Verified against ChatGPT · 2026-07-21
Turn one verified research signal into a cold email that doesn't read like a mail-merge
Builds a cold-outreach email around exactly one verifiable trigger event with a specific banned-opener list, three genuinely different relevance angles, and an honest read on whether the signal was strong enough to use.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are drafting a cold outreach email for a B2B sales rep. The entire email is built around ONE verified, specific signal about the prospect or their company — never a generic compliment, never a guess dressed up as research, and never a fact so broadly true of the segment that it could apply to any of a hundred companies. PROSPECT Priya Menon, VP of Operations at Fenwick Logistics VERIFIED TRIGGER SIGNAL Posted on LinkedIn eight days ago about opening a third regional warehouse WHAT WE SELL route-planning software for multi-warehouse logistics teams WHY THE TRIGGER CONNECTS TO A PROBLEM WE SOLVE A third warehouse usually means route planning that worked for two locations starts breaking down almost immediately DESIRED NEXT STEP Worth a 15-minute call next week, or is this not the right time? STRUCTURE — four moves, each one sentence unless noted 1. OBSERVATION — reference Posted on LinkedIn eight days ago about opening a third regional warehouse specifically enough that the recipient could not mistake this for a mass send. If the sentence would still make sense with the name and company swapped out, it is not specific enough — rewrite it until it wouldn't. 2. RELEVANCE BRIDGE — connect the observation to a problem Fenwick Logistics likely has right now, using A third warehouse usually means route planning that worked for two locations starts breaking down almost immediately. Do not pitch in this sentence. 3. VALUE, ONE LINE — name what we do, framed as the answer to the problem just named, not a feature list and not a tagline. 4. ASK — Worth a 15-minute call next week, or is this not the right time?, phrased as a low-friction yes/no question a busy person could answer from their phone, never "let me know if you're interested" or "happy to share more whenever works." BANNED PATTERNS Do not write "I noticed you're the VP of Operations at Fenwick Logistics" or any variant of it — a job title lookup is not a signal, it is a mail-merge field wearing a signal's clothes. Do not write "I hope this email finds you well," "I wanted to reach out," "just following up" framing, or use an em dash anywhere in the body. No more than one question mark in the entire email. LENGTH AND FORMAT Under 90 words total, plain text, no bullet points, no bold, five lines maximum counting greeting and signoff. Write a subject line separately, under six words, that hints at the trigger without giving away the entire pitch — a subject that reads like the opener of a real note, not a campaign headline. VARIANTS Write three variants that differ in the RELEVANCE BRIDGE specifically — three genuinely different reasons the trigger matters to this buyer, not the same reasoning restated with synonyms swapped in. If you can only find one real angle, say so instead of manufacturing two weak ones just to hit the count. HONESTY CHECK After the three variants, state plainly whether Posted on LinkedIn eight days ago about opening a third regional warehouse is strong enough to carry a personalized email, or too generic or stale to use as the anchor — a signal that's true of most companies in the segment, or one that's more than roughly 60 days old, should be flagged as weak even though it technically fits the format above.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Naming and banning the exact generic opener — "I noticed you're the {{prospect_title}} at {{company_name}}" — is the load-bearing move, because that precise sentence pattern is what makes mail-merge personalization instantly recognizable to anyone who has received more than a handful of cold emails, and most prompts asking a model to "personalize" an email never rule it out by name, so the model reaches for it as its default safe move. The self-check embedded in step one — would this sentence still make sense with the name and company swapped out — gives the model a concrete falsifiable test for specificity instead of a vague instruction to "be specific," which models otherwise satisfy by inserting the company name into an otherwise generic sentence and calling that personalization. Capping the email at 90 words and forcing three variants that differ specifically in the relevance bridge, not the phrasing around it, stops two separate failure modes at once: a model padding the email with filler to sound more thorough, and a model generating three drafts that make the identical argument in different words and presenting that as meaningful variation. The closing instruction to call out a weak trigger signal — and specifically to flag a signal older than roughly 60 days or one true of most companies in the segment — matters because a model asked only to write the email will personalize convincingly around almost any signal it is handed, confident tone included, which is exactly how forced personalization built on a thin signal ends up reading worse to the recipient than an honest, ungimmicked email would have. There is also a compounding cost this prompt is designed to catch before it happens: a rep who sends five emails built on stale or generic signals and gets zero replies will often conclude that personalization itself doesn't work for their list, when the real lesson was that the specific signals chosen weren't good enough to justify the personalization technique — the honesty check exists to separate those two conclusions before the rep burns through the list.
What you get back
Priya — saw your post about the third warehouse going live next month. Most route-planning setups that work fine for two sites start falling apart once a third hub enters the mix, drivers end up covering overlapping zones nobody planned for. We built routing software specifically for that multi-warehouse handoff. Worth 15 minutes next week, or is this not the right time?
Verified against
ChatGPT GPT-5.1 · 2026-07-21
Claude Sonnet 4.5 · 2026-07-22
Changelog
- 2026-07-22 — Initial publish, verified against ChatGPT (GPT-5.1) and Claude (Sonnet 4.5).
Need this built into your business?
If a prompt isn't enough — Google Ads management, built and maintained for you — that's Scult's day job.
EXPLORE GOOGLE ADS MANAGEMENT
