Let's Talk AI Chat

AI moves fast enough that it is easy to go deep on one set of tools and quietly stop looking around. Dan opens this AI Exchange with the ground rule that frames everything after it: we don't marry a tool. We find AI that gets the job done — and the tools are changing too fast for loyalty to survive contact with the market.

This session steps back and looks hard at one thing: ChatGPT Work, and where it actually fits next to Chat and Codex. Dan walks the setup end to end — the desktop app, the Chat/Work toggle, choosing whether a job runs locally or in the cloud, connecting only the tools the job needs, the built-in browser, and the handful of settings that matter.

Then he runs two real Workuity workflows live. An expense report, found and filled from emailed receipts — driving the real Concur website, with no API access anywhere in the chain. And two thousand LinkedIn contacts turned into a usable Phoenix commercial real estate list, exported to a clean Excel file.

Between the demos he shows why agent browsing burns tokens so quickly, why the boundary gets set before the first click, and what happened the morning he arrived with no deck and handed his narrative to every AI on his machine to see which could build it best.

The through-line is not which model is best. It is that the model matters less than the harness around it, and that the work you do well should be saved once as a skill and reused — then scheduled, once it has earned the trust.