Building ZOL AI taught me something unexpected about AI products
I think most AI startups are building for the wrong users.
A lot of products today are focused on power users who already understand prompts, workflows, and AI tooling.
But small business owners usually don’t care about any of that.
They just want outcomes.
While building ZOL AI Studio, I noticed fashion sellers don’t actually want to “learn AI.”
They want:
better product photos
faster catalog creation
lower content costs
less operational work
That completely changed how we started designing the product.
Instead of making users write prompts or configure settings, we’re trying to reduce the experience down to:
Upload clothing image → generate usable campaign assets.
The biggest lesson so far:
Sometimes the real product innovation is removing complexity, not adding more AI features.
Curious if other founders here noticed the same thing while building AI products?
Yes, the 'outcomes not prompts' insight matches exactly what I'm seeing across early stage AI investments. The fastest-growing AI startups right now are the ones where users don't even know they're using AI, they just see results. Products that frame themselves as 'use our AI tool' force users to absorb a new mental model before they get any value, and most users never make it past that wall. The related lesson for small business owners specifically: the magic moment isn't 'wow this is impressive,' it's 'I just saved 90 minutes and didn't have to think.' If your AI delivers the second feeling, you don't need to teach anything. Worth optimizing your first-run experience around 'time saved on the actual workflow' rather than 'showcase what the model can do.'
This is incredibly well said — especially the part about users not wanting to “learn AI” before getting value.
We’ve been noticing the same pattern while building ZOL AI Studio. The strongest reactions usually aren’t “wow the AI is cool,” it’s more like “wait, this saved me hours.”
That shift in thinking has honestly changed how we’re approaching onboarding and workflow design now. Trying to make the experience feel invisible and outcome-first rather than AI-first.
Really appreciate you sharing this perspective 🙌
On the power-user retention concern: what's worked without losing them is the simplest workflow as default, with controls a click deeper. New users get working output with zero configuration. Power users find the deeper options when they want them. The product stays outcome-framed for the buyer, but the controls are there for the user who wants them.
That makes a lot of sense. I think balancing simplicity for first-time users while still supporting deeper creative control is probably the hard part in products like this.
Right now we’re trying to keep the default workflow extremely lightweight, then gradually expose more controls around pose, lighting, framing, etc. only when users actually need them.
Really appreciate these insights — this has been genuinely helpful from a product thinking perspective 🙌
ran into this same thing. I’ve shipped it both ways. ‘AI insights dashboard’ = zero engagement. ‘here’s where your team is overloaded’ = people screenshot and share it.
This is actually a really strong insight, and it matches what we’re seeing across a lot of AI tools right now.
The gap between “AI builders” and “real-world users” is still huge. Most users don’t want to interact with models, prompts, or workflows they just want a finished result that solves a problem in the shortest possible time.
The approach of going from upload → usable output instead of configure → experiment → refine is exactly where a lot of successful AI products seem to be heading.
Reducing friction often ends up being more valuable than adding more capability.
Same lesson here building aisa.to (AI skills assessment through conversation). Our first version showed users all 5 skill dimensions, scoring criteria, the whole framework. Users were overwhelmed.
What actually worked: just have a conversation, then show one clear result — here's where you are, here's what to work on next. All the complexity still runs underneath but the user never has to think about it.
I think the reason so many AI founders resist this is because we're proud of the technical work. We want people to see the sophistication. But the user doesn't care about your model architecture any more than they care about how their car engine works. They just want to get somewhere.
Curious what your retention looked like after simplifying the flow — did usage actually go up?
This hit home — especially the line about small business owners not wanting to "learn AI."
So many AI products today force the user to become a prompt engineer just to get a decent output. But a fashion seller doesn't want to tweak parameters — they want the photo to just work.
What you said about "removing complexity, not adding more AI features" is something I wish more founders understood. The best AI is invisible AI.
One question I'd love to ask: when you made that shift from "AI tool for power users" to "upload → generate campaign assets," did you lose any early users who were power users? And how did you handle that trade-off?
Also, I'm Bexra — helping entrepreneurs find, build & grow. Really appreciate you sharing this lesson publicly. It's the kind of insight most people only learn after burning months on the wrong direction.
In my experience with digital marketing, this is what always makes the difference. Users often care more about the results and they don't care about the metrics as much. If you can gather 10 backlinks in a month, it makes zero difference to a business owner, but if you can get sales, they'll be the happiest. So, yes your inference is correct and that's how real products and service-based businesses thrive, by driving results that actually matter to business owners.
The "outcomes not features" reframe is the one most AI product builders resist because removing complexity feels like shipping less. But the fashion seller who just wants usable campaign assets from an upload isn't failing to understand AI she's correctly identifying that the AI is not the product. The outcome is the product. The same tension came up building the AI writing assistant in ReleaseLog, founders don't want to prompt their way to a polished changelog entry, they want to type rough notes and have something publishable come out. The less the AI is visible, the better the product feels. What's been the hardest complexity to remove so far?
A lot of founders accidentally build for Twitter power users instead of normal business owners. Like that you build this ZOL AI Fashion Studio to solving the real pain point its a top notch.