AI Agent Workflow for Content Writing: Research → Write → Review in 5 Minutes

TL;DR: A three-agent content pipeline — Researcher (Gemini, free), Writer (Claude Sonnet, $0.08), Reviewer (Claude Haiku, $0.02) — produces blog posts, reports, and marketing copy for $0.10 per piece. This guide includes exact prompts, real output examples, and a 5-minute setup.

Most people use one AI chatbot for content writing. They paste a prompt, get a draft, then spend 30-60 minutes editing. The result is usually generic because the chatbot didn't research first.

A multi-agent content workflow splits writing into specialized phases. Each agent does one thing well. The result is content that's researched, well-written, and reviewed — produced automatically in under 5 minutes.

Here's the exact workflow.

The Three-Agent Content Pipeline

Why Three Agents?

Single Agent	Three-Agent Pipeline
Generic content	Research-backed content
No data or examples	Current data and specific examples
No quality check	Automatic review and polish
30-60 min editing	5 min review
$0 (subscription)	$0.10 (API)

Agent Roles

Agent	Model	Role	Cost
Researcher	Gemini 2.5 Pro	Gather data, trends, competitor angles	Free
Writer	Claude Sonnet	Create content from research	~$0.08
Reviewer	Claude Haiku	Check quality, accuracy, completeness	~$0.02

Total per piece: ~$0.10. Time: 3-5 minutes.

Setup (5 Minutes)

Step 1: Create an Account

Go to ivern.ai/signup. Free, no credit card.

Step 2: Add Your API Key

In Settings, add your Anthropic API key ($5 at console.anthropic.com). No markup on usage — BYOK model.

Step 3: Create a Content Squad

Click Create Squad → name it "Content Team". Add three agents:

Researcher (Gemini 2.5 Pro — free):

"You are a content researcher. Given a topic, provide: 1) Key statistics and data points with sources, 2) Current trends and developments, 3) Competitor content angles and gaps, 4) Unique insights and contrarian takes. Structure findings clearly for a writer to use."

Writer (Claude Sonnet):

"You are a professional content writer. Using the research provided, write [content type] about [topic]. Requirements: specific data from research, practical examples, clear structure with H2/H3 headings, engaging opening, actionable takeaways. Target length: [words]. Tone: [professional/conversational/technical]."

Reviewer (Claude Haiku):

"You are a content editor. Review the content for: 1) Factual accuracy against the research, 2) Grammar and readability, 3) Structure and flow, 4) Completeness — did it cover what was asked, 5) Engagement — is the opening compelling? Score each dimension 1-10 and provide specific improvement suggestions if any score below 8."

Step 4: Assign Your First Task

Create a pipeline task:

"Write a 1500-word blog post about email marketing strategies for SaaS companies with under 1000 subscribers. Include current statistics, 3-5 actionable strategies with examples, and a clear CTA."

The task flows: Researcher → Writer → Reviewer. Each output streams to the dashboard.

Real Output Example

Input Task

"Write a blog post about using AI agents for small business automation. Target: small business owners who are not technical. 1200 words. Include real cost examples and time savings."

Phase 1: Research Output (Gemini, free)

The Researcher produces structured findings:

Key statistics: 67% of small businesses plan to adopt AI in 2026 (Salesforce), AI automation saves small businesses 15+ hours/week on average (McKinsey)
Trends: Shift from chatbots to agent teams, BYOK pricing model reducing costs 90%, no-code tools making AI accessible
Competitor angles: Most content focuses on enterprise AI, gap in small-business-specific content
Unique insights: The real cost of AI agents for small business is $3-8/month, not the $20-50 subscriptions most people think

Phase 2: Writer Output (Claude Sonnet, $0.08)

The Writer takes the research and produces a 1200-word blog post with:

Engaging opening with a specific statistic
4 strategies with cost breakdowns
Real examples: "Sarah's bakery saves 12 hours/week on inventory management"
Proper H2/H3 structure
CTA at the end

Phase 3: Reviewer Output (Claude Haiku, $0.02)

The Reviewer scores and improves:

Dimension	Score	Notes
Accuracy	9/10	All statistics match research findings
Grammar	9/10	Clean, no issues
Structure	8/10	Good H2/H3 hierarchy, could add a summary section
Completeness	8/10	Covered all strategies, missing one cost example
Engagement	7/10	Opening is good but could be stronger — suggest starting with the $3-8 cost surprise

Final result: Ready-to-publish content in 4 minutes for $0.10.

