I've seen this play out before.

I spent 9 years selling on Amazon. Over that time period I sold $16 million in products, and built an Amazon FBA community to $340,000 in recurring revenue before I sold it.

At the peak of the Amazon FBA hype, more than 1,000 new seller accounts were opening every single day. I knew kids buying fake Nike socks off Alibaba and clearing $10,000 a month reselling them (not condoning this, it's just something I observed).

Then Amazon raised fees every year, supply exploded, and the margins went to 5 to 7% net if you were lucky. This opportunity went from a gold rush to a commodity market in about 4 years.

Now look at AI. Builds that sold for $10,000 a few years ago are being done by freelancers overseas for $500 a pop.

It's the exact same situation playing out across a different business model.

Here's why that matters, and what I recommend doing about it.

A commodity market runs on four rules. I learned them the hard way on Amazon and they're playing out in AI right now.

Rule 1: The newest seller sets the price. On Amazon it was the kid willing to lose money to win the Buy Box. In AI it's the 19-year-old with an n8n account who'll build your workflow for $500.

Rule 2: When the build cost hits zero, the build price follows. Claude can build workflows now. People now pay for the outcome and the guidance, not the build.

Rule 3: The course boom marks the top. Everyone and their mother had an Amazon course from 2021 to 2024, right as margins collapsed.

Rule 4: Every build you ship becomes a liability. If (and when) it breaks, they call you. Offering a maintenance retainer can make sense for certain builds, but it also tells the client you don't trust your own work.

People will often counter the above points by saying something like "but the models keep getting better, so my builds get easier."

Yes. But that's the problem. Anthropic and OpenAI are building your product into the model with skills, Cowork/ChatGPT Work, and Claude Code/Codex. If you sell builds, every model update makes your business less valuable.

You want the opposite.

A business where the better the models get, the more valuable you become.

The people who survived on Amazon owned relationships, distribution, and exclusivity. Not products. Same here.

Here's how to apply these principles to your AI business:

Step 1: Take the word "automation" out of your offer this week.

Business owners don't want automation. They want outcomes. An "AI automation agency" sells a black box the owner can't see, and the cheapest bid wins every time.

Step 2: Lead with a $999 AI assessment, not a build.

One 45-minute call where you uncover where AI can create an outcome in their business, then a short report with 3 to 7 recommendations. You implement nothing. It takes about two hours, which works out to roughly $500 an hour. I've sold about 26 of these in 2026. It's the easiest sell in the world because there's nothing to break.

Step 3: Upsell the assessment client to an AI Concierge retainer.

Two 45-minute calls a month, done with them, not for them. They share their screen, you tell them what to click and what to build inside their own Claude account, so they own the skill and the outcome. I charge $1,500 to $2,000 a month for about an hour and a half of work. That's over $1,200 an hour.

Step 4: Teach them to fish anyway.

"Won't they fire me once they can build it themselves?" No. Every call makes the relationship stronger. The more they can do, the more they lean on you to find the next outcome.

Compare the three. One-off builds take 10 to 30 hours, come with maintenance forever, and your hourly rate drops every month. Assessments pay about $500 an hour. Concierge pays over $1,000 an hour, and it goes up as the models improve.

So I would recommend selling the diagnosis and the guidance, letting Claude do the building, and you become the person whose value compounds with every model release.

Rewrite your offer this week to focus more on outcomes and position yourself as the coach/guide to your clients and work WITH them to get the outcomes.

- Corey

A few notes/tweets/cool things sourced from the AI community.

1)

Elon Musk @elonmusk Important note regarding Grok @Bot: Going forward, @SpaceX will use the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs. Whatever is most likely to give you the best outcome. 6:46 AM • Oct 7, 2026 4826 Retweets 64228 Likes Read 4461 replies

​

Huge for Grok Bot users.

2)

🥔🥔🥔 @argofowl everyone thinks codex computer use only works inside codex it doesn't, it's a local mcp server in the chatgpt mac app and claude code can just call it, even headless with claude -p same 8-task test: opus 5.5 on codex's engine got 6/8, codex itself got 6/8, cua driver got 3-4/8 2:23 PM • Oct 3, 2026 65 Retweets 1351 Likes Read 130 replies

​

Get the power of Codex computer use inside Claude.

3)

Tavus @tavus Introducing Griffin, the first model to pass the video Turing test. 48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video. It’s the first Human Interaction Model (HIM). 4:59 PM • Oct 1, 2026 4295 Retweets 39275 Likes Read 3369 replies

​

This went mega viral for a reason. Crazy.

Thanks for reading the Corey's Notes newsletter. I'd appreciate it if you sent it to a friend that might find it interesting.

Be sure to check out t he Build With AI podcast on Apple, Spotify, and YouTube. If you're a non-technical entrepreneur who wants to learn how to integrate AI into your business, you'll love it.

Be back next week.

-Corey