signals only after human validation and testing.

The immediate problem: Teams hand sensitive data to models, get plausible-sounding patterns, and change product flows — then face privacy issues, biased features, or wasted development time.

Why this matters: One good insight implemented safely can move KPIs (conversion, retention). One bad insight implemented carelessly can cost reputation and users.

What I’ve seen work: Start with a narrow question, a minimized anonymized sample, and a strict human-review + A/B test workflow. That combination surfaces useful leads while limiting risk.

What you’ll need

• One clear question (single sentence).
• Minimal, anonymized dataset (only columns required).
• Consent/legal basis and an audit log.
• Secure storage and role-based access.
• One product owner and one independent reviewer for bias checks.

Step-by-step (do this)

1. Define outcome: write a single measurable goal (e.g., increase trial-to-paid conversion by 10% in 30 days).
2. Create minimal sample: include only the fields required, replace IDs with random tokens, and drop direct identifiers.
3. Aggregate where possible: use cohort/week or bucketed ranges to avoid single-user signals.
4. Run the AI on the sample with explicit safety guardrails (don’t infer demographics or re-identify).
5. Human review: product owner + reviewer evaluate up to 5 hypotheses, rank by expected impact and ease of test.
6. Design 1–2 A/B tests for highest-priority hypotheses; define primary metric and sample size estimate.
7. Run test, measure, and iterate. Update documentation and retention policies for AI outputs.

Metrics to track

• Primary KPI lift (e.g., conversion rate change %).
• Validation rate: % of AI insights confirmed by human review and testing.
• Time-to-insight: days from question to test-ready hypothesis.
• Access audits: number of model queries and who ran them.

Common mistakes & fixes

• Sending raw PII to models — Fix: anonymize and aggregate first.
• Trusting the model blindly — Fix: require human sign-off and an experimental test before product changes.
• No consent/legal check — Fix: pause, confirm lawful basis, or use synthetic data.

7-day action plan (exact next steps)

1. Day 1: Write one-sentence question and success metric.
2. Day 2: Pull minimal sample and anonymize it.
3. Day 3: Run AI query (use prompt below).
4. Day 4: Joint review with product and compliance reviewer.
5. Day 5: Design A/B test for top insight, calculate sample size.
6. Day 6: Implement experiment and monitoring dashboard.
7. Day 7: Start test and collect early signals; schedule review at minimum n per sample plan.

Copy-paste AI prompt (use as-is)

Analyze this anonymized dataset and provide up to 5 ranked hypotheses explaining the behavior related to [GOAL]. Columns: user_token, cohort_week, sessions_per_week, time_on_site_min, feature_x_used (yes/no), converted (yes/no). For each hypothesis include: observed pattern, estimated confidence (low/medium/high), one clear product recommendation, and a single A/B test idea. Do NOT attempt to re-identify users, infer demographics, or provide instructions for de-anonymization. Flag any potential bias you detect.

Your move.

Quick win (5 minutes): grab the latest patent alert you received, paste the title+abstract into the prompt below and ask the LLM for a 2‑sentence summary + a Relevant/Maybe/Ignore label. You’ll see immediately how much time a summarizer saves.

The problem: patent databases overwhelm with noise. Non‑technical users either spend hours scanning or miss important developments.

Why it matters: a tight surveillance workflow turns distraction into strategic insight — you save time, reduce missed opportunities, and keep control of decisions without hiring a developer.

Short lesson from practice: start narrow, automate capture+summarize, and always include a one‑line human review. That single human step prevents most mistakes and keeps the system useful.

What you’ll need (5–30 minutes to prepare)

1. An account on a patent database that supports alerts (email or RSS).
2. A simple automation tool (email-to-spreadsheet or a connector like a drag‑and‑drop automation).
3. Access to an LLM summarizer (web service or API access via a service).
4. A spreadsheet with columns: title, link, abstract, 2-sentence summary, label, confidence, reviewer, notes.

Step-by-step setup (1–2 hours)

1. Create one focused search: 3–6 keywords + 1 classification code. Time: 15–30 minutes.
2. Activate alerts (daily or weekly) and route them to a single inbox or RSS. Time: 10 minutes.
3. Automate capture: pipe new alerts into the spreadsheet, auto-fill title, link, abstract. Time: 20–40 minutes.
4. Attach the LLM task: for each new row, run the summarizer to produce: 2-sentence summary, label (Relevant/Maybe/Ignore), 3 keywords, and confidence. Time: 15–30 minutes to set template.
5. Weekly routine: review summaries (15–30 minutes), confirm labels, move true hits into a working list for deeper review.

