What do you hate, why do you hate it, and who do you wish could do it?

The best AI training doesn't start with AI. It starts with a question.

Shubham Chandra, Head of Curriculum8 min read

Key takeaways

The most effective AI training opens with a question about the job: "What do you hate about it, why, and who do you wish could do it?"
Saying the frustration out loud, in a room where everyone else has the same one, is what moves people from skeptical to impatient.
Participants leave with a working agent for a problem they just named, a map of one real process, and a queue of the next three to five builds.
Teams that get a return on AI tend to start from a task someone already resents, and work backward to the tool.

The question that starts every program

The first thing participants hear is three questions. What do you hate about your job? Why do you hate it? And who do you wish could do it? The definition of a large language model can wait.

The answers are consistent across roles and industries: reformatting data between systems, writing the same status update in three formats, drafting boilerplate that nobody reads carefully but everyone demands. All of it is repetitive work someone wishes they could hand off.

That "someone" they wish could do it? More often than not, an AI can. But nobody has drawn the line between "I hate writing weekly status reports" and "an agent can draft that from the tracker."

Discussion gets a room further than a lecture does

The format is borrowed from the University of Chicago tradition: discussion over lecture, questions over slides. The instructor sets up the conditions, and the room works the problem through for itself.

When a room full of professionals says its pain points out loud, the first thing they notice is that the frustration is shared. The operations manager hates the same kind of work the marketing director hates.

That recognition creates energy, and the room shifts from audience to working group. It carries people from "I'm not sure AI is relevant" on the first morning to "I built an agent for my own work" by the end of day one, and to a mapped process with the next builds queued by the end of day two.

The moment a room realizes they all lose hours every week to the same kind of task, skepticism turns into urgency.

From pain to a working agent in two days

Day one opens with those three questions, then a live demo in which the instructor builds an agent on one volunteer's real work. Each person then builds their own from a starter written for their role, reads the first output critically, tightens it, and runs it again. Day two belongs to the team: one real process mapped end to end, the next three to five builds pulled out of it, and one named action for Monday.

By the end of the second day, three things leave the room with the team:

A working agent: built by each person on day one, from a starter written for their role
A map of one real process: how it runs today and how it runs rebuilt, with the agent steps marked
An implementation queue: the next three to five builds, with one named action for Monday

Why this scales

The organizations getting a real return on AI share a starting point: a list of repetitive, time-consuming tasks, and a tool worked backward from each one. An awareness campaign never made that list.

The question is the same whoever is in the room: what do you hate, and who do you wish could do it? People who name their own problem tend to own the solution.

One question tells a leader where the gap is

A team that can answer "what do you hate" but has nothing to show for "what have you built to fix it" has located its own gap.

Spending on AI training is growing fast, and most of it is headed for platforms nobody finishes. The programs that return the money start with a real problem and end with a tool somebody uses the following week.