Skip to content
Notes archive
NotereflectionsEnglish

RuleFrame: From AI-Assisted Research to a Digital Product

RuleFrame began as an AI-researched opportunity for an online business, leading to practical tools and paid templates without revenue so far.

This article was drafted with AI assistance and reviewed by Haobin Li.

RuleFrame is one of my first experiments in earning money online. Its topic came entirely from research and recommendations made with an AI agent. As of November 15, 2026, the project has not earned any revenue.

I want to describe that beginning plainly. This was an opportunity I chose to investigate and build around, rather than a product that grew out of an established professional practice in AI governance. The research helped me identify a possible direction; it did not establish that the direction would become a business.

From recommendations to a product idea

I asked an AI agent to research topics that might be commercially promising. Among the possibilities it returned, I selected some for further research. RuleFrame grew out of that process.

The appeal of using an agent this way is that it can help turn a broad question into a set of concrete possibilities. Each possibility still needs scrutiny. An account of a market can sound coherent without demonstrating that a specific audience will trust a new product or pay for it.

That distinction matters especially here. The product concerns how small teams use AI, which brings questions about sensitive information, internal rules, review, and responsibility. A business opportunity suggested by research is not a basis for claiming legal expertise or guaranteeing compliance.

The project therefore has to stand on the usefulness and limits of what it actually offers.

A free-tool and paid-template experiment

RuleFrame combines free tools with paid downloadable templates. The free tools turn structured answers into policies, checklists, registers, and other practical documents. The paid packs offer editable materials people can adapt to their own circumstances.

The business hypothesis is that a useful free result might help someone understand the product and consider a more complete template pack. That is the intended path, not a claim about purchases that have happened. So far, there has been no revenue.

The downloadable-product model also sets a boundary on the implementation. The site does not need to become a full workspace application just to provide a document. It can deliver a focused result without building accounts, a database of customer policies, or an ongoing monitoring service.

AI in the research, deterministic logic in the tools

There is an important distinction between how the idea was researched and how the published tools work. AI helped with the opportunity research. The free tools themselves do not call an AI model to generate their results.

They use structured inputs, validation, and deterministic generation. That makes the relationship between an answer and the resulting document inspectable. It also makes it possible to test how particular combinations of answers change the output.

The free-tool inputs are processed in the browser. Readers can copy their result or use the browser-based PDF output flow without submitting those answers to an AI service. This describes the document tools; buying a pack is a separate interaction with the checkout provider.

The paid materials have a different delivery path. A build process assembles the downloadable documents, and the storefront handles access to those files. Keeping those responsibilities separate avoids treating every document interaction as a reason for more backend infrastructure.

A useful document still has limits

A generated policy or checklist is a starting point for a human decision. It does not establish that a business follows the policy, that an AI use is safe, or that legal requirements have been met.

That affects the language of the interface and the outputs. Questions should make assumptions visible, and results should preserve open issues rather than turning incomplete answers into false certainty. The product offers practical operational guidance, with review and adaptation left explicit.

What the experiment has shown so far

RuleFrame has gone from a research recommendation to a published product with functioning tools and downloadable packs. That demonstrates the ability to build and deliver the offering. It does not yet demonstrate a viable business.

The absence of revenue is an important part of assessing it. I can describe the research process, the product decisions, and the engineering work without presenting the original recommendation as validated.

For me, RuleFrame is a record of trying to bridge that gap: using AI to discover a possible opportunity, choosing to pursue it, and finding out what remains uncertain once the website exists.