aaron

Can AI design custom fonts or tweak existing typefaces? Short answer: yes—for concepting and batch variation. But don’t expect a production-ready, perfectly licensed typeface without human work.

The problem: Many expect AI to output a finished font file they can ship. Reality: AI accelerates ideation, generates vector suggestions, and automates repetitive tasks—but a designer and a font editor are still required.

Why it matters: Fonts affect legibility, brand perception and conversion. Using AI correctly saves weeks on iterations and reduces design cost. Using it incorrectly exposes you to quality, accessibility and licensing risk.

Practical lesson: I use AI to create glyph variations, spacing suggestions, and batch-modify style weights. Then a human cleans, tests and finalizes kerning, hinting and licensing.

Do / Do not checklist

• Do start from an open-source or licensed base.
• Do use AI for concept variations, not final exports.
• Do plan for manual kerning, hinting and accessibility checks.
• Do not copy proprietary fonts or ask AI to reproduce copyrighted glyphs.
• Do not skip license clearance before commercial use.

Step-by-step: what you’ll need, how to do it, what to expect

1. Gather: a license-cleared font or sketches, a short style brief (tone, contrast, x-height), and reference words for testing.
2. Generate: prompt an AI to produce 5 glyph variations per character (or SVG path ideas). Expect rough vector outlines or raster sketches—these are starting points.
3. Import: bring the best variants into a font editor (Glyphs, FontLab, or FontForge). Clean paths, unify metrics, set anchors.
4. Refine: manual kerning, spacing, hinting and test across sizes/devices. Run legibility checks and user-readability tests.
5. Export & license: produce OTF/TTF and document licensing and usage rules.

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

AI prompt (copy-paste): “You are a professional type designer. Given an open-source serif font with high contrast and a short brief: make five distinct design variations for the lowercase ‘a’ that increase character personality for a luxury brand. Provide each variation as a simple SVG path string and a one-line rationale (e.g., higher contrast, bracketed serif). Do not copy any existing proprietary fonts.”

Metrics to track

• Time to first usable glyph (hours)
• Iteration count per character
• Readability test: % correct in timed reading task
• Conversion lift for assets using the new font (A/B test)

Common mistakes & fixes

• Mistake: Treating AI output as final. Fix: Always import to font editor and refine manually.
• Mistake: Licensing oversight. Fix: Use open-source or get rights before commercial use.
• Mistake: Poor spacing/kerning. Fix: Automated spacing suggestions + manual pair testing.

1-week action plan

1. Day 1: Choose a license-cleared base and write a one-paragraph style brief.
2. Day 2–3: Run AI prompt to generate glyph options; select top 10 per key letters (a, e, o, n, t).
3. Day 4–5: Import into font editor, align metrics and set preliminary kerning.
4. Day 6: Run quick readability tests and gather feedback from 5 users.
5. Day 7: Finalize a pilot OTF for an A/B test in one channel (homepage headline or email).

Your move.

Quick win: start forecasting inventory in one weekend — no data scientist, no expensive software.

The problem: You order by gut. That leads to stockouts, excess holding costs, and missed sales.

Why it matters: Cutting stockouts by even 10% and trimming 5–10% of excess inventory improves cash flow and customer satisfaction immediately.

My practical lesson: Begin with rules + human checks. Forecasting is a tool to turn 6–24 months of sales into reliable reorder points. Use it to make consistent decisions, then improve.

Do / Don’t checklist

• Do: Start with top 20 SKUs, weekly data, and a one-week safety stock.
• Do: Tag promotions, holidays, and supplier delays in your data.
• Don’t: Try to forecast every SKU at once.
• Don’t: Ignore model errors — review weekly.

Step-by-step (what you’ll need, how to do it, what to expect)

1. Prepare: Export 6–24 months of sales by SKU (date, units). Note promotions and closures.
2. Clean: Remove obvious errors and mark anomaly weeks (one-offs).
3. Compute baseline: In a spreadsheet, get average weekly sales per SKU and standard deviation.
4. Set safety stock: Start with 1× avg weekly sales. Increase for high volatility (use SD to adjust).
5. Calculate reorder point: (avg weekly sales × lead time in weeks) + safety stock.
6. Pilot: Apply to top 10–20 SKUs for 4 weeks. Each week, compare forecast vs actual and record forecast error (MAPE or simple % error).

