My AI-first method for validating a product that only exists in my head

Creating a product is easy. Creating a product that people actually want is another story.

When an idea for a new product comes into my head, before writing a single line of code or designing a single screen, I run it through a simple method to confront an early-stage concept with the real world, even if it’s still just vaporware.

Let me show you exactly how I do it, with the help of mighty AI tools and with a real example.

Meet Storymate, a window into the creator’s mind

A few months ago, my friend Vince and I became obsessed with a specific problem we kept seeing among solo creators and small video teams: they all filmed themselves with the iPhone’s rear camera (for better image quality) but they had no idea what they were capturing. They'd press record, hold the phone awkwardly hoping they'd be in frame, then spend hours rewatching and deleting unusable takes.

Add to that the stress of forgetting their script, inconsistent framing, messy focus, and you've got a frustrating and time-wasting workflow for anyone trying to create professional-looking content.

That insight led to a hardware product idea: Storymate.

What is Storymate?

Storymate is a compact hardware screen that magnetically snaps on the back of your iPhone. It acts as a rear-facing viewfinder, showing a miniaturized version of your screen into a physical display, letting creators see what they were filming in real-time. No need to charge it, just plug it using USB-C and you're good to go.

But that was only half the story.

The real power lies in the software layer, a powerful embedded software combined with a custom camera app featuring AI features:

• AI-generated script templates to stay on-message,
• on-screen framing guidance,
• automatic scene detection and quick edits,
• a smart teleprompter controllable by gestures,
• and a suite of feature dedicated to make the content creators' workflow more streamlined.

In short, Storymate was designed to let creators feel in control, save time, and deliver better content without needing a crew or reshooting endlessly.

But before building anything, we needed to validate one thing: would content creators care? And more precisely: which part of the product actually mattered to them?

TL;DR - my process for validating a product that does not exist

1. Fix only one variable to test
2. Run three ad campaigns, each with a distinct USP
3. Create one landing page per campaign with polished, realistic visuals (GenAI + 3D)
4. Track CTR and email conversion
5. Start conversations with leads via email through an automated funnel
6. Do 1-on-1 interviews to extract real insights
7. Store and analyze everything
8. ...
9. Profit!!

Step 1: define what you want to test

So we went to work and unfolded the following framework. But the first question we asked ourselves was: what should we test?

One of the biggest mistakes PMs make is trying to test several things at once. Message, price, visuals, CTA, branding... and in the end, they’re left with Gloubi Boulga: a chaotic mix that tells them nothing.

Golden rule: fix one variable to test per iteration. One angle. One hypothesis.

Here are examples of testable variables:

• Unique Selling Proposition
• Copywriting
• Visual hook
• Landing page structure
• Tagline

For Storymate, we wanted to test which emotional hook resonated most with creators.

Step 2: make your product look real

Even if the product doesn’t exist yet, everything must look real. This isn’t the time for half-baked visuals, you have to convince people you are going to sell something and create a compelling website.

Please avoid stock photos as it's difficult to keep a cohesive brand with photos you took on Google Images. Instead, we used Midjourney and Visual Electric to generate lifestyle scenes.

If your product is software, time to fire up Figma and make the sleekest visuals you can. If your product is hardware, you have to bootstrap you way to an industrial design, then fiddle with 3D renders and Photoshop.

• Blender and Fusion 360 to make 3D product renders
• Photoshop to combine both into convincing hero visuals

So we created a bank of dozens of assets for Storymate: close-ups, lifestyle shots, usage demos. These assets powered not just the landings, but the ad creatives too.

Pro tip: hire a freelance designer on Behance or Upwork to mock up your product in 24 hours. It’s cheap and will make your tests 10x more credible.

Step 3: build three campaigns, each with one clear USP

We defined three potential USPs for Storymate:

1. “Stay in control” (focused on confidence and precision)
2. “Save time” (focused on productivity and efficiency)
3. “Deliver powerful messages” (focused on storytelling and clarity)

Each of these hooks has a different emotional charge:

• One targeted anxiety and control ("stop the guesswork when shooting")
• One targeted frustration with wasted time ("reduce your wasted takes by xx%")
• One targeted ambition and professionalism ("create your best stories")

So we went to work and created a fake website with three distinct landing pages and brand narratives, each focused on one USP.