Copy-paste LLM prompt (use as-is)

“You are a technical summarizer. Given the patent title, abstract, applicants, and publication date, do the following in plain text: (1) Write a 2-sentence summary of the invention. (2) List 3 concise keywords. (3) Assess likely novelty vs general field (answer: high / medium / low) and explain in one short sentence. (4) Recommend a label: Relevant / Maybe / Ignore. (5) Suggest one search term or classification code to add or remove to improve future alerts. Do not provide legal advice and only use the supplied text.”

Metrics to track (weekly/monthly)

• Weekly triage time (target: 15–30 minutes/week).
• False positive rate (LLM says Relevant but you mark Ignore) — target: under 40% first month, <25% after tuning.
• Hits/month (items moved to deep review) — target: 2–8 depending on topic.
• Sampled false negatives (check 10% of Ignored items monthly) — look for missed high‑priority items.

Common mistakes & fixes

• Too broad search: add classification codes or exact-phrase filters.
• Full-text automation: avoid parsing full PDFs — use abstract+bibliographic data to reduce noise and cost.
• No review cadence: if triage exceeds 30 minutes/week, tighten filters or add an extra label so the LLM prioritizes higher-confidence items.

1-week action plan

1. Day 1 (1 hour): build one focused search, enable alerts, create the spreadsheet.
2. Day 2 (30 minutes): set up the automation to capture alerts into the sheet.
3. Day 3 (30 minutes): hook the LLM template using the prompt above and test with 5 sample abstracts.
4. Day 4–7: run the system, perform one 20‑minute review session, and adjust keywords/class codes based on results.

Your move.

Quick win: you can turn your newsletter pile from guilt to fuel in a single afternoon with an AI-powered triage that makes reading a 2-minute daily decision and a 10-minute focus block.

The problem

Newsletters arrive constantly. You don’t need to read them all — you need to surface what helps you make decisions, learn, or act. Without a simple system everything becomes noise and stress.

Why this matters

Time is finite. If your inbox is a firehose you lose both focus and opportunity. A predictable routine converts noise into prioritized actions and saves hours each week.

Short lesson from practice

I set up the same pattern for non-technical executives: one capture point, three tags, and an AI that returns a one-line summary + action. Within a week unread volume drops, and you actually act on the useful items.

1. What you’ll need
   • An email filter or aggregator to send all newsletters to one folder/feed.
   • Three folders/tags: Now, Maybe, Archive.
   • An AI summarizer (built-in assistant, Zapier+AI, or copy-paste to a chat tool).
   • Daily: 2 minutes triage + 10 minutes for Now. Weekly: 20–30 minutes review.
2. How to set it up — step by step
   1. Capture: Create an email filter so every newsletter lands in a single folder.
   2. Auto-filter: Tag by sender or keywords to auto-assign probable Now items (trusted authors, your topics).
   3. AI summary: Configure the AI to return a one-line takeaway, a 1–2 item action, a relevance 1–5 score, and an estimated read time.
   4. Daily triage (2 minutes): Scan AI one-liners. Move items with actions or relevance ≥4 to Now. Everything else → Maybe or Archive.
   5. Workblock (10 minutes): Open Now items and perform the one-line action or schedule it. Archive afterwards.
   6. Weekly tidy: Review Maybe, promote 5–10% to Now, archive the rest.

Copy-paste AI prompt (primary, use as-is)

Read the newsletter below. Give me: 1) one-sentence summary (clear takeaway), 2) a relevance score 1–5 for my goals (business strategy, productivity), 3) up to two concrete actions I should take (each 6–10 words), 4) estimated read time. Keep answers short and label each part.

Variants

• Quick-scan: “One-line takeaway + relevance score only.”
• Action-mode: “List 1–3 actions, priority 1–3, suggested deadline.”
• Deep-read rec: “Explain why this matters in 3 bullets and what to read next.”

Metrics to track (weekly)

• Unread newsletter count (start vs end of week).
• Now items per day and % completed within 24 hours.
• Average daily time spent on newsletters.
• Archive rate (items archived / total received).

Common mistakes & fixes

• Over-tagging: If too many items become Now, tighten the relevance threshold to ≥4.
• Relying on AI for decisions: Use summaries for triage only; read full when the action matters.
• Privacy risk: Don’t send sensitive internal newsletters through third-party tools; use local or enterprise options.

