[{"data":1,"prerenderedAt":118},["ShallowReactive",2],{"\u002Fwork\u002Fcase-studies\u002Fai-assisted-automation":3},{"id":4,"title":5,"body":6,"businessProblem":87,"confidentiality":88,"description":89,"employer":90,"extension":91,"featured":92,"heroStat":90,"meta":93,"navigation":92,"ogImage":90,"order":94,"path":95,"publishedAt":96,"results":97,"seo":107,"stem":108,"summary":109,"technologies":110,"updatedAt":90,"__hash__":117},"caseStudies\u002Fwork\u002Fcase-studies\u002Fai-assisted-automation.md","AI-Assisted Business Automation",{"type":7,"value":8,"toc":78},"minimark",[9,14,18,22,27,34,40,46,52,58,62,71,75],[10,11,13],"h2",{"id":12},"the-problem","The problem",[15,16,17],"p",{},"There's a wide gap between \"AI wrote some code\" and \"AI-assisted development that's actually\ntrustworthy in production.\" The first is a demo. The second requires a workflow — one where AI tools\nhandle the parts they're genuinely good at (drafting, boilerplate, exploring an unfamiliar API,\nfirst-pass tests) while a human still owns system design, correctness review, and the decision to\nship.",[10,19,21],{"id":20},"the-workflow","The workflow",[23,24],"architecture-diagram",{":steps":25,"title":26},"[{\"label\":\"Business problem\",\"detail\":\"Defined by me, not the model\"},{\"label\":\"System design\",\"detail\":\"Architecture + constraints set upfront\"},{\"label\":\"AI-assisted implementation\",\"detail\":\"Claude Code, Codex, Cursor, Gemini\"},{\"label\":\"Tests + human review\",\"detail\":\"Nothing ships unread\"},{\"label\":\"Deploy + monitor\",\"detail\":\"Same ops discipline as any change\"}]","How AI fits into the process",[15,28,29,33],{},[30,31,32],"strong",{},"Business problem first."," Every task starts with the actual operational problem — a catalogue\nmismatch, a slow report, a manual process eating hours every week — not a prompt in search of a use\ncase.",[15,35,36,39],{},[30,37,38],{},"System design stays mine."," Architecture decisions, data models, and failure-mode thinking happen\nbefore AI tools touch the problem. This is the part that determines whether the resulting system is\nmaintainable six months later, and it's the part I don't delegate.",[15,41,42,45],{},[30,43,44],{},"AI-assisted implementation."," Different tools for different jobs: Claude Code and Codex for\nlarger, multi-file implementation work and refactors; Cursor for fast in-editor iteration; Gemini\nfor research and second opinions on approach. I treat model output as a strong first draft, not a\nfinished product.",[15,47,48,51],{},[30,49,50],{},"Tests and human review."," Nothing generated ships without tests and a real read-through — the same\nbar I'd hold a human collaborator's pull request to. AI-generated code gets the same scrutiny as\nanyone else's, arguably more, because it doesn't yet have a track record with this specific codebase.",[15,53,54,57],{},[30,55,56],{},"Deployment and monitoring."," Once something ships, it's watched the same way any production change\nwould be — logging, alerting, and a rollback plan, not \"the AI said it works.\"",[10,59,61],{"id":60},"why-this-matters","Why this matters",[15,63,64,65,70],{},"The honest pitch isn't \"AI writes my code.\" It's that AI tools meaningfully speed up the\nimplementation phase of systems I'd have designed and validated either way — the catalogue pipelines,\nmarketplace sync jobs, and internal tools described in my other ",[66,67,69],"a",{"href":68},"\u002Fwork","case studies"," get built\nfaster because of this workflow, not because judgment got outsourced to a model.",[10,72,74],{"id":73},"what-im-still-refining","What I'm still refining",[15,76,77],{},"The weakest link right now is turning implicit operational knowledge — the stuff I know about how a\nspecific supplier feed misbehaves, or how a marketplace's edge cases work — into context the AI tools\ncan actually use, instead of re-explaining it every session. That's the direction most of my current\ntooling investment is going.",{"title":79,"searchDepth":80,"depth":80,"links":81},"",3,[82,84,85,86],{"id":12,"depth":83,"text":13},2,{"id":20,"depth":83,"text":21},{"id":60,"depth":83,"text":61},{"id":73,"depth":83,"text":74},"AI coding tools are good at producing code fast. They're not automatically good at producing code that's correct, maintainable, and safe to run against real supplier data, real inventory, and real customer orders — that gap is where most \"AI-generated\" work quietly fails.\n","public","The workflow I use to fold AI tools into real ecommerce systems work, from problem framing through deployment and monitoring — with human review at every step that matters.",null,"md",true,{},4,"\u002Fwork\u002Fcase-studies\u002Fai-assisted-automation","2026-03-01",[98,101,104],{"label":99,"value":100},"AI tools used daily","Claude Code, Codex, Cursor, Gemini",{"label":102,"value":103},"Human review","Every deploy",{"label":105,"value":106},"Judgment retained by","Me, not the model",{"title":5,"description":89},"work\u002Fcase-studies\u002Fai-assisted-automation","How Claude Code, Codex, Cursor, and Gemini fit into a real development process — not just code snippets.",[111,112,113,114,115,116],"Claude Code","Codex","Cursor","Gemini","Python","CI\u002FCD","4e_ExX31zgQuu7opLtcYSbLIwiTkCTsvT9qGnt_TRP8",1785688078907]