Worked example (copyable)

• SKU: Coffee Beans — avg weekly sales 20, lead time 2 weeks, safety stock 20 → reorder point = (20×2)+20 = 60.
• If weekly variability (SD) is 8, raise safety stock to 1.5× = 30 → new reorder point = (20×2)+30 = 70.

Copy-paste AI prompt (use in ChatGPT or a forecasting tool)

“You are an inventory analyst. I will provide a CSV with columns: date (YYYY-MM-DD), SKU, units_sold, promotion_flag. For each SKU, generate an 8-week weekly forecast with confidence intervals, flag anomalous weeks, and recommend a reorder point given lead_time_weeks and safety_stock_rule (default = 1 week or X×SD). Output as CSV: SKU, avg_weekly_sales, sd_weekly_sales, forecast_week_1..8, safety_stock, reorder_point, anomaly_notes.”

Metrics to track

• Forecast accuracy (MAPE %) weekly.
• Stockouts per SKU per month.
• Weeks of inventory and holding cost %.

Common mistakes & fixes

• Mistake: Using raw promotional spikes. Fix: Tag and exclude or model separately.
• Mistake: One-off reorder changes. Fix: Keep a log of manual overrides and why.
• Mistake: No review cadence. Fix: 15–30 minute weekly review, monthly recalibration.

1-week action plan

1. Day 1: Export top 20 SKU sales (weekly) for last 12 months and list lead times.
2. Day 2: Clean data, tag promos.
3. Day 3: Calculate avg weekly, SD, safety stock, and reorder points in a sheet.
4. Days 4–7: Start the pilot: place orders by new reorder points, track actual vs forecast each week.

Your move.

— Aaron

Good point — speed and clarity are the right priorities. Here’s a direct, outcome-first plan to get a professional personal website and portfolio live fast using AI.

Why this matters: a simple, optimized website converts curiosity into opportunities — clients, interviewers, and referral traffic. You don’t need perfect design; you need clear messaging, trust signals, and a path to contact.

Do / Do not (checklist)

• Do: pick one goal (hire me, consult, sell a service), one clear CTA, and one page that does that.
• Do: use AI for copy, images, and layout suggestions — then edit for your voice.
• Do not: try to show everything. Avoid long menus and multiple CTAs.
• Do not: skip basic proof (one case study, one testimonial, contact).

Worked example: Freelance consultant — single-page site with hero, services, 1 case study, testimonial, contact form. Goal: get 3 qualified leads in 30 days. Time to publish: 48 hours.

Step-by-step (what you’ll need, how to do it, what to expect)

1. Choose a builder (fast: Carrd, Webflow, Squarespace or a template-based host). Expect a drag-and-drop UI and a simple publishing flow.
2. Gather assets: one headshot, 1–3 portfolio images/PDFs, one short case study (problem → action → result), one testimonial, short bio, and your contact method.
3. Generate copy with AI: use the prompt below to create a hero headline, 3-sentence bio, service bullets, and a case-study blurb. Edit to match your tone.
4. Assemble the page: hero, services, case study, testimonial, contact form. Keep CTA above the fold and again after the case study.
5. Optimize: set a clear page title, meta description, and a single UTM-tagged contact link for tracking.
6. Publish & test: share to 10 known contacts for feedback, fix one quick issue, then announce on LinkedIn and email.

Copy-paste AI prompt (use as-is)

“Write a 6–10 word hero headline for a freelance marketing consultant who helps small B2B tech firms generate qualified leads. Then write a 20–30 word subheadline that explains the outcome and a 3-sentence bio that establishes credibility (years, specialties, measurable outcome). Create 3 short service bullets and a 50–70 word case-study blurb that includes metrics (e.g., % uplift, revenue, leads). Keep tone professional and confident, for an audience of non-technical executives over 40.”