One-week action plan

1. Day 1: Create the newsletter filter and three tags. Route one newsletter in as a test.
2. Day 2: Plug in the AI summarizer or copy-paste prompt; run triage on today’s mail.
3. Days 3–6: Use the 2-minute triage each morning; act on Now items in a 10-minute block.
4. Day 7: Run the weekly tidy, measure the metrics above, adjust the relevance threshold.

Your move.

Quick win (under 5 minutes): Take the three most recent reminders (phone, flagged email, Slack mention) and paste them into one Google Sheet row with columns: title, source, created_date. You’ll instantly see the benefit of a single daily inbox.

A useful point you made: Starting read-only and batching 10–50 items for the AI step is smart — it gives confidence without automation risk. I’ll build on that with a focus on measurable results and a tighter pipeline.

The problem: Reminders are fragmented across apps, duplicated, and inconsistent — so decisions are deferred, things slip, and attention fragments.

Why it matters: Consolidation is not about neatness — it’s about improving execution. Target outcomes: fewer missed deadlines, higher completion rate, and less time hunting for context.

Experience-led lesson: I’ve run this for busy leaders — start with 3 sources, a clear schema, an AI dedupe/prioritization step, and measurable acceptance thresholds. That delivers a trustworthy list within days, not months.

1. What you’ll need
   • Accounts for your sources (email, calendar, Slack, Todoist).
   • An automation tool (Zapier/Make/Power Automate) — start with free tiers.
   • Destination: Google Sheet or Notion DB for visibility.
   • AI access (API key via your automation tool).
2. Step-by-step setup (do this)
   1. Inventory: List sources and available triggers (email forward, webhook, connector).
   2. Schema: Create columns — title, notes, source, created_date, due_date, link.
   3. Connect 2–3 sources read-only into your pipeline and map to schema.
   4. Batching: Every hour collect new items into a batch (10–50) and call the AI for dedupe + priority.
   5. Write-back: Save cleaned output to destination and send a single daily digest email or Slack DM.
   6. Review: Manually review a sample daily for 7 days before any write-back to originals.

Copy-paste AI prompt (drop into your automation):

Please process this batch of reminder items. Each item: title, notes, source, created_date, due_date (optional). Return a JSON array where you: 1) remove exact and near-duplicates, 2) infer priority (High/Medium/Low) using due_date and keywords, 3) assign category from {Call, Email, Errand, Admin, Project, Meeting, Follow-up}, 4) produce a one-line standardized title, 5) include confidence (0-100) and a suggested due_date when missing. Output fields: title, category, priority, due_date, source, confidence, original_id.

Metrics to track (and targets)

• Consolidation rate: % of identified sources feeding into one list — target 90% in week 2.
• Duplicate reduction: duplicates before vs after — target ≥80% reduction.
• Task completion lift: weekly completed tasks — aim +25% in month 1.
• Trust score: % of AI items with confidence ≥70 — target 80%.

Common mistakes & fixes

• Over-automation: start read-only to avoid loops; enable writes after trust built.
• Poor dedupe: increase semantic similarity threshold and add example pairs.
• Missing dates: use rule no due_date = today + 7 days and let users override.

1. 1-week action plan
   1. Day 1: Inventory + choose destination (30–60 min).
   2. Day 2: Set up 2 connectors read-only (60 min).
   3. Day 3: Build schema + test data flow (45 min).
   4. Day 4: Add AI batch step and run 50 items (60–90 min).
   5. Day 5: Measure metrics, tweak prompt, add one more source (45–60 min).
   6. Days 6–7: Monitor confidence and completion lift; prepare for selective write-backs (30–60 min).

Your move.

Yes — and you can make it repeatable, compliant, and conversion-focused in under 60 minutes. The win isn’t the draft; it’s the system: extract, translate, prove, and test — every time.

Do this, not that

• Do start with a measurable outcome and write the benefit first, then add the technical “because.”
• Do build a quick Spec→Benefit matrix: feature, buyer pain, benefit, metric, proof source.
• Do tier claims: Verified (exact number), Directional (range), Qualitative (no number) — and label them in draft.
• Do produce two tones per asset: technical buyer vs business buyer.
• Don’t invent numbers or “industry-leading” superlatives without evidence.
• Don’t lead with architecture; lead with time/cost/risk outcomes.
• Don’t ship without a quick SME check and a claim scan.

