Prompt Examples for Sales, Content, Outreach, and Universal Tasks

Sales Prompts with a Fully Filled Example

Sales prompts sell a specific outcome, not a feature. Five patterns cover the most common tasks: a system prompt for voice and quality, a full service-page draft, targeted revision, headline variations, social-proof generation. The showcase pattern S-4 shows how headline variations come from conversion angles instead of fixed counts.

S-4 · Headline Variations (compact example):

Generate headline variations across the conversion angles below · variety, not numerical fill · for the following offer:

Service: AI-driven predictive maintenance for mid-market manufacturing
Target: Plant managers at US manufacturers, 50-250 employees
Promise: 10 percent OEE lift in six months, no production stop for rollout

Variations per angle as the angle yields substance:
- Problem-focused (names the pain)
- Outcome-focused (names the specific result)
- Curiosity-based (raises a question without giving the answer)
- Direct/rational (clear value statement, no metaphor)

No exclamation points. No em-dashes. No clickbait. Every headline names an industry-specific detail.

Further sales patterns: S-1 universal system prompt for voice and quality rules, S-2 full service-page draft with 5 sections, S-3 targeted single-axis revision of one section, S-5 testimonial generation from verified numbers. Full templates in the splicelog Prompt Engineering Guide.

Content Prompts for Blog, SEO, and Research

Content prompts typically run in three steps: topic research first, long-form writing next, revision last. Six patterns cover the usual path. The showcase pattern C-3 shows a topic-research brief that activates web search and returns structured results.

C-3 · Topic Research (compact example):

Research the topic of embedded banking for B2B platforms.

Audience for follow-up articles: FinTech CTO or Head of Engineering at a Series A-C SaaS company evaluating providers.

Return in structured form:
- Three recent technical architecture decisions (source + URL)
- Three compliance aspects with US-market relevance (source + URL)
- Three common selection mistakes from practitioner reports (source + URL)
- Two comparison axes that decide real evaluations

Per finding: one sentence claim, one sentence justification, source link.
No marketing sources. No unsourced numbers.

Further content patterns: C-1 full SEO blog article with H1/H2 structure and meta description, C-2 content refresh from an existing article plus new research, C-4 LinkedIn post series from one article, C-5 section revision with one axis, C-6 white-paper outline from a topic brief. Full templates in the Prompt Engineering Guide.

Outreach Prompts for Cold Email and Follow-Up

Outreach prompts produce emails that do not read as AI-generated. Four patterns cover the sales funnel: cold email, follow-up after silence, internal preparation for objection handling, cover email for a proposal. The showcase pattern V-1 shows a cold email with a specific reason and a low-commitment next step.

V-1 · Cold Email (compact example):

Write a cold email to the operations leader of a US mid-market manufacturer, 80-150 employees.

Reason: recent press release on a plant expansion in the Southeast.
Our offer: AI-driven predictive maintenance, 10 percent OEE lift in six months.
Proof from own work: comparable mid-market shop, 11 percent OEE lift.

Constraints:
- First sentence: specific reason for the recipient, no generic opener
- Maximum 120 words
- Direct, peer-to-peer tone, no sales voice
- One concrete question as the next step, not a generic meeting request
- No attachments, no links except one reference URL
- No exclamation points, no em-dashes, no three-adjective chains

Further outreach patterns: V-2 follow-up after two weeks of silence with a new reason, V-3 internal prep for objection handling in the next call, V-4 cover email for proposal handoff. Full templates in the Prompt Engineering Guide.

Universal Prompts for Research and Brainstorm

Universal prompts work across categories: revising text, summarizing long content, building comparisons, generating research briefs, moderating structured brainstorms. The showcase pattern U-4 shows a research brief that runs before any larger content or strategy effort.

U-4 · Research Brief (compact example):

Create a research brief for the project below.

Project: market analysis of predictive-maintenance providers in US mid-market.
Purpose: make-or-buy decision for our platform.
Time horizon: two weeks.

Deliver:
- Five concrete research questions, one per decision axis
- Per question: two hypothesized answer paths plus one data point that would support each path
- Three source types that yield reliable answers
- Two source types that are unreliable (with reason)
- One anti-pattern: what would make the analysis worthless

No marketing sources as primary evidence. No LLM assumptions as fact.

Further universal patterns: U-1 text revision with one style axis, U-2 long-content summary with hierarchy, U-3 comparison table from multiple sources, U-5 structured brainstorm with four perspectives. Full templates in the Prompt Engineering Guide.

When Each Pattern Applies

The four categories overlap less than the names suggest. A decision heuristic by task type:

When the output sells: service page, headline, testimonial block, proposal email. Sales, with S-1 as the system prompt first, then the matching sub-pattern. S-1 covers voice, banned phrases, and market conventions and applies to a whole conversation, not to a single prompt.

When the output explains: blog article, white paper, LinkedIn series. Content, with topic research (C-3) before writing and revision (C-5) after. Running C-1 without a preceding C-3 produces a plausibly written article without verifiable sources and doubles the downstream fact-check effort.

When the output addresses: cold email, follow-up, proposal cover email. Outreach, with a specific reason tied to the recipient instead of a generic sales opener. Every outreach email needs a recipient reference that does not come from the company profile, but from a current data point about industry, market, or event.

When the output structures: research brief, summary, brainstorm, comparison. Universal, with no upstream system prompt required. U-4 typically runs before any larger project, U-2 after every long source, U-3 before every make-or-buy decision.

How These Examples Were Built

The patterns collected here are distilled from daily work and theoretically grounded in the splicelog Prompt Engineering Guide. Every prompt is fully filled for a concrete industry or a concrete reason. To build your own pattern, take the closest example as a template, replace the industry-specific slots, and check the result against the CRAFT framework (Context, Role, Action, Format, Target).

The order of patterns within a category is historical, not workflow-optimized. In practice, content typically runs C-3 before C-1 before C-5, sales runs S-1 as a system prompt first, then S-2 or S-4. Reading the patterns as a linear catalog overlooks the workflow prerequisites noted in each pattern. In the Guide, prerequisites are listed explicitly per pattern.

Deepening Into Adjacent Topics

Anyone using the patterns systematically soon meets three related practices: Prompt versioning protects reusable templates from output drift, hallucination reduction shows how structured prompts prevent faulty answers, and Custom GPTs anchor S-1-style system prompts persistently in a reusable configuration.

For more complex application scenarios, three further deepenings are useful: AI agents combine multiple patterns into a multi-step workflow, AI automation in business shows how templates flow into process tools, and the 2026 AI tool comparison helps with model choice per pattern type.

Prompt Examples · Next Step

The four showcase patterns are ready to use. For the full templates with slot system, see the splicelog Prompt Engineering Guide with compatibility picker and copy buttons. For the theoretical background, start with the CRAFT framework in the Guide.

The most common stumbling block in practice is not the pattern itself, but the temptation to relax structural constraints. Anyone cutting anti-patterns because the output is faster without them loses the pattern substance and gets average AI text. Anyone keeping the constraints and varying only the industry slots gets comparable results week after week.