Metrics to track

• Time to publish (goal: <72 hours)
• Unique visits/week
• Contact form submissions/week
• Lead quality: % of submissions that are qualified (target 30%+)
• Conversion rate: visits → contact (target 1–3% first month)

Common mistakes & fixes

• Too many CTAs — fix: pick one CTA and remove distractions.
• Vague outcomes — fix: add a single measurable result in the hero or case study.
• No tracking — fix: add UTM tags and a simple spreadsheet to log leads.

1-week action plan

1. Day 1: Choose builder, gather assets, run AI prompt for copy.
2. Day 2: Build layout, add content, upload images, set CTAs.
3. Day 3: Publish draft, share with 10 contacts for feedback.
4. Day 4: Iterate on feedback, set tracking (UTMs), prepare announcement.
5. Days 5–7: Launch announcement, collect leads, log results, adjust copy/CTA based on first data.

What to expect: a functional, credible site in 48–72 hours; measurable leads in 7–30 days if you share it to networks and use one targeted CTA.

Your move.

Hook: You can automate 70% of your monthly board and stakeholder reports in two cycles without losing control. The play: delta-first narrative, tight thresholds, and three simple controls that prevent errors.

The problem: Reports bloat, tone varies, and manual assembly risks wrong numbers. That erodes board confidence and wastes leadership time.

Why it matters: A repeatable, exception-led pack cuts time-to-draft from hours to minutes, reduces follow-up questions, and lets you focus the discussion on decisions, not recounting data.

Lesson learned (keep it simple): Lock your “truth map,” write only to material changes, and reuse approved language blocks. AI drafts fast; your reviewer certifies the message.

Do / Don’t checklist

• Do set a materiality threshold (e.g., ±3% or ≥$25k) and only narrate exceptions.
• Do snapshot each month’s KPIs into a dedicated tab so numbers never shift after sign-off.
• Do name KPI ranges (Revenue_MTD, Churn_Rate) and ask AI to echo source names for quick verification.
• Do keep a “narrative block library” (approved one-liners for common causes/risks/actions) to standardize tone.
• Don’t let AI guess causes. If unknown, label it and suggest one test to confirm.
• Don’t mix units. Standardize $, %, and months in the prompt and run an auditor pass.
• Don’t automate every section at once. Start with the executive summary, then KPI highlights.

Insider template: the Exception Ledger

• Create a short table in your snapshot: KPI, Last_Month, This_Month, Delta, Status (R/A/G via threshold), Cause (pick list: Pricing, Pipeline, Seasonality, Ops, Unknown), Risk, Next_Action, Source_Name.
• Feed only this ledger to the AI. It forces a clean, delta-first story and makes audits painless.

What you’ll need

• One trusted sheet or BI export with named ranges for 5–10 KPIs.
• A one-page template: Executive Summary (120 words), KPI Highlights (5 bullets), Risks, Actions/Asks, Appendix.
• An AI assistant for drafting and a reviewer checklist for sign-off.

Step-by-step

1. Define thresholds: e.g., call out if change ≥3% or ≥$25k. Map R/A/G: Red >5% adverse, Amber 3–5%, Green otherwise.
2. Snapshot: copy KPIs into “Snapshot_YYYY-MM” with named ranges and your Exception Ledger.
3. Narrative blocks: write 6–10 approved one-liners (e.g., “Lower churn driven by improved onboarding completion (+8 pts).”)
4. Draft: use the prompt below with this month, last month, thresholds, narrative blocks, and any context notes (campaigns, outages).
5. Audit: run the auditor prompt to flag math, unit, or threshold errors before human review.
6. Review: confirm numbers, tone, and the single clear board ask. Export “Board_Report_YYYY-MM_v1.0.pdf”. Log edits.

Copy-paste prompt: Delta-first board pack draft

“You are a company secretary preparing a delta-first monthly board summary. Use only the data provided. If a cause is unknown, write ‘unknown’ and propose one test to validate. Echo each KPI’s [Source_Name].

Inputs: This_Month & Last_Month values for: Revenue_MTD [$], Gross_Margin [%], Churn_Rate [%], Cash_Runway [months], SQLs [#], NPS [score]. Thresholds: material_change = 3% or $25k; Red >5% adverse; Amber 3–5% adverse; Green otherwise. Context: {paste short context}. Narrative_Blocks: {paste 4–8 approved one-liners}. Sources: {KPI → [Source_Name]}.