Insider trick: Claim Tiering + Evidence Tags

• Create [E1], [E2], [E3] tags that point to exact lines in the spec or an approved source. Insert them inline after each claim in your draft. It speeds approvals and slashes rework.
• Use tier labels in the draft: [Verified], [Directional], [Qual]. If Legal pushes back, you can instantly downgrade a claim without rewriting the whole page.

What you’ll need

• Product spec (even one paragraph is fine).
• Buyer persona snapshot (top pain + decision trigger).
• Tone example (one short sample you already like).
• LLM access and one SME for a 10-minute fact check.

Process (fast, repeatable)

1. Extract signals (benefits, metrics, proof) from the spec with the Extraction Prompt below. Output a mini matrix.
2. Draft variants for two personas (technical and business) using the Copy Prompt. Force a measurable benefit in each.
3. Tag proof by inserting [E#] after each claim and add tier labels.
4. Run a claim QA pass with the Guardrail Prompt. Remove or downgrade anything not backed by [E#].
5. A/B test the headline and 150-word section. Keep the CTA constant for clean read.

Copy-paste prompts

• Extraction Prompt: “You are a product marketer. From the following technical spec, extract: 1) top 5 buyer pains implied by the spec, 2) 5–7 benefits in buyer language, 3) all measurable outcomes with their exact spec quotes, 4) risks/limitations. Output as bullet lists with Evidence Tags [E1..En] referencing the verbatim spec lines. Spec: [paste]. Persona: [describe].”
• Copy Prompt: “You are a senior B2B copywriter. Using these benefits and Evidence Tags [paste Extraction output], write: a) 3 headlines, b) a 50-word elevator blurb, c) a 150-word feature→benefit section, d) 2 CTAs. Produce two versions: V1 for [technical buyer], V2 for [business buyer]. Each claim includes a [Verified]/[Directional]/[Qual] label and [E#] tag. Tone: confident, helpful, concise.”
• Guardrail Prompt: “You are a fact-checker. Review this copy. For each sentence: 1) flag any claim without [E#], 2) suggest a compliant rewrite using available [E#], 3) replace risky words (best, never, always) with precise alternatives, 4) highlight any implied guarantees. Copy: [paste].”
• Objection Handler Prompt: “Act as a skeptical buyer. List the top 5 objections to this copy and provide a one-sentence pre-bunk for each. Keep them in customer language. Copy: [paste].”

Worked example (spec → marketing copy)

Spec snippet: “New ingestion pipeline sustains 2× throughput at 99.95% availability; average deployment time 2 hours; SOC 2 Type II; AES-256 at rest; typical customer reduces nightly batch window by 38%.”

• Headline options (business)
  • “Close your batch window 38% faster — without weekend fire drills.” [Verified][E5]
  • “Double the data in the same window. Same team, fewer after-hours.” [Directional][E1]
  • “Ship dashboards before 9am — even on peak nights.” [Qual][E1]
• 50-word blurb (technical): “Sustain 2× throughput at 99.95% availability with AES-256 and SOC 2 Type II controls. Typical customers cut batch windows by 38%, so ops teams finish on schedule without paging on-call. Deploy in about 2 hours, no code changes for common connectors.” [Verified][E1][E2][E3][E5]
• 150-word feature→benefit: “Your reports don’t slip because the pipeline keeps up. The new ingestion path sustains 2× throughput under load, so the same hardware clears more data before 9am. That typically trims nightly windows by 38%, which means fewer after-hours escalations and happier stakeholders. Availability holds at 99.95%, backed by SOC 2 Type II controls and AES-256 at rest, so Security stays comfortable while you move faster. Deployment takes about 2 hours with prebuilt connectors — no re-architecture. Bottom line: faster closes, fewer overtime hours, and fewer ‘where’s my dashboard?’ emails.” [Verified/Directional][E1][E2][E3][E4][E5]
• CTAs: “See your data window by tomorrow morning” and “Get a 2-hour guided setup.”

What to expect

• First drafts: 70–90% usable. Expect to tweak numbers, tone, and compliance wording.
• Time-to-publish drops by 50–70% once the Evidence Tag habit is in place.

Metrics to track

• Headline CTR (email/landing) ≥ baseline +15% within two tests.
• Landing conversion rate lift ≥ 10% relative.
• Dwell time on feature section (scroll depth + seconds on section).
• Time-to-publish (hours saved vs. prior method).
• Support tickets referencing unclear messaging (target: -25% in 30 days).
• Legal/SME revision cycles (target: ≤1 pass) — Evidence Tags reduce back-and-forth.

