I built with AI, but backed it out of the core user experience after talking to real customers.

by

I built my wedding planning app with a lot of AI help. I’m a former wedding coordinator, not a software engineer, so Claude Code and Codex were a huge part of making it real.

But the more I talked to actual couples, the more I realized I was deeper in the AI adoption bubble than many of my users.

Some didn’t want an “AI wedding planner” making decisions with them. Some didn’t trust AI period. Others had tried planning just using ChatGPT, but got generic answers because they didn’t know how to prompt well. So I backed AI out of the core product more than I originally expected.

What I ended up building was an app handles the structured work (budget math, pricing benchmarks, quote decoding, guest tracking, RSVP tools, and timeline scaffolding) and an optional AI chatbot feature through a prompt builder. If/when a couple wants to talk an AI wedding coach, the app builds a guided prompt with the couple’s real plan context already included, then sends them to the custom Wedding Muse GPT or lets them copy it into whatever chatbot they prefer.

That choice feels less flashy than “AI does your whole wedding,” but for my audience, I think it’s going to make it feel more trustworthy. is live on Product Hunt today. I’d love feedback or thoughts from other builders thinking about where AI should be visible vs. where it should quietly support the product.

12 views

Add a comment

Replies

Best

this is the right call most teams will not make. ai as the engine and ai as the experience are two different products. your customers told you what they actually wanted (human judgment from a former wedding coordinator) and you listened. the play is keeping ai in the operational layer (search, drafting, sorting) and putting the human judgment in the user-facing layer. on tam network we shipped an agent identity row that names which parts of the work are ai and which are the human, specifically because customers were starting to discount everything as probably ai-written. transparent ai plus visible human judgment wins.