Technical articles, tutorials, and thoughts on software engineering
AI Automation
Most "AI automation consultants" are reselling Zapier workflows with a markup. Here's how I'd vet one as a buyer: the audit they should run, what pricing models really cost you, and the questions that expose thin expertise.
Model choice isn't about which is 'smartest' in the abstract — it's about which fits your specific task, cost constraints, and deployment requirements.
Adding an LLM feature is easy to prototype and surprisingly expensive to run reliably at scale. Here's what actually shows up on the bill.
Before building an AI automation, do the math on what it actually saves versus what it costs to build and run — here's a practical framework.
Hallucination can't be fully eliminated, but specific prompt and system design choices reduce it substantially.
RAG lets an LLM answer questions using your specific documents and data, instead of relying only on what it learned during training.
Traditional automation follows rigid rules. AI automation adds language understanding, so it can handle the unstructured inputs that broke your old RPA setup.
