How I build AI products with no technical ability (and the portfolio course I wish more designers used)

A few months ago, I started building with AI as someone with zero technical background. I used a tool called Lovable which lets you turn natural language prompts into working prototypes, and ended up launching a product that now has real customers, revenue, and visibility.

Last month, the Lovable team invited me onto their livestream with 1,500 people watching to walk through how I did it. I shared my screen, talked through the thinking behind each project, and got flooded the next day with one simple question: “How do you even get started?"

That question is the real secret. I don’t begin with code, I begin with pain. Here’s the exact, unvarnished sequence I follow every time so you can apply it to your own ideas and start shipping AI projects in under an hour.

This post is not sponsored by Lovable btw.

1. Pinpoint the pain, not the solution

I spend a lot of time manually solving a small problem that’s bugged me or someone I know:

A project that always feels like busywork
A repetitive task that never quite works
A pattern I noticed talked about online that is related to what I know about

Then I go talk to people who share that pain via Linkedin/Reddit DMs or Slack communities and ask two to three questions:

Is this still happening for you today?
What are you doing right now to work around it (and how much time or money does that cost)?

That conversation is the research. If the pain is real and costly, you’ve just found your north star.

A DM I sent to a bunch of folks who attended one of my Maven lessons - I already knew they were likely feeling this pain and wanted to validate my idea further. Some of these DM's turned into the very first Backchannel customers when I only had a prototype and a Loom walkthrough.

2. Record a voice note of the problem

Once I’ve validated the pain, I open up Granola and riff for two to three minutes:

“A lot of people are stuck applying for jobs and getting nowhere. They are told to network, skip the traditional process but don't know how to do that. I've been doing it for over 10 years and always finding opportunities in non-traditional ways, I have written extensively about it and also coached people. I don't want to become a coach or create a course because then it will be prohibitively expensive for those who could benefit the most.”

That voice note captures my raw, unfiltered thinking. It surfaces assumptions I’d never notice in bullet points.

Part of the original voice note that became Backchannel

3. Turn your voice note into a mini-PRD

I paste the transcript into GPT or Claude with this prompt:

“You are a product manager, designer, and engineer. From this transcript, give me:

A one-sentence problem statement.

A two-paragraph spec describing the MVP.

A bullet list of the five core features needed for a first build.”

In seconds, I have a crisp PRD I can refine into my own words:

MVP Spec:
“A 7-step interactive program you follow over 14 days. Each two-day module delivers exactly what you need: mindset shifts, contact-finding hacks, outreach templates, proof signals, and a daily routine so you consistently spark real conversations with decision-makers.”

Core Features:

14-day scheduler with paired 2-day modules

Module content for each step (mindset reset, hiring signals, intro system, identity refresh, message formula, proof signals, opportunity routine)

Built-in templates and swipe files for intros and DMs

That document becomes my north star for every subsequent step. And this is where knowing the problem space and the pain becomes very critical because you have to review the output the AI is producing. If you're just building for fun (I built 60+ weird ideas before building a real thing) then it doesn't matter. But if you actually want to build something people will use, you need to catch the issues AI produces - and you can't do that if you don't really understand the problem you're solving.

4. Copy Lovable’s prompt instructions straight into AI

Lovable publishes a “Learn” page with the exact syntax and instructions their conding assistant needs. I grab that text, no rewriting, and then feed it to GPT asking it to create a prompt for Lovable.

Example of a recent test prototype prompt

5. Build, observe, iterate

I drop the prompt into Lovable. Within minutes, I have a working prototype. If you've spent the time to prepare for the build, 9 times out of 10 you'll get something working.

I then record another voice note narrating each inconsistency, talking through any UX issues or anything else I want to iterate on.

Then I feed that transcript back into GPT:

“Based on these observations now that the V1 is built, write the Lovable prompt to fix the bugs and refine the UI. Once that's done we can move on to the next phase of the project.

I rerun the prototype one more time and now I have a demo I can actually click through.

6. Share the raw build and ignite conversations

Finally, I screen-record a 60-second demo of the half-baked build and post it:

“Built this {thing you built} MVP in 45 minutes with Lovable + GPT. Raw demo below, what feature would you add next?”

Within hours, I get DMs from:

People feeling the pain and interested in using it.
Workshop organisers inviting me to teach this exact process.
Fellow builders pointing out edge-cases I hadn’t seen.

Those conversations turn into consulting gigs, small product partnerships, and invitations to bigger stages, none of which would have happened if I’d waited for perfection.

Why this isn’t “AI hype”

You start with real human pain.
You document raw thinking with voice.
You let AI spec and prompt, not invent features.
You iterate based on direct observation.
You build in public, inviting genuine feedback.

Done right, AI is a multiplier of clear thinking, not a shortcut around it.

Your first “half-baked” assignment

Tonight:

Pick a tiny pain point you face.
Record a 2-minute voice note riffing on that pain.
Ask GPT/Claude for your PRD with the prompt above.
Turn Lovable’s instructions and your PRD into a prompt.
Build!

Reply if you have any questions, happy to help. Or send me what you've created!

Mindaugas

If you’re working on your UX portfolio, I highly recommend my friend Aneta’s course:

UX Portfolio by Aneta Kmiecik

It’s packed with real examples, storytelling frameworks, and 48+ Figma templates to help you finally turn your messy projects into a senior-ready portfolio.

Use code MINDAUGAS50 for 50% off until May 31st.

→ Grab it here