In this run I took a crack at a resume and COV generation pipeline. It's a Cron task to check a ClickUp jobs list where I've pasted nothing but the job posting url. Then made it an Agent to scale up the difficulty -Firecrawl scrapes to the page for company and posting information -analyzes and extracts keywords from the Job Posting, -uses a reference file of my resume and builds a new version specific to that job posting along with an optimized COV -updates the ClickUp item with a flushed title, company info, salary details and PDF versions of the generated files This one took me deep with what broke and went wrong. From Unicode formatting in the PDF tasks to context bloating with the resume handoffs. It was a fun build and I like the result, but I'm definitely wanting to be a bit more structured with how I'm building.

Hey everyone here is my introduction (again with more context): 20+ years in performance improvement. 5 months into AI and automation. That combination is why I'm here. My name is Chris Sharkey. I run Sharkitect Digital out of Kansas City. Most people in the AI space sell automations. I don't. I sell an AI Transformation Partnership — built around one mission: helping small businesses understand, trust, and actually implement AI into their operations. Not by handing them tools. By diagnosing their systems and solving at the root. Good doctors don't write prescriptions the second you walk in. They diagnose first. That's the model. Three months ago I started applying this with real clients. First engagement: a construction company that reached out for a completely different project. I sat with them. Ran the diagnosis. Mapped their workflow end to end. What they called for wasn't the real problem. It was a problem — just not the one bleeding them out. Once I showed them what was actually happening, it was clear: they were spending 80–100 hours every week doing nothing but manually moving data from one place to another. By hand. 35 minutes per estimate. Calculation errors costing them real bids. They didn't fully realize how much it was costing them until we put it on paper. So I built the solution around how they already work. No new system to learn. No process overhaul. It adapted to them — not the other way around. Training took less than two hours. Adoption was immediate. 35 minutes → 30 seconds. They're saving $104K–$156K annually in labor alone. That's the original project they called about? Still on the list. Just no longer a priority. Within the first month, they saw the shift. We're now working on two additional projects together — each one targeting a different operational bottleneck. None of it is built in isolation. Every system is designed to stack, integrate, compound, and scale. Each build talks to the last. The business gets smarter with every layer, not just bigger.