Common mistakes & fixes

• Vague “industry-leading” claims — fix: swap for a measured or directional metric with [E#].
• Feature dumps — fix: one-sentence benefit first, “because” clause second.
• One-tone fits all — fix: publish technical and business variants; route by traffic source.
• Claims without proof — fix: Guardrail Prompt + Evidence Tags before any stakeholder review.

7-day action plan

1. Day 1: Paste one spec into the Extraction Prompt; build your Spec→Benefit matrix.
2. Day 2: Generate two persona-based drafts with the Copy Prompt; insert [E#] and tiers.
3. Day 3: Run the Guardrail Prompt; SME reviews only flagged lines.
4. Day 4: Finalize two variants; prep A/B test (headline + 150-word section).
5. Day 5: Launch test; keep CTA constant.
6. Day 6: Monitor CTR, scroll depth, and early conversions; capture objections via the Objection Handler Prompt and add pre-bunks.
7. Day 7: Pick the winner; document the Evidence Tags used; templatize for the next spec.

Your move.

Hook: Yes — AI can turn reminders scattered across apps into one clean, prioritized list. It’s not magic; it’s a repeatable process that saves time and removes anxiety.

The core problem: You have reminders in several places (phone reminders, email flags, calendar events, Slack, Todoist, Outlook) and no single, trustworthy view.

Why it matters: Fragmentation costs attention and causes missed tasks. Consolidation reduces cognitive load, increases completion rates, and frees hours per week.

Lesson from practice: Start simple. Pick a single destination for the consolidated list, automate ingest from 3–5 sources, and use an AI step to normalize, deduplicate, and prioritize. Iterate.

1. What you’ll need
   • Accounts for your reminder sources (Google, Apple, Outlook, Todoist, Slack, etc.).
   • An automation tool with connectors (Zapier, Make, or Microsoft Power Automate).
   • A destination list: Google Sheet, Notion database, Todoist project, or a single Reminders list.
   • Access to an AI model (via the automation tool or a simple API key you paste into the automation).
2. Step-by-step setup
   1. Inventory: List all apps and note how items can be exported (webhooks, email forwards, API, or Zapier triggers).
   2. Pick a destination: Choose one place you will check daily.
   3. Create connectors: In your automation tool, create triggers for each source that send new/updated reminders to a central pipeline.
   4. Normalize: Map fields into a standard schema (title, due date, source, link, notes, created date).
   5. AI step: Send batched items to the AI to remove duplicates, infer priority, and assign simple categories (call, email, errand, follow-up).
   6. Write back: Save the cleaned list to your destination and optionally push a daily summary to email or Slack.

What to expect: Initial accuracy ~70–85% for dedupe/priority. Improve by adding examples and rules. Expect a 30–90 minute setup plus minor tuning.

Metrics to track

• Consolidation rate: % of sources feeding into the single list.
• Duplicate reduction: # duplicates before vs after.
• Task completion change: % increase in tasks completed weekly.
• Time saved: estimate minutes saved per week by checking one list.

Common mistakes & fixes

• Missing fields — Fix by adding simple default rules (no due date = today+7).
• Rate limits or auth failures — Use email-forward fallback or stagger polling.
• Over-automation — Start with read-only consolidation, then add write-backs after confidence rises.

Copy-paste AI prompt (use in your automation AI step):

“You receive a list of reminder items. Each item has: title, notes, source, created_date, due_date (optional). Return a cleaned list where you: 1) remove exact and near-duplicate items, 2) infer a priority (High, Medium, Low) using due_date and keywords (urgent, ASAP, follow up, call), 3) assign a category from {Call, Email, Errand, Admin, Project, Meeting, Follow-up}, 4) provide a one-line standardized title and a confidence score (0-100) for the priority. Output JSON array of items with fields: title, category, priority, due_date, source, confidence.”

1-week action plan

1. Day 1: Inventory sources and choose destination (30–60 min).
2. Day 2: Set up 2–3 connectors into automation (60 min).
3. Day 3: Create normalization schema and test data flow (45–60 min).
4. Day 4: Add AI dedupe/prioritization step and test with 50 items (60–90 min).
5. Day 5: Review results, tweak prompt/rules, add one more source (45–60 min).
6. Days 6–7: Monitor, measure metrics, and finalize daily digest delivery (30–60 min total).

Your move.