Seven interactive sketches showing how AI agents can plan, decide, and act inside real marketing work. Each one is a self-contained HTML page you can open and run.
These are exploratory — written to think through the patterns, not to sell anything. The goal is to make the moving parts visible.
Something happens — a lead appears, a page is visited, a date is reached. The workflow is triggered.
It looks at the data, enriches it, and routes down a branch. This is where the "agent" part earns its name.
Emails, CRM updates, Slack pings, ad launches. The outcome is observed and feeds back into the next run.
A visitor browses but doesn't buy. The agent watches what they did, sorts them by intent, and picks an appropriate email + ad combo.
intent levelThree specialist agents work in parallel on the same page — one on copy, one on SEO, one on conversion. Their findings merge into a PDF.
requests + BeautifulSoup scrape the target URLContent, SEO, ConversionReportLab assembles the PDF, Slack notifies the teamThe gentlest version. A form submission turns into either a nurture sequence or a sales alert, depending on company size. Good starting point.
companySizeA list of leads from different sources walks through the agent. Unsubscribes are skipped, each remaining lead gets the right template, CRM records update live.
leads[] queue from Pipedrivesource → emailContactedA three-phase launch sequence — pre-launch drafting, launch-day publishing and outreach, post-launch monitoring and metrics review.
A new lead arrives in the CRM. The agent enriches it, checks company size, and routes big accounts straight to sales with a deal created and a Slack ping.
employees, industryType a product or topic and the agent drafts a full social plan — slogan, Instagram, X, Facebook, and a simple posting schedule.