38
39 Comments

I'm 17 & I built an "Agentic AI" to kill the 15-minute Shopify listing grind

Hey hackers,
I’ve been obsessed with the operational side of Shopify lately.
One thing that consistently blows my mind: even in 2026, most merchants are still spending 10-20 minutes manually copy-pasting data to list a single product. It’s a massive growth killer.

I built Filleo to turn that into a 15-second task. It’s an agentic AI that takes messy product data (supplier sheets, PDFs, raw notes) and transforms it into a ready-to-publish listing.

It handles the titles, formatted descriptions, SEO tags, HS codes & everything else automatically.
The Tech/Strategy: I’m currently in the pre-launch phase, focusing on the "high-frequency listers" who add products one-by-one.

Try The App For Free: https://apps.shopify.com/filleo

30s Demo shows it all: https://youtube.com/shorts/K_uncUg2NVc

I’d love your feedback on:

  • Would You Buy It Starting At 1.99$ for basic plan?

  • Do You Want to try it for free to test it out? If yes then click the above URL

  • Does the need to clean up messy data feel like a real pain point?

Happy to answer any questions about the build & get feedback!

posted toAvatar for product Filleo
Filleo
  1. 2

    17 and already thinking in agentic workflows — respect.

    One suggestion: "15-second task" is the real hook, lead with that even harder. The $0.99 early access is smart urgency but 47/100 spots — consider showing a live counter on the landing page, converts better.

    Answered your question: most manual task AI hasn't solved well yet? Cloud cost anomaly detection that understands context — not just thresholds. Still mostly manual for me.

    1. 1

      Hey, really appreciate you taking the time to share this super helpful info.

      You’re right about the “15-second task” angle, I think I’ve been underplaying that a bit. I’ll bring it much more front and center since it’s honestly the strongest part.

      And yeah, the live counter idea makes a lot of sense. Right now it’s static, but making it dynamic would definitely add more urgency and make it feel more real — I’ll look into implementing that.

      Also your point about cloud cost anomaly detection is interesting. Makes sense since most tools just flag spikes but don’t actually explain why, so you still end up doing the manual digging. That “context-aware” layer feels like a real gap.

      Thanks again for the thoughtful feedback 🙌

  2. 2

    the 10-20 min copy-paste tax is one of those small repetitive bottlenecks that's perfect agent-bait. one ask: what's your retry/recovery story when a listing fails halfway through? that's where these agents earn their keep at boutique scale. 17 and shipping > a lot of us at 30+, keep going.

  3. 2

    the 10-20 min copy-paste tax is one of those small repetitive bottlenecks that's perfect agent-bait. one ask: what's your retry/recovery story when a listing fails halfway through? that's where these agents earn their keep at boutique scale. 17 and shipping > a lot of us at 30+, keep going.

  4. 2

    Great idea! Best of luck!

    1. 1

      Thanks for your support!

  5. 1

    This is actually really impressive, especially at 17 👀
    How did you build the AI part – did you use existing APIs or train something yourself?

  6. 1

    17 is crazy.

  7. 1

    But success will depend less on “agentic AI” and more on reliability, edge-case handling, and seamless Shopify workflow integration.

    1. 1

      Totally agree with you! The video above shows exactly how well it integrates with Shopify to get you the idea. Feel free to join the waitlist if you're interested in getting early access

  8. 1

    Love the 'multi-cycle' approach you mentioned. Cleaning the data first, then feeding it back in for generation is exactly how you handle the messy edge cases that trip up single-pass agents. That retry logic is where the real value lies for high-frequency listers. How are you handling the context window limits if a supplier sheet has hundreds of SKUs?

  9. 1

    Good stuff man, love to see a fellow 17 year old like me trying to achieve.

  10. 1

    Cool idea! What's been your biggest challenge getting

    the first users?

  11. 1

    This is a sharp problem to tackle — messy supplier data is exactly where most “AI listing tools” fall apart.

    How stable is Filleo when the input formats vary wildly between vendors?

  12. 1

    curious what the agent's instruction set looks like once you're handling 100+ product types. the reusable instructions - 'what counts as a good listing' - are where these break down. everyone's thinking about the AI model, nobody's thinking about the instruction set.

  13. 1

    Respect for building and shipping this early.

    One thing I’m curious about — was the idea driven more by a real workflow problem you faced, or by exploring what’s possible with AI agents?

    I’ve noticed a lot of AI tools struggle not with capability, but with fitting into actual user workflows. How are you thinking about that part?