How did we come up with the wording of the website? ChatGPT and Claude, obviously. We kept it short and sweet, and made sure to remove any trace of the em dash ("—") it's a dead giveaway that AI helped write it.

We then used no-code website builders like Squarespace or Webflow to create all three websites, each with its own URL. There's no need to create three separate brands (which would be a pain to manage), but we made sure to use a different domain name from our main one to avoid brand dilution.

Then, for each one, we ran a Meta Ads campaign with 3 elements (a reel, a story, and a static post), all driving traffic to the corresponding landing page.

To push the illusion further, we intentionally designed the pages to make it look like the product was available for purchase. When users clicked on the "Order Now" button, they were redirected to a page that said, "We're out of stock, sorry," and invited them to leave their email to be notified once it was back. This allowed us to measure actual purchase intent without having to ship anything... and get email addresses.

Step 4: measure clicks and conversions

The metrics we tracked were simple:

• Click-Through Rate (CTR) on the ads
• Email signups on each landing page

This helped us answer: what USP worked the best, who was clicking, and who cared enough to want more?

The results were clear. The "control" narrative resonated far more than the others. CTR was significantly higher, and people weren’t just clicking, they were commenting the posts, sharing them, and leaving their email. That was the signal we needed.

Step 5: qualitative discovery starts with one email

Once we had collected emails, we moved to the qualitative phase: real conversations.

We created a new domain just for outreach (never, ever use your main brand domain, or risk getting flagged as spam). Then we set up a simple automated email flow.

Before launching this outreach, I made sure to properly warm up the email address. Email warming is crucial: if you're using a new domain or address, email providers are likely to flag your messages as spam unless they see organic behavior first. We spent a few days sending manual emails to friends and colleagues, initiating real back-and-forth conversations. You can refer to this comprehensive guide from Za-zu on how to warm up email addresses (spoiler: it's long and thorough but one of my year's best reads). The goal is to make your domain look trustworthy before you start any cold outreach at scale.

Each email asked just one question. No surveys. No forms. Just a plain, honest "Hey, founder of Storymate here, I have a very important question: what is your biggest time-waster in your content creation process?"

To avoid spam filters, we used a language model to generate variations of the message ("CEO" instead of "founder", "flow" instead of "process"...). The meaning was the same, but the phrasing changed. Why? To avoid being flagged by email providers' spam detection systems. Again, it’s a cat-and-mouse game.

When someone replied, I took over manually. The goal: book a 30-minute call to dive deep.

Step 6: Call'em and let'em talk

These calls are the core of discovery. We asked open-ended questions like:

• What’s the most frustrating part of filming content?
• What do you wish you could do faster?
• How do you feel about using scripts?

We listened. A lot. And never asked yes/no questions (you should use them as little as possible), always trying to get a long answer from the client.

And when a user said, unprompted, "I hate not knowing if I’m in frame," we knew we were onto something.

Others told us:

• They wasted hours reviewing takes
• They struggled to sound confident on camera
• They avoided the rear camera because of uncertainty

Those insights were gold. They weren’t just opinions. They were real pain points.

Step 7: capture everything and analyze

We asked users if they agreed being recorded (most of the time they were) then transcribed the conversation with Whisper. We stored everything in Notion, tagged it, and highlighted recurring themes.

These transcripts became:

• Input for product roadmap decisions
• Testimonials and messaging inspiration
• Data to pitch potential partners or investors

When someone says, in their own words, "this would save me 20 minutes per shoot," that’s not a guess. That’s validation.

If you want to know whether your idea holds up, don’t pitch it to friends. Don’t pitch it to investors. Put it in front of strangers who owe you nothing. If it resonates, you’ll feel it.

And remember, the goal isn’t to “validate an idea.” It’s to build conviction. The kind that you can get back to when you're deeply in doubt.

Happy discovery!