Building a product around a friction I kept noticing

Tap to Share began while I was researching NFC refrigerator magnets for KikiMajo Studio, my creative side project. NFC was already being used to open Google Review pages, which removed the work of finding the right link.

But navigation was only one barrier. A customer still had to decide what to say, write it, find usable photos or other material, choose a platform, and spend more time after the experience had ended. I recognized that pattern in my own behaviour: I could genuinely want to support a business and still put the task off.

There was also a merchant-side problem. Social participation campaigns often depend on staff repeatedly explaining the rules, where to post, what content is available, and how to transfer images or posters. When a social campaign uses an incentive, staff may also need to manage fulfilment. That is separate from voluntary reviews and does not mean rewarding positive Google reviews.

Some willingness to share may already exist at the moment a customer is especially satisfied. The barrier may be effort rather than motivation. Tap to Share is a way to test that possibility without assuming it has already been proven.

This artifact is supporting context rather than the main product evidence. It shows where the question started.

I had seen the other side of this problem at BabySquare. A dissatisfied customer publicly criticized an after-sales experience even though the store had been helping with the manufacturer's warranty process. The identifying details are not part of this case; the useful lesson was the imbalance in who feels motivated to speak.

The first Framer version was intentionally simple: give customers a ready-to-use caption so they did not have to write from scratch. It made the flow faster, but introduced another problem. If customers received similar copy, their posts could begin to look coordinated rather than genuine.

The preset version made sharing easier, but the content was too predictable. Generating different captions with AI was the obvious next step. The harder problem was making those captions sound appropriate to the person, context, language, and platform rather than like variations from the same AI system.

Different wording does not automatically mean natural wording. Generic AI output can still repeat familiar structures, use broad marketing language, or sound overly polished for the platform. Tap to Share therefore uses a controlled caption-generation system rather than a simple prompt-to-caption flow.

Before this workflow, moving an idea toward a testable product involved two kinds of cost: research friction from gathering and sorting information, and implementation friction from technical labour, handoffs, waiting, communication, and revision.

The research step includes structured market exploration, query and keyword discovery, search-intent analysis, and API-based research tools such as Monid. Those tools help me explore the category and its language; they do not make the product decision. Judgment stays in the middle of the workflow.

I own the research direction, product decisions, MVP scope, requirements, AI behavior and rules, QA, and iteration. AI helps with research, analysis, and structuring specifications, while Codex handles the coding implementation. That lets me move through build, test, and refinement with fewer external handoffs.

AI also changed the economics of experimentation. Work that would previously have required external development time, communication, and repeated revisions became something I could test independently before committing more resources.

As the product evolved, I started thinking beyond the caption itself. Merchants needed ways to guide what customers could share, while the interaction could also record what customers selected and did within the flow.

This self-service flow may reduce some repeated explanation and content handoff for staff, although it does not remove every operational task. The value may also extend beyond the resulting posts: observed participation can show what customers select, which media they use, which platforms they open, and which touchpoints generate activity.

The growth system is now being built around the product. The sequence is observation, potential signal, marketing interpretation, then decision—not preference inferred from a single action. SEO/GEO, acquisition, and conversion performance will be evaluated once there is enough live data to measure it.