Produce exactly: 1) Executive summary (max 120 words, delta-first). 2) KPI highlights (5 bullets): Status (R/A/G), KPI name with delta, cause (or ‘unknown’), risk (if any), next action, and source echo in brackets. 3) One clear board ask (decision or resource). Tone: calm, precise, non-technical.”

Quality guardrail prompt: Auditor pass

“Act as a compliance reviewer. Compare the draft text to the KPI inputs. Flag math errors, unit mismatches, threshold mislabels (R/A/G), and any claim with no stated cause/test. Output a bullet list of fixes only; do not rewrite.”

Worked example (what good looks like)

• Inputs: Revenue $1.28M (was $1.20M), Churn 2.1% (was 2.3%), Runway 8.7m (was 9.0m), SQLs 410 (was 360), NPS 46 (was 44). Threshold 3% or $25k.
• Executive summary (sample): Revenue rose $80k (+6.7%) on stronger enterprise closes; churn improved 20 bps, extending ARR stability. SQLs increased 14% from the Q3 campaign; NPS ticked up to 46. Cash runway dipped to 8.7 months due to hiring and annual prepaids. Risks are limited; focus is sustaining top-of-funnel and restoring runway to 9+ months. Ask: approve reallocating $30k to paid search and pausing two non-critical hires.
• KPI highlights (sample):
  • G — Revenue +$80k (+6.7%); cause: enterprise deals; risk: none near-term; next: replicate playbook in mid-market [Revenue_MTD].
  • G — Churn -0.2 pts; cause: onboarding completion +8 pts; risk: unknown durability; next: cohort test Q4 [Churn_Rate].
  • A — Cash runway -0.3 months; cause: hiring + prepaids; risk: covenant buffer; next: freeze 2 roles [Runway_Months].
  • G — SQLs +50 (+13.9%); cause: Q3 campaign; risk: lead quality; next: MQL-to-SQL audit [SQLs].
  • G — NPS +2; cause: faster support SLAs; risk: none; next: extend to weekends [NPS].

Metrics to track

• Time to first draft (target: under 20 minutes).
• Reviewer edit time (target: -50% by month two).
• Error rate (numerical mismatches per report; target: zero post-audit).
• Board follow-ups (clarification questions per meeting; target: -30% in two months).
• Decision cycle time (days from report to decision on the “ask”).

Common mistakes and fixes

• Speculation in causes: force “unknown + 1 test” in the prompt; reviewer confirms or replaces.
• Version chaos: adopt “YYYY-MM_v1.0” and only increment after approval; keep a simple changelog.
• Metric drift: freeze monthly snapshots and add a “Methodology Changes” note for any definition updates.
• Unit confusion: specify units in the prompt and run the auditor pass every month.
• Bloat: enforce thresholds; move non-material items to the appendix automatically.

One-week action plan

1. Day 1: Pick 5 KPIs and thresholds; create Snapshot_YYYY-MM with named ranges.
2. Day 2: Build the Exception Ledger and the one-page template.
3. Day 3: Draft 6–10 narrative blocks (approved one-liners).
4. Day 4: Run the delta-first prompt with this and last month’s numbers.
5. Day 5: Run the auditor pass; reviewer edits and approves v1.0.
6. Day 6: Generate the stakeholder variant from the approved board summary.
7. Day 7: Log metrics (time, edits, questions) and schedule next month’s snapshot and draft.

Expectation set: Month one saves heavy drafting time; month two adds reliability as your template and blocks stabilize. After two clean runs, extend to more KPIs and automate the snapshot export.

Quick win (under 5 minutes): Paste this one-paragraph policy at the top of your syllabus and you’ve started—”Students may use AI tools for brainstorming and drafting with instructor permission. All AI-generated content must be identified and accompanied by a short reflection on how the tool was used. Use of AI to produce assessed work without disclosure is academic dishonesty.”

Good starter—this thread needs concrete, measurable steps. Below is a practical playbook you can implement this week.

The problem: Teachers often ban or ignore AI because there’s no clear, consistent classroom policy. That creates confusion, inconsistent grading, privacy risk and lost learning opportunities.

