ChatGPT prompts for AI automation agencies are completely different from the "100 beginner prompts" lists you've seen.
The beginner lists help you write a blog post.
These prompts help you build pipelines that pay £1,500-£5,000/month per client.
This page is 30 prompts I actually use in production with agency clients inside AI Profit Boardroom.
They're battle-tested.
They go into real n8n flows that run in real businesses every day.
🔥 Want my full agency prompt library (200+ prompts)? AI Profit Boardroom has the n8n integration prompts, the API prompts, the agency pitch decks, and weekly coaching — $69/mo. → Join here
How These Prompts Are Different
These prompts assume you're running them via API inside n8n, Make, or Zapier.
Not chat-window prompts.
System message + user message format, structured output expected, JSON-friendly.
Drop them straight into your pipeline nodes.
30 ChatGPT Prompts For Automation Pipelines
Pipeline 1: Cold Email Personalisation (5 prompts)
1 — First-line generator from LinkedIn
System: "You are a cold email writer. Output ONE personalised first line for a cold email based on a LinkedIn profile. UK English. No clichés. Max 20 words. Return JSON: {first_line: '...'}."
User: "[LinkedIn profile JSON]"
2 — Subject line generator
System: "Output 5 cold email subject lines for outreach to a [JOB TITLE] at a [COMPANY TYPE]. Each under 6 words. Return JSON: {subjects: ['...']}."
3 — Email body
System: "Write a 4-sentence cold email body offering [SERVICE] at £[PRICE]. UK English. Friendly. Goal: book a 15-min call. Return JSON: {body: '...'}."
4 — Reply classifier
System: "Classify this email reply as: interested, not_interested, more_info, out_of_office, auto_reply. Return JSON: {category: '...', reasoning: '...'}."
5 — Auto follow-up
System: "Write a 2-sentence value-first follow-up to my cold email. Adds insight, no pressure. Return JSON: {follow_up: '...'}."
Pipeline 2: Lead Scoring + Routing (5 prompts)
6 — Lead intent scorer
System: "Score this inbound lead 1-10 on buying intent based on the form fields. Output JSON: {score: int, reasoning: '...', recommended_route: 'sales|nurture|spam'}."
User: "[Form data JSON]"
7 — Lead enrichment summary
System: "Summarise this lead's company in 2 sentences using these data points: [DATA]. Then suggest 3 likely pain points. Return JSON: {summary: '...', pain_points: ['...']}."
8 — Routing rules
System: "Given this lead score and company size, recommend the next action: 1) book demo, 2) trial offer, 3) nurture email, 4) discard. Return JSON: {action: '...', reasoning: '...'}."
9 — Personalised demo invite
System: "Write a personalised demo invite based on this lead's pain points. UK English. Under 100 words. Return JSON: {invite: '...'}."
10 — Slack alert template
System: "Format this lead as a Slack alert for the sales team. Include: name, company, score, pain points, next action. Return JSON: {slack_message: '...'}."
Pipeline 3: Content Engine (5 prompts)
11 — Content brief from keyword
System: "Given the keyword [KEYWORD], output a 600-word content brief. Include: target intent, outline (H2s + H3s), 5 key questions to answer, 3 stats to find. Return JSON: {brief: '...'}."
12 — SERP gap analysis
System: "Analyse these top 10 SERP results for [KEYWORD] and output: 3 angles competitors missed, 3 questions nobody answered, 1 unique POV to lead with. Return JSON: {gaps: [...], angle: '...'}."
13 — Long-form draft
System: "Write a 3,000-word blog post on [TOPIC] using the brief: [BRIEF]. UK English. Conversational. Add personal opinion every 500 words. Return JSON: {draft: '...'}."
14 — Meta description
System: "Write 5 meta descriptions under 155 chars for the article on [TOPIC]. Include keyword [KEYWORD]. Return JSON: {descriptions: ['...']}."
15 — Internal linking suggester
System: "Given this draft and these existing posts: [PASTE], suggest 5 internal links to add. Return JSON: {links: [{anchor: '...', url: '...', insert_after: '...'}]}."
Pipeline 4: Customer Support Triage (5 prompts)
16 — Ticket classifier
System: "Classify this support ticket as: billing, bug, feature_request, how_to, account, urgent. Return JSON: {category: '...', priority: 'low|medium|high|critical'}."
17 — RAG-powered reply draft
System: "Using ONLY this knowledge base context: [KB CONTEXT], draft a reply to the user's question. If KB doesn't cover it, return 'escalate_to_human: true'. Return JSON: {reply: '...', confidence: 1-10}."
