I’ve got an idea… tell me why it won’t work (or why it might)

To post in this sub-forum you must have made 100 posts or have Trust status or have completed our ID Verification

Forums › Startups › I’ve got an idea… tell me why it won’t work (or why it might)

yudumcan

105 posts

Master Geek
+1 received by user: 39

ID Verified

#324640 7-May-2026 09:05

hi all, I'm the one who was planning to take on Trade Me and FB Marketplace ...but that dream didn't last very long. I’ve been working on something else since and keen to get your thoughts before I go too far down the rabbit hole and tell me why this is a bad or a great idea.

It's basically an AI tool for landscapers and garden designers. They take a photo of a client’s yard and later upload it to generate a realistic concepts (with prompts and image generation) then directly from the platform send it to the client for feedback. The goal is to speed up that whole back and forth and spending hours in Photoshop or doing hand drawn sketches.

I’ve spoken to four designers so far and all of them are still doing things the slow way. Using Photoshop and start mocking things up or sketching by hand. It takes time and clients still struggle to picture the end result. And making further changes are as painful.

Yes, there are AI tools out there but from what I’ve seen they’re pretty generic. You can get something back but it often doesn’t reflect NZ conditions. You get random plants, weird spacing and layouts.

The plan is to build this in stages. First step is just validating the idea and whether people would actually use it. If that works, next step could be pushing that job straight into their job management tool, pulling in rough plant and hardscape costs from local suppliers and helping them put together a quote with materials and labour. Feels like there’s a lot you could build on top once the core idea works, it's just selling the vision.

So, I'm keen to check what you all think? Does this sound useful (think about the future iterations) What would make it a killer idea from the start? If you know someone in this space I’d love to hear from them too.

Thanks

gehenna

8673 posts

Uber Geek
+1 received by user: 3900

Moderator

Trusted

Lifetime subscriber

#3488358 7-May-2026 10:17

You need your idea to have something that the customer can't do themselves. Your value add might be more effective prompting to get a reliable output, but I don't agree with what you've observed of other existing AI tools (noting it's subjective). IMO there's nothing you describe that a semi-digitally-literate customer can't do in the same scenario. You'll need to figure out your hook. Why do I use you instead of the "good enough" I can do myself for no/low cost? Once you convince me (i.e. customers) why I'd do that, you're on your way.

CokemonZ

1116 posts

Uber Geek
+1 received by user: 315

#3488374 7-May-2026 11:08

If you could input soil types, take into account shading, and growth characteristics, and pull through NZ plants as well as non natives could be cool.

My parents get their places landscaped when they move, and a large reason is so the right plants get put in the right spots for size and survivability.

The impression I get it's a bit more complicated than a pretty picture.

yudumcan

105 posts

Master Geek
+1 received by user: 39

ID Verified

#3488393 7-May-2026 13:04

gehenna:

You need your idea to have something that the customer can't do themselves. Your value add might be more effective prompting to get a reliable output, but I don't agree with what you've observed of other existing AI tools (noting it's subjective). IMO there's nothing you describe that a semi-digitally-literate customer can't do in the same scenario. You'll need to figure out your hook. Why do I use you instead of the "good enough" I can do myself for no/low cost? Once you convince me (i.e. customers) why I'd do that, you're on your way.

Yeah, someone could go to chatgpt and get something good enough. What I’ve seen is how much effort it takes to get there. You need to know what to ask, tweak it a multiple times and even then the result can drift. We have built a system that already knows what sort of project you're working on (I won't tell how :))

So this isn't just generating an image, it’s producing it predictable usable outputs. Instead of fiddling around trying to get something decent, the output is closer to something you could show to a client. We're also planning to implement some extra features like a estimate cost of each plant find it for you, creating a job in your job management tool, like Fergus, sending an email to client with the generated image attached with your logo and brand.

Our killer feature is the accuracy and the knowledge of AI that we'll use. it's going to have knowledge of all NZ plants, trees, which environment they live in the best, how far they should be planted apart, every other feature will be added on top.

Like I said the people I talked to have used AI before, they're semi-confident but the time they spend to explain AI where the feijoa tree needs to go while breaking other things in the image, they rather jump to Photoshop and do it there even knowing it'll take them longer.

yudumcan

105 posts

Master Geek
+1 received by user: 39

ID Verified

#3488394 7-May-2026 13:06

CokemonZ:

If you could input soil types, take into account shading, and growth characteristics, and pull through NZ plants as well as non natives could be cool.

My parents get their places landscaped when they move, and a large reason is so the right plants get put in the right spots for size and survivability.

The impression I get it's a bit more complicated than a pretty picture.

Shading and growth characteristics wasn't included in our AI structure I'll make sure it's added. thank you ;)

muppet

2665 posts

Uber Geek
+1 received by user: 1681

Trusted

#3488411 7-May-2026 14:44

I think you'd be mad to build a business on AI like this. At the moment the good AI is with the big players, but some of the local AI you can run is getting very, very close. In ~5 years I expect you'll have models on your computer where you can do your idea, and 2000 other things, all without needing "cloud" AI or someone's product.

History is littered with businesses who built atop of someone else's platform who then cried foul (and closed down) when that platform closed its API, or put rate-limits in place, or jacked the prices up x 50.

Basically (correct me if I'm wrong please) your idea is just to write a very very customised (New Zealand specific) prompt that will do the hard yards AI magic, right?

I'm sure there is short term money to be made while the tech is still new, but I don't think you've come up with a product that has much of a real growth curve or future.

Personally I'd think the better market would be the "I know landscaping and I know the right prompt and things to ask an AI to get useful ideas/visualizations that match New Zealand conditions" - you control the experience for the customer much better then. Based on your post and the previous one I get the idea you're more after a "thing to build that just makes money" so the "going out and talking to people" consultancy idea probably doesn't land either.

Anyway, that's my thoughts. I'm old, bald and grumpy though, so keep that in mind when considering my position!

roobarb

733 posts

Ultimate Geek
+1 received by user: 731

Trusted

#3488934 8-May-2026 19:27

I do wonder how did people like Capability Brown ever manage without AI. I particularly liked the reference in that link to shrubberies

If you build anything with AI, what is your property that you own and can license, copyright etc?

What stops anyone else simply copying exactly what you did? There are thousands of small utilities that were created by enthusiasts and then simply copied or cloned into products/components of companies with deeper pockets.

I suggest you may just end up being the unpaid researcher building a prototype that gets copied wholesale by someone else.

My understanding is that in NZ you can't patent software, and "do X but using AI" is a thousand times worse than "do X but using a computer", because with AI you can't actually explain how anything is done. In reality it is "do X by asking somebody else to do it"