Five emails. Nothing you could not defend.

Every email in the sequence draws from a dossier assembled by a research prompt before the first send. Inputs required:

• AI feature observed. Specific feature name and what it does. From their product page or changelog.
• Recent public writing. CTO blog post, podcast, LinkedIn post, engineering blog. Last 90 days.
• Hiring signal. Open roles involving AI/ML/LLM. From their careers page.
• Tech stack hint. From their engineering blog or job description: do they use Anthropic? OpenAI? Cohere? Internal?
• Competitor or industry context. Something relevant happening in their space.

If the dossier has fewer than 3 of these, the prospect is dropped. The sequence is useless without specifics.

To make the emails concrete, every {{placeholder}} below resolves against this dossier.

To: elena@moorlink.example

From: hello@nexcur.ai

Subject: 60 days to Shipment Intelligence

Hi Elena,

Your March post about building Shipment Intelligence in 60 days caught my eye - specifically the line about deciding to wrap Claude rather than training from scratch.

We do product-development work for Series-A and B teams shipping similar features. Evals, cost modeling, prompt libraries. Usually the first fix is turning the prompt into something you can change without holding your breath.

One question while you are hiring a Staff ML Engineer: is the eval harness something the new hire is expected to build, or is it already in place?

- Leo

Why this works. The hook is a specific fact (her post, her 60-day timeline, her decision on Claude). The bridge is one sentence, not a value-prop avalanche. The ask is a question she can answer in one line.

Hook citation: Blog post 2026-03-04.

Expected reply rate: 8 to 12% on this ICP.

Subject: A cost-model template you might find useful

Hi Elena,

Following up on last week. Not pushing for a call - just sending something that might be useful given where you are.

We maintain a token-economics template for AI features: cost per action, per user, P90 sensitivity, caching ROI. It is the spreadsheet we run in week 1 of every engagement. Free, no email gate.

Link: nexcur.ai/samples/cost-model-template.html

The Shipment Intelligence scoring workflow fits neatly into Scenario B (document-style input, structured output, no caching because every shipment is unique). You will see whether margin is a concern or not in about 10 minutes.

- Leo

Why this works. Pure value, no ask. Provides a tool that applies to her specific product. The sample link proves our side of the pitch without asking her to click "book a demo".

Expected reply rate: 2 to 3% (additive to Email 1).

Subject: The eval harness shape we ship

Hi Elena,

One more follow-up because I think this one is the most useful.

We publish a sanitized example of the eval harness we build in engagements: golden dataset, scorers, pass-rate dashboard, regression tracking. It is the shape of thing most teams wish they had built before their first AI feature slipped quality.

Link: nexcur.ai/samples/eval-harness-example.html

For Shipment Intelligence, the scorers I would build are: (1) factual accuracy on the risk score, (2) no hallucinated route data, (3) structured output parses correctly. Happy to sketch what a golden set looks like for your shape of problem if useful.

- Leo

Why this works. The third email is where generic outbound dies. We land a specific artifact and a specific application to her product. The "happy to sketch" ask is still low-commitment.

Expected reply rate: 1 to 2% (additive).

Subject: Three ways we might help, ranked

Hi Elena,

If you are weighing whether a conversation is worth your time, here are the three concrete things we do, ranked by what I think fits Moorlink best.

1. Eval Sprint. 4 weeks, $25k. We build a production eval harness for Shipment Intelligence. Golden set, scorers, regression alerts. You keep all of it.

2. Cost Model + Prompt Library. 3 weeks, $18k. Token economics, caching strategy, versioned prompt library, runbook. Smaller bite if eval is not the priority.

3. Full Product Discovery. 6 weeks, $60k. Covers both of the above plus observability and fallback strategy. For teams planning a second or third AI feature.

All fixed price, no per-hour billing. 30-minute conversation before we quote.

- Leo

Why this works. Explicit scope and price removes the "is this going to waste my time" objection. CTOs respect directness. Fixed price signals no hourly-billing games.

Expected reply rate: 1%.

Subject: Closing the loop

Hi Elena,

Last note from me. Assuming the timing is not right, I will drop off here. No ghost-closing.

If the shape of problem comes up in the next 6 to 12 months - eval drift, cost surprises, prompt chaos - you know where to find us. Most of what we publish is at nexcur.ai/samples/ and stays useful regardless.

Wishing you and the team a clean Q2.

- Leo

Why this works. A clean exit outperforms a "breakup email" that tries to trigger FOMO. Reply rate on Email 5 is disproportionately high because the prospect reads it as confident, not desperate.

Expected reply rate: 2 to 3%.

What is different about this sequence

• Every email has a hook citation. You could audit the sequence and show why each email was sent to this specific prospect. No generic blasts.
• Emails 2 and 3 give before asking. The samples are useful whether or not Elena replies. They accumulate credibility.
• Email 4 names prices. Hiding prices earns a skip from a good CTO. Respecting their time means being direct.
• Email 5 is a real close. No "still the right person?" bait. The graceful exit produces the most replies of any email after the first.
• Three-week window. Any longer is pestering. Any shorter and the prospect has not had time to think.