Why this matters: Clear policy protects student data, keeps assessments fair, and turns AI from a cheating risk into a learning tool. It also creates measurable outcomes you can improve.

Lesson from practice: Policies that are short, specific, and paired with simple checks (disclosure + reflection) get adoption. Complex legalese does not.

1. Define scope & objective
   • What you’ll need: current syllabus, list of tools students use.
   • How: State whether AI is allowed for brainstorming, drafting, editing, or not for final submissions.
   • Expect: Immediate clarity for students and fewer disputes.
2. Stakeholder & consent check
   • Need: Brief note to parents/admin explaining benefits and safeguards.
   • How: One-paragraph email; collect concerns.
   • Expect: Faster approval from leadership.
3. Acceptable uses & examples
   • Need: 3 positive examples and 3 prohibited ones.
   • How: Put these in the syllabus and review on day one.
   • Expect: Fewer gray-area incidents.
4. Privacy & data rules
   • Need: List of banned data (student PII, assessments).
   • How: Require anonymization and no uploading of tests.
   • Expect: Lower risk of data exposure.
5. Assessment & attribution
   • Need: Disclosure form + brief reflection with submissions.
   • How: Use a checkbox in LMS or a one-paragraph statement.
   • Expect: Easier grading and academic integrity enforcement.
6. Review cadence
   • Need: Schedule to revisit policy each term.
   • How: Set calendar reminders and collect metrics below.
   • Expect: Policy that stays relevant as tools change.

Metrics to track

• Adoption rate: % of classes that include the policy in syllabus.
• Disclosure compliance: % of submissions with AI disclosure/reflection.
• Academic incidents: number of suspected misuse cases per term.
• Student outcomes: average grade change on assessed tasks using AI.
• Teacher confidence: quick survey (1–5) after each term.

Common mistakes & fixes

• Mistake: Policy too long or vague. Fix: Reduce to 3–5 clear rules and examples.
• Mistake: No enforcement. Fix: Add a simple verification step (disclosure box).
• Mistake: Ignoring privacy. Fix: Ban uploading of student PII and require local/approved tools.

1-week action plan

1. Day 1: Add the one-paragraph policy to your syllabus and send a short note to parents/admin.
2. Day 3: Create three examples of allowed/prohibited uses and post them in class.
3. Day 5: Add a disclosure/reflection checkbox to the next assignment in your LMS.
4. Day 7: Run a 5-minute student poll on understanding and adjust wording.

AI prompt you can copy-paste

“Create a one-page classroom AI use policy for high school students that includes: purpose, allowed uses, prohibited uses, data privacy rules, a 2-sentence disclosure students must include with AI-assisted submissions, three examples of allowed use and three examples of misuse, and a short teacher checklist for enforcement.”

Your move.

Quick win (under 5 minutes): open last month’s board report, pick three KPIs (revenue, churn, runway), and paste them into an AI assistant with this instruction: “Draft a 2-sentence executive summary and 3 bullets explaining trend, likely cause, and one recommended action.” Read and tweak one sentence — you already have a usable start.

The problem: board and stakeholder reports take too long, vary in tone, and risk factual errors when hand-assembled every month.

Why this matters: consistent, fast reports mean clearer decisions, fewer follow-up questions, and lower overhead for your leadership team. Get the draft automated; keep a human to certify accuracy.

Experience-driven lesson: start with the executive summary and one trusted data source. Automating full reports before you’ve proven the summary flow creates governance gaps. Proven flow: single source of truth → template → AI draft → human sign-off.

What you’ll need

• A single trusted data source (Google Sheet, Excel or BI export).
• A one-page template (exact lines for each KPI and chart placeholders).
• An AI drafting tool (chat assistant or simple automation).
• A named reviewer for final sign-off.

Step-by-step (do this in order)

1. Map inputs: document where each KPI comes from (sheet name and cell or dashboard field).
2. Automate extraction: schedule a monthly export or connect the sheet so values populate the template automatically.
3. Auto-populate visuals: link charts to the template so they refresh with new numbers.
4. Generate draft: feed the numbers and a short context note to the AI to create the exec summary and 3–5 bullets.
5. Human review: the reviewer verifies numbers, edits tone, adds risks/actions, and approves the PDF/email.
6. Distribute & log: send the report and keep a changelog of edits for auditability.