  14. 1

    This is a solid problem to go after — that “10–20 min per product” grind is real.

    What I find interesting is that even when tools exist, people still stick to manual workflows because of trust and edge cases in the data.

    Curious — how accurate is it right now with messy inputs? Like, do users still need to heavily review/edit before publishing?

  15. 1

    Hey, same here — I actually built an AI tool myself when I was 18. It’s a highly personalized outbound email tool that consistently lands in the primary inbox. I’m offering a free trial right now, so if you’re interested, just let me know and I’ll unlock it for you with no time limit.

  16. 1

    The "high-frequency listers" focus is smart. Trying to serve every Shopify merchant is a trap — the person adding 50 products/week feels this pain 10x more than someone adding 2/month.

    One thing I'd watch: the $0.99 early bird might get you waitlist signups but could anchor your price so low that upgrading to $7.99+ feels like a 8x jump. Have you thought about offering the first 3 listings free instead? That way people experience the value without you training them to expect $0.99 forever.

    Re: "would I buy it" — I'm not a Shopify seller, but I built something similar for launch marketing (turning raw product info into press pages + social copy). The pattern is the same: people will pay to skip repetitive formatting work. The 30s demo sells it better than any copy could.

    What's your conversion rate from waitlist signup to actually trying the demo?

    1. 1

      Really appreciate the suggestion. every user gets a free trial of first 12 products, simply to test out if the app works well for them. After that, if they’re interested, they can move out to one of the paid plans. Also, a thing to note is that 0.99$ will grant the user generation of 100 products per month, but if it is a larger store that uploads products more frequently, it will have to upgrade to a higher plan accordingly.

  17. 1

    This is a really strong problem to go after—anyone who’s touched Shopify knows how painful listing products can get at scale.

    I’m curious about one thing though: how well does it handle inconsistent supplier data? Like when different vendors format things completely differently or leave gaps—does your AI standardize that reliably or does it still need manual cleanup?

    Also, the $0.99 early access is smart for traction, but I wonder if positioning it around “time saved per 100 listings” might convert even better—because the value here feels very ROI-driven.

    I’m working on something in AI for fashion visuals, and messy input data has been one of the hardest parts—so this definitely resonates.

  18. 1

    "Really interesting approach. The manual listing problem is real — I've seen it firsthand with small Shopify stores. Curious how you're handling the edge cases where supplier data is completely unstructured, like handwritten notes or weird Excel formats. Also validating my own idea right now (client vetting tool for freelancers) so I know how valuable early feedback is — good luck with the waitlist!"

    1. 1

      Basically, the AI agent runs not once but multiple times in cycles:

      1) It cleans the data in first cycle

      2) it feeds the structured data into the agent as a prompt in the next cycle

      3) this happens in parallel for each of the fields individually to get a personalised tailored to your store product listing

      1. 1

        Running multiple cycles with parallel field processing is smart — reduces hallucination per field. How are you handling cases where the source data is so messy that the first cleaning cycle produces garbage that then propagates through the rest? Does it fail gracefully or does the whole listing come out wrong?

  19. 1

    best of luck!

    1. 1

      thank you ❤️

  20. 1

    Great job. This is the realistic approach.

    1. 1

      Appreciate that

  21. 1

    Great job at 17ysd. My concern is why shpify did not enable this feature. What’s the moat here? Maybe customers are not that happy with ai features or the products updating frequency is not that high?

    1. 1

      Great question. Shopify has basic auto-fill but it's pretty limited—doesn't handle messy supplier data well or adapt to different product types. The real issue is when you're importing from suppliers/manufacturers with inconsistent formats. Our edge is handling the messy cases Shopify's native tools can't. But you're right to ask—it's why I'm talking to merchants first to make sure this actually solves a real pain point vs. a nice-to-have

  22. 1

    I saw how Filleo cuts the Shopify listing grind down to 15 seconds—that's a brilliant niche to tackle at 17. Since you're in the pre-launch phase with growing traction, would you be interested in discussing how to scale the distribution of your product to reach even more high-frequency merchants?

    1. 1

      Thanks for the kind words! Right now I'm focused on talking directly to merchants to nail product-market fit before scaling distribution. Appreciate the offer though!

      1. 1

        That makes total sense — focusing on PMF first is the right move.

        What we’ve seen is that once you start getting consistent signals, having a distribution layer ready can really accelerate things.

        Happy to reconnect when you get there — would love to see how Filleo evolves.