A Guide to Which AI to Use in the Agentic Era

A Guide to Which AI to Use in the Agentic Era
By Ethan Mollick

If you’ve been feeling a little lost keeping up with AI lately, you’re not alone. A year ago, “using AI” mostly meant chatting with a bot. You typed, it replied. Simple.

Not anymore.

In his latest piece, A Guide to Which AI to Use in the Agentic Era, Wharton professor Ethan Mollick explains how we’ve quietly shifted from chatbots to something much bigger, AI agents that can actually do things for you.

And that changes everything.

Mollick breaks the landscape into three parts: Models, Apps, and Harnesses. Think of the model as the brain, GPT-5.2, Claude Opus 4.6, Gemini 3 Pro. The app is where you interact with it, like ChatGPT’s website or Claude’s desktop tool. The harness is the real game changer. It’s what lets the AI use tools, browse the web, write and run code, manipulate spreadsheets, or build slide decks.

Same brain. Completely different outcomes.

He shows how a model inside a simple chat window behaves very differently from that same model inside a powerful coding or research environment. It’s like comparing someone giving advice versus someone rolling up their sleeves and finishing the project for you.

What really stands out is his practical advice. If you’re serious about using AI for work, pay for the advanced model. Choose it manually. Don’t rely on defaults. Then go beyond casual chatting. Upload real documents. Assign complex tasks. Treat the AI less like a search engine and more like a junior colleague you manage.

And that’s the heart of this shift.

We’re moving from AI that talks… to AI that acts.

It’s still early. The tools can be confusing. Sometimes messy. But learning how to work with agents now is a bit like learning the internet in the 90s. Awkward at first, then suddenly essential.

If you care about where work is heading, this guide is worth your time.