Copy-paste AI prompt (use as-is)

“You are a seasoned company secretary. Input: revenue = $1.2M (up 6% MoM); churn = 2.3% (down 0.4%); cash runway = 9 months (no change). Produce: 1) a 2-sentence executive summary that states trend and one likely cause, 2) three concise bullets: one operational implication, one risk, one recommended action. Cite the source field for each KPI in brackets (e.g., [Sheet:KPI!B3]). Be factual, avoid speculation, and keep language non-technical and confident.”

Metrics to track

• Time to first draft (target: <20 minutes).
• Reviewer edit time (target: reduce by 50% in 60 days).
• Error rate (mismatched numbers in published reports).
• Board follow-ups (number of clarification questions after each report).

Common mistakes & fixes

• Mapping errors: run a one-month dry run and reconcile line-by-line. Fix: lock cell ranges and add a checksum row.
• AI hallucinations: require source fields be echoed in the draft. Fix: add “cite source field” to the prompt and enforce reviewer check.
• Over-automation: don’t automate commentary until summaries are stable. Fix: automate the exec summary first.

1-week action plan

1. Day 1: Pick 3 KPIs and build the one-page template.
2. Day 2: Map inputs and document source fields.
3. Day 3: Set up automated export/refresh into the template.
4. Day 4: Run the AI prompt and produce the first draft.
5. Day 5: Reviewer verifies, edits and finalizes the report.
6. Day 6–7: Capture time saved and feedback; adjust prompt and template.

Your move.

Agree: your “title + DOI” check is the fastest filter. Keep that habit. Now let’s drive hallucinations toward zero with a tight, repeatable system and visible KPIs.

Hook: Make AI work like a junior analyst under audit — every claim has a source, a quote, and a confidence score.

Problem in one line: AI writes fluent prose that blends true studies with plausible fiction. Reviewers notice. You pay for it later.

Why it matters: One fabricated citation can trigger reviewer distrust, delays, and rewrites. A verification-first workflow preserves credibility and speeds acceptance.

Lesson learned: Use “source-locked mode.” Feed the model the abstracts or excerpts you trust, and force it to quote only from those. Treat everything else as hypothesis until verified.

• Do: Require verbatim quotes, DOI, and a confidence reason for every claim.
• Do: Verify two details (title + DOI or title + authors + year) in a database.
• Do: Keep a simple claims register (ID, claim, source, quote, DOI, status).
• Do: Run a second pass where the AI acts as a citation auditor on your draft.
• Don’t: Accept paraphrases as proof; you need the exact sentence from the paper.
• Don’t: Mix preprints and peer-reviewed sources without clearly labeling.
• Don’t: Ask broad questions; specify population, timeframe, outcomes, and study type.

What you’ll need

• AI chat assistant.
• Access to PubMed, Google Scholar, Scopus, or your institutional library.
• A reference manager or a simple spreadsheet for your claims register.
• 10–20 minutes per major claim for verification.

Step-by-step: zero-hallucination workflow

1. Frame tightly. Define the claim scope (population, intervention/exposure, comparator, outcome, timeframe). Note acceptable evidence types.
2. Run strict sourcing. Use the primary prompt below. Extract 2–3 claims you intend to keep. Paste abstracts or key excerpts when you have them and tell the AI to use those only.
3. Verify independently. In PubMed/Google Scholar, confirm at least two details (title + DOI or title + authors + year). If either fails, mark the claim “Unverified.”
4. Quote or cut. Store the exact one-sentence quote and DOI in your claims register. If no verifiable quote exists, remove the claim or label as speculative.
5. Draft from evidence. Write your paragraph using only verified claims. Keep in-text references aligned to your register IDs.
6. Audit pass. Run the auditor prompt on your draft and references. Fix any flagged items before submission.

Copy-paste AI prompt — Strict Verify

“You are my research verifier. Topic: [insert topic]. Produce 3–5 bullet claims. For each claim, return only peer‑reviewed sources. For each source include: full citation (authors, year, journal), article title, DOI (required if available), one verbatim one‑sentence quote that directly supports the claim, and a confidence level (high/medium/low) with a one‑line reason. If you are unsure or cannot find a source, say ‘I don’t know’ and list the exact search queries I should run in PubMed or Google Scholar. Do not invent citations. If I paste abstracts/excerpts, only cite from those and label them [S1], [S2], etc., and map each quote to the claim.”