Prompt Templates for Different Content Types

Blog Post

"Write a [length]-word blog post about [topic] for [audience]. Include: current statistics, practical examples, clear H2/H3 headings, and actionable takeaways. Tone: [professional/casual]. End with a clear CTA."

Email Newsletter

"Write a [length]-word email newsletter about [topic] for [audience]. Include: compelling subject line (3 options), engaging opening, 2-3 key points with data, and clear CTA. Tone: conversational but authoritative."

Social Media Batch

"Create 5 social media posts about [topic]. Each post should: stand alone, include a hook, provide value in the first line, and end with engagement question or CTA. Optimize for [platform: Twitter/LinkedIn/Instagram]."

Case Study

"Write a case study about [company/use case]. Structure: Problem → Solution → Results. Include specific metrics, quotes (realistic), and lessons learned. Length: [words]. Tone: professional."

Product Comparison

"Write a comparison of [Product A] vs [Product B] for [audience]. Include: feature comparison table, pricing breakdown, pros/cons of each, use case recommendations, and clear verdict. Length: [words]."

Cost Analysis

Per Piece Cost

Content Type	Avg Cost	Time
Blog post (1500 words)	$0.08-0.15	3-5 min
Email newsletter	$0.04-0.08	2-3 min
Social media batch (5 posts)	$0.03-0.06	1-2 min
Case study	$0.10-0.20	4-6 min
Product comparison	$0.08-0.15	3-5 min

Monthly Cost by Volume

Pieces/Month	Total Cost	Compare to Freelancer
4 (weekly blog)	~$0.50	$400-2000
12 (3x/week)	~$1.50	$1200-6000
20 (daily)	~$2.50	$2000-10000
30 (daily + social)	~$3.50	$3000-15000

The agent pipeline produces content at 1/1000th the cost of human writers, and 1/20th the cost of subscription tools like Jasper.

Quality: AI Pipeline vs Alternatives

Approach	Research	Writing Quality	Editing Needed	Cost/Piece
ChatGPT manual	None	Good	30-60 min	$0 (sub)
Jasper template	None	Moderate	15-30 min	$0 (sub)
Freelance writer	Manual	Varies	15-30 min	$50-500
AI Pipeline	Automatic	Very good	5 min	$0.10

The key differentiator is the automatic research phase. Content with real data and specific examples consistently outperforms generic AI output in search rankings and reader engagement.

Tips for Best Results

1. Be Specific in Your Task Description

Bad: "Write a blog post about marketing" Good: "Write a 1500-word blog post about content marketing strategies for B2B SaaS companies with $1-10M ARR. Include 3-5 specific strategies with examples from real companies, current statistics, and actionable implementation steps."

2. Let the Research Run First

Don't skip the research phase. The writer produces dramatically better content when it has real data to work with. Gemini's 1M token context means thorough research at zero cost.

3. Review the Reviewer's Feedback

The Reviewer's scores tell you where to focus your 5-minute edit. If engagement scores 7/10, spend your time on the opening paragraph. If accuracy is 10/10, skip fact-checking.

4. Build a Prompt Library

Save your best task descriptions as templates. The more specific your prompts, the better the output. After 10-20 pieces, you'll have templates that produce consistent quality.

5. Batch Similar Content

Create one research task, then route the findings to multiple writers:

Research "AI agents for business" → Writer 1 (blog post) + Writer 2 (email) + Writer 3 (social batch)

One research phase produces three different content formats.

Frequently Asked Questions

Is AI-generated content good for SEO?

Yes, when it's researched and reviewed. Google has stated AI content is acceptable if it provides value. The research phase is key — content with specific data, examples, and unique insights outperforms generic AI output in search.

How does this compare to hiring a writer?

The pipeline produces first drafts at $0.10/piece. A human writer adds strategy, brand voice, and domain expertise. Best practice: use the pipeline for drafts, have humans add the final polish for important content.

Can I customize the tone and style?

Yes. Modify the Writer's system prompt to match your brand voice. Include examples of your existing content as style references. The more specific the prompt, the more the output matches your voice.

What about plagiarism?

AI-generated content is original — it doesn't copy from sources. However, always review AI output and verify any specific claims or statistics before publishing.

How long should the content be?

For SEO: 1500-2500 words for pillar content, 800-1200 for supporting posts. The Writer agent follows whatever length you specify in the task description.

Get Started

Sign up free at ivern.ai/signup
Add your Anthropic API key ($5 minimum)
Create a Content Team squad (Researcher + Writer + Reviewer)
Assign your first content task
Get researched, written, and reviewed content in 5 minutes

Your first 15 tasks are free. That's 15 pieces of researched content at no cost.

Set up your content writing pipeline →