architecture

Marketing AI Agents

AI agents for marketing: content creation, competitive intelligence, and campaign analysis. Includes off-the-shelf options, custom build approaches, and training examples.

Overview

Marketing has high content volume needs and data synthesis requirements. Focus areas: content creation, campaign optimization, competitive intelligence, and performance analysis.

Efficiency Potential: 7-9 hrs/week at Adoptive tier

Agent Summary

Agent Efficiency Target Complexity Priority
4-6 hrs β†’ 1-1.5 hrs (75%) Medium High
Continuous vs quarterly Medium-High High
3 hrs β†’ 30 min weekly Medium Medium

Content Brief & Draft Generator

Problem Statement

Content marketing requires constant output. Briefs take 1-2 hours, first drafts take 2-4 hours. Most content follows similar patterns with variable research.

Efficiency Target: Brief + first draft from 4-6 hrs β†’ 1-1.5 hrs (75% reduction)

Off-the-Shelf Options

Tool Capability Pricing
Jasper Long-form content, brand voice $49-125/mo
Writer Enterprise content, style guides Enterprise
Copy.ai Workflows, templates $49/mo
Claude/ChatGPT Flexible, requires prompting $20/mo

Training Example: Content Brief

Generate a comprehensive content brief.

CONTENT REQUEST:
- Topic: {{topic}}
- Format: {{format}} (blog post, whitepaper, case study, etc.)
- Target Length: {{word_count}}
- Target Audience: {{audience_persona}}
- Stage: {{buyer_journey_stage}} (awareness, consideration, decision)

BUSINESS CONTEXT:
- Campaign: {{campaign_name}}
- Goal: {{content_goal}} (traffic, leads, nurture, etc.)
- Key Message: {{key_message}}
- CTA: {{desired_cta}}

SEO REQUIREMENTS:
- Primary Keyword: {{primary_keyword}}
- Secondary Keywords: {{secondary_keywords}}
- Competitor Content: {{competitor_urls}}

GENERATE:

## Content Brief: {{topic}}

### Objective
[What this content should achieve]

### Target Reader
- Who: [specific persona]
- Their situation: [context]
- Their question: [what they're trying to answer]
- After reading: [what they should think/feel/do]

### Key Messages (priority order)
1. [Must communicate]
2. [Should communicate]
3. [Could communicate]

### Outline

**Hook/Intro** ({{word_count * 0.1}} words)
[Specific angle; why reader should care NOW]

**Section 1: {{section_title}}** ({{word_count * 0.25}} words)
- Key points to cover
- Data/proof points to include
- Transition to next section

[Continue for all sections]

**Conclusion/CTA** ({{word_count * 0.1}} words)
[Summary approach; CTA language]

### Research Required
- [Specific stats/data to find]
- [Expert quotes if needed]

### SEO Checklist
- [ ] Primary keyword in H1 and first 100 words
- [ ] Secondary keywords in H2s
- [ ] Internal links: [suggest 2-3]
- [ ] External links: [authority sources]

Competitive Intelligence Monitor

Problem Statement

Competitors move constantlyβ€”new features, pricing changes, messaging shifts, new content. Marketing can't manually track 5-10 competitors across websites, social, reviews, and news.

Efficiency Target: Always-current competitive intel vs. quarterly manual audits

Off-the-Shelf Options

Tool Capability Pricing
Crayon Competitive tracking, battle cards Enterprise
Klue Compete enablement, alerts Enterprise
Kompyte Automated tracking Mid-market
Feedly + AI Content aggregation $18+/mo

Custom Build Approach

Data Sources:

  • Competitor websites (changelog, pricing, product pages)
  • G2/Capterra reviews
  • LinkedIn company + employee posts
  • Press releases / news
  • Job postings (signals priorities)
  • Social media

Architecture:

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ Scheduled       β”‚
β”‚ Crawlers        β”‚
β”‚ (5 competitors) β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”˜
         β”‚
         β–Ό
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”     β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”     β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ Change          │────►│ Analysis        │────►│ Battle Card     β”‚
β”‚ Detection       β”‚     β”‚ Agent           β”‚     β”‚ Updater         β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜     β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜     β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
         β”‚                      β”‚                       β”‚
         β–Ό                      β–Ό                       β–Ό
  [Slack Alert]         [Weekly Digest]         [Sales Enablement]

Training Example: Competitive Change Analysis

Analyze this competitive intelligence data and generate insights.

