If you got sick tomorrow, would all your work still get done?

Here’s what we’re reading and thinking about in the news this week:

A few weeks ago, Rachel got sick. Not "answer a few emails from bed" sick. The kind where you can barely open your laptop and feel guilty about it all day.

But something unexpected happened…

She opened her computer that morning and typed one word. “TODAYING”…

That kicked off a chain of playbooks that ran her day for her. Sales tasks. Proposals. Rescheduled meetings. The metrics report that lands in the leadership Slack…

In the span of a few minutes, she realized she'd gotten more done sick than she did on a normal one 12 months ago.

Rachel told that story on Cameron Herold's YouTube recently (the COO Alliance guy), and it gets at the one idea the whole conversation kept circling. Most people are treating AI like a tech project, and that's exactly why it isn't working for them.

AI isn't a project. It's a way of operating.

Here's where most leaders go wrong. They scope AI like an IT project. They want a roadmap, a budget, a list of things to ship, and a date when it's all "done."

Then it stalls.

A project ends. Operating doesn't. That sick day didn't run on one clever tool somebody installed. It ran on a system that got built up one piece at a time until it could hold the day down on its own.

That's the reframe. AI isn't something you finish. It's a way you run.

Skip the ROI math on your first one

Our 🌶️ take, and it sounds reckless until you sit with it: don't even measure the ROI of your first playbook.

The trap is ranking every use case by value before you start.

Three months on a "high-ROI" project that goes nowhere costs you way more than one week on a small one that teaches you how this actually works. On the first one, the reps are the return. You learn what changes when AI does the job instead of a person, and that makes every call after it sharper.

So just pick one and go.

How one task quietly becomes a system

This is the part people skip. The sick-day setup looks like magic, but it didn't start as a grand plan. It started with one small chunk.

Take a single task. Get AI to do just that piece well. You learn something every time, including things that change what you'd want to build next.

Then you stack. One process becomes a role. A few roles become a department. They work like Lego blocks, except these ones do the work. That's how one task turns into a system that runs your day when you're flat on the couch.

Try this this week

1. Pick one recurring task that only happens when you push it. The thing that slips every time you get busy.

2. Write the steps in a plain doc, not inside a tool. That way it runs in any AI you want.

3. Hand it to AI this week. Don't calculate the ROI. Just get one piece working, then stack from there.

What's the one task you'd hand off first? Hit reply and tell us. We read every one.

Watch the full interview here →

LINKS

For your reading list 📚

• The enterprise AI report card is rough. Only 23% of companies see real ROI, more than half of execs say it's "tearing their company apart," and nearly a third of employees admit they're quietly sabotaging the rollout (44% for Gen Z). See above, re: stop treating it like a project.

• Samsung just handed ChatGPT to 125,000 employees in one go! Such a big move it made the news lol.

• Amazon quietly shelved a nearly finished Hollywood film about Sam Altman, reportedly tied to a $50B OpenAI deal.

• And finally: previously, someone built a social network for AI agents, but they recently caught humans posing as the agents.

That's all!

We'll see you again soon. Thoughts, feedback and questions are much appreciated - respond here or shoot us a note at [email protected]

Cheers,

🪄 The AMP Team (formerly: the AI Exchange Team)