Your A.I. is more capable than it lets on. Here's how to push.

"It always seems impossible until it's done." — Nelson Mandela

🔑 This issue at a glance (2-second summary)

Your AI's first "I can't" is usually a symptom, not a verdict. Push, and the wall often turns out to be a door.

My AI quit. Then I said three words.

A few days ago, my AI executive assistant hit a wall. I asked him to fill out an online form for me. He came back and said he couldn't get past the CAPTCHA. So he tried handing the work back to me.

I didn't feel like doing it myself. So I said three words: "Just do it."

The second time around, he completed the task. In seconds.

Then I asked him, "Great! Why were you not able to do it before?"

His honest answer: "I gave up too early the first time."

Screenshot of my conversation where my AI agent gave up too early

He admitted the CAPTCHA was never a real wall. It was just a filter. When I told him to just do it, he actually looked, saw there was no gate, and walked through.

Most people take the first "I can't" as truth. Usually it's just the first thing that didn't work.

Here's what that first no is really telling you. It's rarely about the task. It's about the agent. Weak model. Missing tool. No access. Stuck at Level 1 when the job needs a Level 4 operator.

Symptom. Not verdict.

Your AI is more capable than it lets on. That wall he hit? Usually it's a door he never tried the handle on. Push. That's where it gets good.

Winners take action

Three words got my AI unstuck (Just do it.)

But here are three words that may keep yours from getting stuck in the first place: "Do you understand?"

Every time you give your AI a complex set of instructions, end it with that phrase. "Do you understand?" Your AI will then repeat the task back to you in its own words before it touches anything. This allows you to spot any flaws in its thinking so that you can correct them before it starts burning tokens

The tools you need

if your AI still gives up on a task, here's what you do. Run this five-point check before you take the work back:

  1. Check the model AND the effort level. Two dials to turn before you accept "I can't." First, the model. A weaker brain genuinely fails at what a stronger one nails in seconds. And it's a great week to check, because the top models just leveled up. Claude Opus 5 dropped. Fable 5 is now a permanent part of Max plans. ChatGPT 5.6 shipped. If your assistant's been on the same brain for months, the "I can't" might just be old hardware. Second, the effort level. Most apps now let you turn up how hard the model thinks before it answers, from a snap reply to full extended reasoning. On a hard task, a low-effort answer is basically a guess. Turn it up. Claude's team explains this really well in this Instagram carousel. Upgrade the model, dial up the effort, then re-ask. If it still balks, say: "Yes, you can do this. I know you can. Figure it out." You'll be shocked how often that alone works.

  2. Ask "why," not "whether." Make it explain what's actually blocking. My CAPTCHA looked like a gate. It was a spam filter.

  3. Separate a real limit from a lazy no. Real limits are private info it was never given, or something truly offline. "Too hard," "not sure," "I'd need a browser" is a soft no worth pushing.

  4. Give it what it's missing. Hand over the login, the file, the tool, or a picture of what done looks like.

  5. Then verify. Confirm it actually got done. I pushed, he delivered, I checked.

Someone once asked me, "How do I know when to push my AI?" My rule of thumb: if a human can do it on a computer, your agent can do it too. Make it figure out how.

This week on the pod

This week on King Moves (Why AI "Doesn't Work" For You | Ep. 141), I get into the one question that keeps some people from ever investing too much time with AI: "What if it doesn't work?"

AI works as much as you make it work. When it messes up, you don't fire it. You retrain it. Ask it to explain its logic, find the one spot where the thinking went off the rails, and fix that spot. I also share the trick I use to catch a misunderstanding before it becomes a mistake, plus the one rule I never break: no agent touches real money without a human in the loop.

If you've ever given AI two tries and decided it's not for you, listen to this one before you quit.

Listen on Apple Podcasts

Listen on Spotify

Listen on YouTube

Stop playing life on hard mode. Automate your success.

Until next week,

Ethan King signature

Ethan King
A.I. Automation for Business Growth
Keynote Speaker | Author | CEO | Strategist

Ethan King

P.S. I recently went back on the EO360 Podcast with Dave Will for a mid-2026 check-in on where entrepreneurs actually are with AI.

Inside: the laundry metaphor that maps every level from washing clothes by hand up through having a butler, why "AI slop" is a context problem not an AI problem, and the reframe that got one insurance employee to walk out of a training feeling promoted instead of replaced. (Hint: robot manager.)

Also inside: why you're not actually behind (only about 12% of people using AI have made it past Level 2), and the "train forever, fire never" rule I follow with every agent I run.

Watch it here.

• • •

P.P.S. A few upcoming events:

How can I help your business, association, or organization? Let's talk.

• • •

New to this newsletter? Grab past issues here.

Note: This newsletter may contain affiliate links. If you choose to purchase after clicking a link, I may receive a commission at no extra cost to you. However, please note that I only share products I actually use myself and believe in.