I’ve wanted my own career coaching platform for years. I could picture how it would work, but the cost and complexity always made it feel out of reach. Even after getting it built, I’d still have to maintain the thing.
With GPT-5.6 and GPT-6 Astra in ChatGPT, I got a working version of Make Me Pro online in a few days. It resembles a junior LinkedIn for my coaching business, with career records, resume tools and coach access.
It’s jarring how straightforward much of this was.
xAI explored a similar idea during its September 15–17 Grok Bot event, where three people set out to build a business in three days with AI.
More people can now attempt this, which also means more people can build something similar to yours.
These are the five things I’d pay attention to before assuming the software alone will make your business worth anyone’s time.
1. Start with work you understand
After more than ten years of career coaching, I wanted to spend less time moving information between documents and more time working with clients. I also wanted to make some of that expertise available to people who might never book a session.
Make Me Pro keeps someone’s career information in one place so it can support different tools without starting over. People can use the free tools without hiring me.
Astra understood that business model surprisingly well. I could explain how I work with clients, and we could discuss what belonged in the platform and what still needed a conversation with a coach.
Having experience gave me something useful to bring to those discussions. I wasn’t relying on the AI to tell me how my own business should work.
2. Learn enough technology to stay involved
I previously wrote about building Vino Tracker with Claude through the Vibecode app. With this project, Astra and the tools around it gave me much more control.
We started in ChatGPT Sites, which lets you build and publish interactive websites. We then moved the code into GitHub and hosting to Cloudflare, where code changes can automatically build and deploy.
ChatGPT also walked me through setting up Clerk, the service handling accounts and Google sign-in.
I could describe a change, review it and get it onto the live site without personally writing all the code.
Sometimes that meant explaining a workflow. Other times it meant sending a screenshot and asking why something looked so bad.
You don’t need to be an expert software engineer, but having a technical thumb helps.
Understanding API keys and how changes reach production saved me some confusion. I’ve worked in technology for years, and I wouldn’t pretend that didn’t help.
3. Build around what customers will actually do
Most of my clients don’t want another profile to maintain. They already put off updating their resumes.
I could build the nicest career-record editor in the world and still have trouble getting people to use it.
So we built a coach workflow around the conversation we were already having. Discovery notes, including notes from a transcribed meeting, can become proposed updates to someone’s record. We also had to work through coach permissions and make sure clients were notified about changes.
The client doesn’t have to repeat the conversation as a data-entry exercise afterward.
That feature came from knowing how clients behave.
A profile editor would have satisfied the basic requirement, but it would have left the most annoying part of the work with the person paying me to help.
4. Be painfully specific about quality
I spent a ridiculous amount of time sending resume exports back for changes.
The text was cramped, dates weren’t aligned properly, and page breaks looked awkward. The Word and PDF versions needed to look like the same document.
I’m putting my coaching business behind these tools.
Someone shouldn’t download their resume and immediately need to fix the formatting. (This is where I think a lot of AI resume services get it wrong. Too many feel like churn-and-burn shops that leave you to clean up the details yourself.)
We also had an import put “core competencies” under projects instead of skills. Anyone who works with resumes would know that was wrong.
Your experience and taste matter here. You have to inspect the actual output, explain what needs changing and keep at it. An AI telling you a feature is finished doesn’t mean it’s ready for a client.
5. Plan for the work after launch
Astra helped me work through pricing and access rules, along with privacy, security and draft legal language. Those areas still need proper review, but it helped me consider decisions I might otherwise have postponed.
There are ongoing costs too. I build through ChatGPT Pro, which has $100 and $200 monthly tiers. During this project, I hit my usage allowance and had to wait for it to reset. The application’s AI usage is billed separately from that subscription.
For the beta, we set a $100 monthly AI budget, with daily caps and per-user allowances. Manual edits and downloads don’t count. That’s the AI budget, not the full operating cost, but it keeps one enthusiastic user from spending the month’s allowance rewriting their resume.
I’d also like agents to help with customer service and social media. We started with Pixel, a visual QA bot intended to check the site after deployments.
Getting it reliably into the private screens became a rabbit hole. We paused it because I wasn’t convinced the benefit justified the setup yet. For now, feedback from people using the beta is a better use of my time.
Give yourself a weekend to try building one useful piece of your idea. Show it to someone who would actually use it and ask what they’d change. You don’t have to decide whether you’re starting a business before you get that far.