Unit 10AI that does real work

The big idea

None of this course is theory. It's how I run my company, Future Summit, today. Seeing real systems makes it clear where AI belongs and where people still have to be.

Why it matters

The most important skill of the next decade is deciding where the human stays in the loop. You learn that by watching a system make decisions step by step and asking, at each step, "Would I trust this without checking?"

Four systems I run

1. A research engine. The same loop from Unit 8. It plans, researches, drafts, checks and fixes. I review the result.
2. A publishing engine. It turns research into posts for about ten platforms. It only works because the input is good. I call that the feedstock. And it checks that every post matches what the research actually said.
3. An outreach pipeline. It takes a list of leaders, finds their contact details, picks the right kind of letter for each person, writes it, and rates which letters are ready and which need a human. Then it produces printed letters and emails. A person checks the data at every point where a mistake would matter.
4. An assistant. It answers simple emails, batches the harder ones for me to approve or decline, and books meetings after checking that they're worth my time.

Research AI vs applied AI

• Research AI helps a human make a decision. It finds, sorts and checks information.
• Applied AI takes action. It produces something real: a post, a letter, a booked meeting.

Applied AI needs more checkpoints, because its mistakes go out into the world.

The honest trade-offs

• These systems take a lot of time to build and they need constant upkeep. Tools change fast.
• Sometimes building it with AI takes longer than just doing it by hand.
• The real question is whether pushing quality from 83% to 90% is worth the effort for that job. Sometimes it is. Often 83% is fine.
• I'd always rather have a human reviewing the work. But I have to weigh that against running the business.
• AI should give you your time back. It shouldn't take over your life.

Do this

1. List three tasks you do every week. School, a job, a club, chores, anything.
2. For each one, write down:
   • Which steps need a fixed, exact answer?
   • Which steps need judgment?
   • Where would a mistake hurt someone?
3. Decide for each step: AI, human, or AI with a human check.
4. Pick one task and design a simple version with the two-window loop. Mark exactly where you'd check it.

Questions to ask yourself

• If this step goes wrong, who gets hurt?
• Is the AI's review actually needed, or is it just telling me something needs review?
• Is this saving me time, or am I just building for fun? (Both are fine. Know which.)

In my words

"Every single stage, right at the point of truth, to make sure we're keeping our standards high and we're not getting garbage."

"The best thing that you can learn is critical thinking, and using AI in ways that are meaningful and add value, but don't take away from your life."

Worksheet

Use Worksheet 10: Human or AI? in the PDF kit.