The Custom GPT Hustle Is Real Money. It's Also a Countdown Clock.
- Rich Washburn
- 2 days ago
- 4 min read
I watched a clip on TikTok this week of a woman explaining her business to a host, and I need to talk about it, because it's one of the best real-world examples I've seen of exactly what I've been saying since 2023 — and it's also proof of something most people building with AI right now don't want to hear.
What She Actually Built
She sells custom GPTs — the ones you build on a standard $20/month ChatGPT account — to coaches, founders, and consultants. She's doing around $100,000 and wants to get to a million. Her offer stack: $497 to $997 for a branded custom GPT, a $997 four-week AI boot camp, and a $3,800-a-year AI platform membership with all her frameworks included. That's not a side hustle. That's a real business, built on a tool that takes minutes to configure, sold to an audience that doesn't know how to build it themselves and doesn't need to.
The most interesting part of the clip wasn't even the AI part — it was the business coaching moment. The host told her to kill the standalone $497 GPT offer entirely, fold it into the boot camp as a free deliverable, and raise the boot camp price instead, because for a founder or consultant, $4,000 a year is often the cost of a single customer. Then: drop the entry price to make the boot camp itself feel like a no-brainer pitch event, get more people in the door, and upsell from there. That's not an AI insight. That's just good offer design — and it's the same skill that will matter no matter what tool she's selling five years from now.
I'll be honest about my first reaction: a little jealousy. I was building and selling custom personas the same year OpenAI opened up custom GPTs, and I never packaged it or scaled it anywhere close to what she's done. So credit where it's due — she saw the opportunity, moved fast, and built a real audience and a real funnel around it. That's the hard part, and she nailed it.
I Was Doing This in 2023. That's the Actual Point.
Back in October 2023, right after custom GPTs launched, I wrote about something I called my "AI Staff" — a lineup of custom personas built as a spinoff of the ARIA project I'd already been developing. Jordan, the personal trainer. Michael, the financial strategist. Each one built with the prompting techniques available at the time, each one positioned as a specialized hire rather than a generic chatbot. I built dozens of these, sold some, gave plenty away, and used them as the testbed for what eventually became the much more capable version of ARIA I run today.
I say that not to claim I called it first — plenty of people were experimenting the same month. I say it because it's useful evidence for the actual lesson here: the tool she's built a six-figure business on on is now the same tool I was treating as an early prototype three years ago. What was frontier in October 2023 is, by mid-2026, something I'd call a screwdriver — a single-purpose tool that does one narrow job well and nothing else. Useful. Still sells. Not remotely close to where the technology actually is now.
Why "Screwdriver" Isn't an Insult
A custom GPT is a fixed persona with a system prompt and maybe a knowledge file attached. It answers questions in a specific voice. It does not remember you between sessions in any meaningful way, it cannot take action on your behalf, it cannot verify that what it told you actually happened, and it cannot operate anything beyond the chat window it lives in. That's not a criticism — it's just what the tool is. For a coach who wants a branded lead magnet, or a consultant who wants a client-facing FAQ bot with personality, that's genuinely enough. It's a real screwdriver doing a real screwdriver's job.
The frontier moved past that a while ago, into agents that hold persistent memory, execute real tasks across real tools, and check their own work before calling something done. That gap — between a fixed persona that answers questions and a standing agent that actually does the job — is the entire distance between where most people selling "AI products" are operating today and where the technology already lives.
The Actual Cautionary Tale
This isn't "I told you so," because being early to a trend that made someone else six figures isn't a flex — it's a reminder that being early and being loud about it aren't the same as building the business around it. She did the part I didn't: she turned a two-minute build into a boot camp, a membership, and a real funnel. That's the transferable skill, and it's the one worth actually studying here.
The warning is about the shelf life of the tool underneath it, not the hustle itself. Custom GPTs went from novel to mainstream to screwdriver-tier in about three years, and the interval between "brand new capability" and "commodity tool a mom-and-pop shop uses without thinking about it" keeps getting shorter, not longer. Whatever looks like frontier AI right now — the agentic, persistent, tool-using systems that are actually where the interesting work is happening — will make the same trip. Probably faster.
So build the thing. Use what's free right now — tools like NotebookLM can get you from zero to a genuinely deep, well-researched custom GPT in twenty minutes if you're willing to put in the work. Sell it, package it, build the boot camp and the community around it the way she did. Just don't mistake the tool for the moat. The tool is disposable. The audience, the trust, and the offer you built around it are the only parts that survive the next jump — and there's always a next jump.
Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital and CIO of Data Power Supply.
Comments