Auditor prompt — second pass

“Act as a citation auditor. I will provide a draft paragraph and a list of references with DOIs. For each sentence, check whether at least one provided reference contains a verbatim quote that supports it. Flag sentences with no direct support or mismatched claims. Output a list of flagged sentences, the missing evidence, and the exact search terms to verify.”

Worked example (process over content)

1. Claim draft: “Moderate aerobic exercise improves insulin sensitivity in adults over 50 within 12 weeks.”
2. Run Strict Verify on that claim. Get 2–3 studies with DOI and verbatim quotes.
3. Open PubMed/Google Scholar; search the title or DOI. If both title and DOI match, copy the DOI and the exact quote into your claims register.
4. If a source lacks a DOI or the quote doesn’t appear in the paper, mark Unverified and swap in the next candidate source.
5. Write the paragraph using only the verified claim(s); keep the quote in your notes for reviewer queries.
6. Run the Auditor prompt on the paragraph + references; resolve any flags.

Metrics to track (make results visible)

• % of AI-cited claims verified on first pass — target 90%+ after week 2.
• Average verification time per claim — target <15 minutes.
• Title–DOI mismatch rate — target <2%.
• Reviewer reference objections on first submission — target zero.

Common mistakes & fast fixes

• Mistake: Relying on paraphrases. Fix: Store one exact sentence from the paper for each claim.
• Mistake: Over-broad prompts. Fix: Specify population, timeframe, and outcome.
• Mistake: Mixing preprints and peer-reviewed papers. Fix: Label preprints clearly or exclude them.
• Mistake: One-pass drafting. Fix: Always run the auditor pass before submission.

1‑week action plan

1. Day 1: Create a claims register (columns: ID, claim, source, quote, DOI, confidence, status).
2. Day 2: Select one section of your paper. Run Strict Verify for 3–5 claims. Verify title + DOI for each.
3. Day 3: Draft the section using only verified claims; store quotes.
4. Day 4: Run the Auditor prompt on the draft + references. Fix flags.
5. Day 5: Repeat for the next section; track your verification time.
6. Day 6: Normalize references in your manager; ensure DOIs are consistent.
7. Day 7: Quick pre-submission audit: sample 5 claims, re‑verify title + DOI. Update KPIs.

What to expect: Slightly more time up front, major reductions in citation errors, smoother peer review, and confidence you can defend every sentence.

Your move.

Smart addition: Your Naturalness Stack (Reader Mirror + 1–1–1 + Reply Magnet) is the right foundation. I’ll layer on a control system that turns those drafts into measurable results fast.

Try this now (5 minutes): Open your last 10 sent emails or replies from customers. Paste them into AI with the prompt below to extract a Voice Bank. Then regenerate Email 1’s subject/preview using those exact phrases. Expect tighter opens within your next send.

Copy-paste prompt — Voice Bank extractor

From the text below, extract: 1) the 12 most common reader phrases, 2) 5 pains, 3) 5 desired outcomes, 4) 5 words to avoid (sound salesy). Format as four bullet lists. Text: [paste 5–10 short customer emails or replies here]

The problem: Great outlines still underperform without a feedback loop. Most sequences stop at “sounds natural” and never convert because subject lines aren’t tested, reply asks are mistimed, and CTAs don’t ladder toward a single offer.

Why it matters: Small, weekly tweaks move core KPIs — open rate, click/reply rate, and booked calls. A 3–5% lift at each stage compounds into real revenue without rewriting everything.

Field lesson: Sequences that place reply CTAs in Emails 1, 3, and 5, use reader phrases in subjects, and show one small proof by Email 3 consistently beat “polite but generic” flows. Keep the message short; make the next step obvious.

