Real Marketing Ops AI workflows from people using AI at work. Browse documented examples, prompts, tools, and outcomes.
Jordan Crawford built a lead validation pipeline using Exa, a LinkedIn scraper, and Claude Haiku. He discovered that raw search data has a 39% error rate. The fix: a multi-step workflow that discovers, enriches, and validates every profile for just $0.04 per lead. Jordan is worth listening to when it comes to digging data gold with the help of LLMs, among many other things. I would recommend checking out his LinkedIn for useful tips.
Nikhil Krishnan's takeaway from healthcare LLM workshops is simple: the first skill is not a workflow, it's learning to ask Claude for help when you're stuck. That means taking a screenshot, asking what to do next, using the model to improve the process itself, and pushing for explanations that match how you think. It is the same muscle as learning to Google well. The real lesson for readers is that AI becomes useful faster when you treat it like a debugging partner, not a person you wait on for answers.
Khalid Warsame uses a simple prompt to cut corporate fluff from his inbox. He forces the AI to rewrite emails to be exactly 50% shorter. It kills the filler while keeping the core message and original tone intact.
Stop paying the 'Blank Slate Tax' by using Claude Skills—markdown files that encode your brand voice and audience. This system uses three core workflows: an Email Sequence Builder, a Competitor Intelligence Brief, and an SEO Article Optimizer.
A step-by-step guide to building an AI review monitoring system using Make.com and OpenAI. It watches Google My Business reviews, filters by sentiment, analyzes with GPT, drafts personalized responses, logs everything to Google Sheets, and sends Slack alerts for critical negative reviews. Handles 200+ reviews per week, cutting average response time from 18 hours to 2 hours.
A triage agent that reads support emails, pulls order context from Google Sheets, applies a structured returns policy, drafts responses in the brand's voice, and queues them for one-click human approval. Built with Claude Projects, Gmail connector, and a decision-tree policy document. Handles 70% of cases autonomously, pre-processes the rest.