COMPETITOR: {{competitor_name}}
MONITORING PERIOD: {{date_range}}

CHANGES DETECTED:
- Website Changes: {{website_diff_data}}
- Pricing Page: {{pricing_changes}}
- New Content Published: {{new_content_list}}
- Social/LinkedIn Activity: {{social_summary}}
- New Reviews: {{review_summary}}
- Job Postings: {{job_posting_summary}}

GENERATE:

## {{competitor_name}} Intelligence Update

### TL;DR
[3 bullets: most important changes and what they signal]

### Product/Feature Updates
[What's new; how it compares to us; customer impact]

### Messaging/Positioning Changes
[Shifts in how they talk about themselves; new claims]

### Pricing/Packaging
[Changes; competitive implications]

### Sentiment Signal
[What customers are saying in reviews; trend direction]

### Strategic Signals
[What job postings, content themes suggest about priorities]

### Recommended Actions
- Sales enablement: [battle card updates needed]
- Product: [features to highlight or gaps to close]
- Marketing: [content or messaging opportunities]

### Raw Evidence
[Links to specific sources for each insight]

Architecture Pattern

Uses ENT-RAG for searchable competitor history:

# Store competitive changes in vector DB
vectorstore.add_documents([
    Document(
        page_content=f"{competitor}: {change_description}",
        metadata={
            "competitor": competitor,
            "change_type": change_type,
            "date": date,
            "source": source_url
        }
    )
])

# Query for trends
similar_changes = vectorstore.similarity_search(
    "pricing strategy changes",
    filter={"date": {"$gte": "2025-01-01"}}
)

Campaign Performance Analyzer

Problem Statement

Marketers spend hours in analytics platforms pulling data, making spreadsheets, and trying to understand what's working. Analysis is reactive, not proactive.

Efficiency Target: Weekly reporting from 3 hrs β†’ 30 min; daily anomaly detection automated

Training Example: Campaign Analysis

Analyze campaign performance and generate actionable insights.

CAMPAIGN: {{campaign_name}}
PERIOD: {{date_range}}
GOAL: {{campaign_goal}}
BUDGET: {{budget}}

PERFORMANCE DATA:
{{performance_metrics_json}}

CHANNEL BREAKDOWN:
{{channel_data}}

CONVERSION FUNNEL:
{{funnel_data}}

HISTORICAL BENCHMARK:
{{benchmark_data}}

GENERATE:

## Campaign Analysis: {{campaign_name}}

### Performance Summary
| Metric | Actual | Target | vs. Target | vs. Benchmark |
|--------|--------|--------|------------|---------------|
[Key metrics with color coding guidance]

### What's Working
[Top 2-3 performers with hypotheses on why]
- Channel: {{best_channel}} - [insight]
- Creative: {{best_creative}} - [insight]
- Audience: {{best_audience}} - [insight]

### What's Not Working
[Bottom performers with diagnostic hypotheses]
- Issue: [specific]
- Likely cause: [hypothesis]
- Recommended test: [action]

### Anomalies Detected
[Unusual patterns worth investigating]

### Optimization Recommendations
1. [Specific action] - Expected impact: [X]
2. [Specific action] - Expected impact: [X]
3. [Specific action] - Expected impact: [X]

### Budget Reallocation Suggestion
[If applicable, where to shift spend]

### Next Period Forecast
[Based on current trajectory]

Chains

Content Pipeline Chain

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ SEO      │──►│ Brief    │──►│ First    │──►│ Editor   │──►│ SEO      β”‚
β”‚ Research β”‚   β”‚ Generatorβ”‚   β”‚ Draft    β”‚   β”‚ Agent    β”‚   β”‚ Optimizerβ”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

Implementation with CrewAI:

from crewai import Agent, Task, Crew, Process

seo_researcher = Agent(
    role="SEO Research Specialist",
    goal="Identify high-value keywords and competitor gaps",
    tools=[serp_tool, semrush_tool]
)

content_strategist = Agent(
    role="Content Strategist",
    goal="Create detailed, actionable content briefs"
)

writer = Agent(
    role="Content Writer",
    goal="Write engaging, SEO-optimized content"
)

editor = Agent(
    role="Editor",
    goal="Polish content for clarity and brand voice"
)

crew = Crew(
    agents=[seo_researcher, content_strategist, writer, editor],
    tasks=[research_task, brief_task, draft_task, edit_task],
    process=Process.sequential
)

See Pattern: Agent Chaining & Orchestration for details.