Do You Actually Need AI?

We’re starting to learn a hard truth: AI isn’t always the right answer.


I’m seeing a smaller version of the “AI bubble” play out in everyday workflows. We’re often overestimating which problems require complex intelligence—and underestimating how effective simple, deterministic automation still is.

It’s tempting to create a chatbot and build an entire agent for a process. But in many cases, a lightweight workflow with a single GPT call for intent detection delivers the same outcome as a complex agent—without burning computing resources or building operational overhead.

I’m excited about MCP servers and agentic modeling. But the real question isn’t “can this be agentic?” It’s “what level of intelligence does this workflow actually need?”

I’m helping local businesses automate their workflows, and often, the most reliable solutions are simple: a few lines of Python + a webhook, or just a few modules in Make

I encourage all you automation enthusiasts to consider a basic low-code solution the next time you’re tempted to reach for LangGraph. You might be surprised at what still works!