Bring a workflow you're tired of and I'll tell you honestly whether AI is the right answer.

I build AI agents and the production systems around them. Not chatbots. Agents that receive real work, do the job end to end, and escalate only what they genuinely can't settle.

Which means I've also watched a lot of automation projects fail, and the failure is almost never the model. It's automating something that should have been deleted, or building an agent where forty lines of plain code would have done the job cheaper and never hallucinated.

So this call is as much about what not to build.

Where the time actually goes in your workflow, and whether the expensive part is the part you think it is

Whether an agent is the right tool, or whether rules, a script, or a process change gets you there cheaper

If it is worth building: the architecture, which model, where a human has to stay in the loop, and roughly what it costs to run per month

The order I'd do it in, and what I'd deliberately leave until later

Founders and operators with a manual process running real volume. Teams with an AI project that works in a demo and misbehaves in production. Anyone who has been quoted for an automation build and wants a second opinion before signing.

If you want someone to validate a plan you've already committed to, I'm the wrong call. I'll tell you what I actually think.