Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Building an AI Agent for Content Management, Re...

Sponsored · Your Podcast. Everywhere. Effortlessly. Share. Educate. Inspire. Entertain. You do you. We'll handle the rest. →

Building an AI Agent for Content Management, Research & Growth [Real Example]

You will learn how to:

✅ Build an agent for a few $$$ + a solid stack
✅ use GA4, GSC, Scraping data to power up your agent
✅ Build processes and skills so your output is excellent

Marco Giordano

November 21, 2026

Transcript

70 slides
  1. Building an AI Agent for Content Management, Research & Growth

    [Real Example] MeasureCamp Zurich 2026 Marco Giordano Web Analytics Consultant
  2. Ugh... no custom attribution and only 14 months in Explorations!?

    Sampling, thresholding... dude you must be joking! Only 16 months and hidden data? We can help!
  3. What is n8n? A workflow automation platform for connecting apps,

    APIs and data. Connects apps, APIs, and data sources Builds workflows from triggers, rules, and actions Automates repeatable processes with little or no code From a trigger to a completed workflow, with human approval where it matters.
  4. What is Hermes Agent? An open-source AI agent for getting

    work done across tools, tasks, and workflows. Plans and executes multi-step tasks Uses tools and connected services Self-learning agent that acts as your personal assistant. Remembers context and learns reusable skills
  5. Hermes Agent: the wrapper around tools Model decides Hermes Agent

    orchestrates Tools execute Call an API n8n Run a workflow Read or write data Hermes Agent Send a message or create a task Hermes Agent turns model reasoning into useful actions by calling the right tools at the right time.
  6. Claude (Code) is a nice intro for coding and basic

    automation: 👉 Creating apps 👉 Automate basic data extraction 👉 Be called by Hermes 👉 Help you with the setups
  7. Before creating content, 3 questions matter 1 Which new articles

    or pages are missing? 2 Do we already have a page for this topic? Without these answers, content teams risk missing opportunities, duplicating existing pages, or overlooking what competitors already cover. 3 What are our competitors doing?
  8. The problem content automation solves Mockup visuals Research + synthesis

    Research is scattered across sources Collating evidence takes time Ideas stay abstract without a visual mockup Research faster Goal: being faster with content production AND management. Bring different sources together Turn the brief into mockup visuals
  9. Running my automation Marco wants a content analysis of these

    terms. Got it, running the analysis. I send you the output!
  10. But the SERP data??? How many pages I need for

    a set of keywords SERP data SERP Clustering
  11. Where is the Google data? I know what to improve

    or where we stand. Let me check what’s in I store all the data. st our 1 party sources.
  12. Evidence.dev + WordPress data + GA4 + GSC. Build your

    personal dashboard with many more features and scalability.
  13. → → → AI costs (month) $10 + DataForSEO (month)

    $4-5 + VPS (month) $10 = around $24-25/month (even exaggerating here).
  14. CONTENT WRITING STAYS HUMAN The automation does: research and summarization

    for you prepares drafts But is NOT suitable for 100% automation.
  15. Scraping, in layers Different layers of scraping based on the

    roadblocks and/or needs Trafilatura Article text out, no selectors Crawl4AI JS pages to clean markdown Scrapling Selectors that survive layout changes Plus curl_cffi for requests that must look like a browser, and ScrapeGraphAI for LLM extraction.
  16. Run on my own site Crawl4AI, a sitemap pipeline and

    DeepSeek Fetch, all on seotistics.com Crawl4AI: 33,946 characters raw, 24,267 after the filter. Pipeline: 25 of 25 pages kept, 1.5s warm.
  17. Research and gap analysis How I decide what to write

    next Collect SERPs from DataForSEO, pages via Crawl4AI Cluster Queries grouped into topics Decide Gaps between my pages and the ones that rank The competitor playbook scraper runs the same flow on a rival's URLs.
  18. Same question, two tools 5/5 0/5 209k+ narrow script tool:

    right figure Google's BigQuery MCP: right figure input tokens per MCP run (script: under 2k) One question, 5 runs each. A diagnosis of one setup, not a benchmark.
  19. What the agent is made of 88 5+ 1 skills

    with my voice and rules tools incl. BigQuery, n8n, Canva human approves before go-live An orchestrator agent runs the day to day. Claude Code does the heavy builds.
  20. Claude Code runs inside Hermes Hermes orchestrates, Claude Code does

    the heavy builds. Hermes calls the Claude Code CLI in print mode on my VPS, on my subscription. This deck's texture review ran that way.
  21. The visual factory One agent, a different tool for each

    shape of content Comparison: infographic editor | Concept: Excalidraw | Product screen: mock-UI | Deck: Canva
  22. ✅ LLM Automation goes beyond Claude Code and ChatGPT ✅

    You still need to do data engineering
  23. ✅ LLM Automation goes beyond Claude Code and ChatGPT ✅

    You still need to do data engineering ✅ Context + providing right sources is your job
  24. ✅ LLM Automation goes beyond Claude Code and ChatGPT ✅

    You still need to do data engineering ✅ Context + providing right sources is your job ✅ Manual labor is still there (but less)

More Decks by Marco Giordano

Other Decks in Marketing & SEO