Step-by-step: the Nurture Control System

1. Inputs (10 minutes): Audience line, single goal, 3 micro-proofs (short quote, stat, mini-story), and your Voice Bank phrases (10–12).
2. Generate (20 minutes): Use the assembler prompt below. Ask for 3 variants of Email 1 and Email 3. Pick the clearest.
3. Calibrate (15 minutes): Apply the 1–1–1 rule. Convert half of “we/I” to “you.” End Emails 1, 3, 5 with a one-line question.
4. CTA Ladder (10 minutes): Map actions: E1 reply, E2 read tip, E3 reply, E4 download/checklist, E5 reply, E6 soft book, E7 deadline. One link max per email.
5. Subject/Preview workshop (10 minutes): Create 12 pairs for Email 1 using Voice Bank phrases. Keep subjects 6–9 words, previews 8–12 words.
6. QA + Load (10 minutes): Insert [FirstName]. Mobile check. Plain language. No hype words. Test links. Schedule 2–4 day gaps.
7. Pilot (5 minutes): Send to 100–500 subscribers. Mark baseline KPIs for two weeks.
8. Iterate weekly (15–30 minutes): Change one variable only (subject, CTA wording, or send time). Track trend lines, not single sends.

Copy-paste prompt — 7-email assembler (natural + conversion)

Write a 7-email nurture sequence for [Audience: one line, include one pain]. Goal: [single action, e.g., book a 20-minute consult]. Voice: warm, clear, human, short sentences, no hype. Use these reader phrases: [paste 10–12 from Voice Bank]. Include first-name token [FirstName] where natural. Length: 3–5 short sentences per email (90–120 words max). For each email provide: 1) subject (6–9 words), 2) preview (8–12 words), 3) body, 4) single CTA. Purposes: 1 welcome/expectations + reply question, 2 quick tip with tiny checklist, 3 short proof or 90-word story + reply question, 4 how-to or checklist + one link, 5 address main objection + reply question, 6 soft offer (what happens, how long, zero risk), 7 reminder with real deadline and easy opt-out. Keep paragraphs short for mobile.

What good looks like: Subjects echo reader phrases, bodies stay under 120 words, three emails invite replies, and every email advances one step toward the offer. Expect a friendly, grounded tone you can read aloud in under 20 seconds.

Metrics that matter (with simple targets)

• Open rate: aim 25–40%. If under 25%, fix subject/preview using Voice Bank.
• Primary action rate (click or reply): 3–10% per email. If under 3%, tighten the CTA to one specific next step.
• Reply rate (Emails 1/3/5): 1–4%. If flat, make the question smaller and easier to answer.
• Unsubscribes: keep under 0.5% per send. If higher, increase relevance or widen gaps.
• Booked calls from sequence: track weekly; aim for steady growth, not spikes.

Mistakes that kill performance (and fixes)

• Vague CTAs: replace “learn more” with a precise ask (“Book a 20-minute call”).
• Walls of text: cap to 3–5 short sentences; add a line break after sentence two.
• Proof too late: move a micro-proof into Email 3 (even a one-line outcome).
• No reply friction: ask a one-line question that’s easy to answer (“What’s the one thing blocking you this week?”).
• Mixed goals: one goal only. Remove secondary links and side offers.

1-week action plan

1. Day 1: Build your Voice Bank (5 minutes) and define one goal.
2. Day 2: Run the 7-email assembler. Generate 12 subject/preview pairs for Email 1.
3. Day 3: Edit with 1–1–1. Add reply questions to Emails 1, 3, 5.
4. Day 4: Load, insert [FirstName], mobile test. Fix line length and link placement.
5. Day 5: Pilot to 100–500 subscribers. Note baseline KPIs.
6. Day 6: Review: underperformers get new subject/preview; keep bodies stable.
7. Day 7: Adjust one variable only. Set next week’s test (subject vs CTA).

Bonus prompt — Subject/preview sprint (for Email 1)

Generate 12 subject/preview pairs that use these reader phrases: [paste 6–8]. Constraints: subjects 6–9 words, previews 8–12 words, no hype, plain language, curiosity without clickbait. Output as a numbered list. Mark 3 safest options for an older, professional audience.

Bottom line: Your outline is solid. Add the Voice Bank, a CTA Ladder, and weekly single-variable tests. That’s how you turn “sounds natural” into booked calls and revenue.

Your move.