18 — Sentiment scorer
System: "Score this support message on sentiment 1-10. 1 = furious. 10 = delighted. Return JSON: {sentiment: int, urgency_modifier: '...'}."
19 — Escalation summary
System: "Summarise this support thread for a human agent in 3 bullets. Include: customer issue, what we've tried, recommended next action. Return JSON: {summary: '...'}."
20 — Customer follow-up sequence
System: "Generate a 2-touch follow-up sequence for a customer who hasn't replied in 48 hours. Touch 1 at day 3. Touch 2 at day 7. Return JSON: {touch_1: '...', touch_2: '...'}."
Pipeline 5: Reporting Automation (5 prompts)
21 — Monthly performance summary
System: "Write a 200-word client-facing monthly summary using these metrics: [METRICS JSON]. UK English. Plain language. Highlight wins, flag concerns. Return JSON: {summary: '...'}."
22 — Win / loss callouts
System: "From these metrics: [METRICS], identify 3 wins to celebrate and 3 concerns to flag. Return JSON: {wins: ['...'], concerns: ['...']}."
23 — Next-month recommendations
System: "Based on this month's performance, recommend 3 specific actions for next month. Return JSON: {recommendations: ['...']}."
24 — Email cover note
System: "Write a 4-sentence cover email for this monthly report. Friendly. Mentions one specific win. Return JSON: {cover_email: '...'}."
25 — Executive 1-line summary
System: "Reduce this monthly report to 1 sentence for the CEO. Return JSON: {one_liner: '...'}."
Pipeline 6: Agency Pitching (5 prompts)
26 — Niche pitch deck outline
System: "Build a 5-slide pitch deck for [PIPELINE] aimed at [NICHE]. Include: problem, solution, deliverables, pricing, proof. Return JSON: {slides: [{slide: 1, title: '...', content: '...'}]}."
27 — Discovery call qualifying questions
System: "Generate 8 qualifying questions for a discovery call about [PIPELINE] with a [NICHE] business. Return JSON: {questions: ['...']}."
28 — Proposal generator
System: "Write a proposal for [PIPELINE] at £[PRICE]/month. Include: scope, deliverables, timeline, what's NOT included, payment terms. Return JSON: {proposal: '...'}."
29 — Objection handler library
System: "List the 10 most common objections a [NICHE] business raises when buying [PIPELINE]. Give a 1-paragraph response to each. Return JSON: {objections: [{objection: '...', response: '...'}]}."
30 — Closing email
System: "Write a closing email after a discovery call. Recap what we agreed. Send the 50% deposit invoice. Confirm timeline. Return JSON: {email: '...'}."
How To Use These Prompts In n8n / Make / Zapier
- Drop the system prompt into the
systemfield of your AI node. - Pass dynamic data via the
userfield (Airtable, webhook, CRM, etc.). - Set
response_format: json_objectin the API call. - Parse the JSON in the next node.
- Route the parsed output to the right destination (email, Slack, CRM, etc.).
That's the entire pattern.
The Agency Pricing These Prompts Support
Cold email pipeline: £1,500-£4,000/mo per client.
Lead scoring pipeline: £800-£2,500/mo per client.
Content engine: £2,000-£5,000/mo per client.
Support triage: £2,000-£6,000/mo per client.
Reporting automation: £500-£2,000/mo per client.
10 clients across these pipelines = £15,000-£40,000/month recurring.
The Mistake Most Automation Beginners Make
They tweak prompts for 6 weeks instead of pitching them to clients.
Prompts are 90% there out of the box.
Land the client. Deploy the prompt. Iterate based on real production output.
🔥 Want my full agency prompt library (200+ prompts)? Inside AI Profit Boardroom you get the production-tested prompts, the n8n integration templates, the API patterns, the agency pitch decks, and weekly coaching — $69/mo. → Join now
Frequently Asked Questions
Do these prompts work via the chat window or only API? Both — but they're optimised for API use inside n8n/Make/Zapier. The JSON output format is what makes them production-grade.
Which prompts make money fastest? The cold email pipeline (1-5) and the lead scorer (6-10). Both close fast because the ROI is obvious to any sales team.
Do these work with Claude API? Yes. The structure is identical. Swap the model — keep the system prompt.
How much does the ChatGPT API cost per pipeline? £20-£200/month per client typically. Pass it through to the client at 2-3x markup.
Should I add temperature: 0 for production prompts?
For structured output (classification, scoring) — yes. For creative output (emails, drafts) — keep temperature 0.4-0.7.
Can I sell these prompts as their own product? No. Use them yourself to power agency services. The money is in the productised pipeline, not in the prompts.
