I Spent 3 Years Building an AI Tools Directory That Runs Itself

There is a particular kind of frustration that comes from knowing a tool exists but not being able to find it.

In 2023, I was doing research for a client in the B2B SaaS space. They needed to know what AI writing tools were worth recommending to their marketing team. I spent the better part of an afternoon clicking through results that were either outdated, clearly sponsored, or so bloated with every AI tool ever created that they were useless as a decision-making resource.

Around the same time, I was reading about how directories and listing sites were one of the fastest-growing content formats in the SEO world. Not because directories are new, they are not, but because the AI tools space was moving so fast that no single publication could cover it comprehensively. There was a gap, and it was opening wider every week.

I had also been thinking about business models. Specifically, how to build something that generates revenue without requiring my constant presence. A directory where tool makers pay to be listed, where the listing process is automated, and where the product essentially sells itself once set up, that model had been in my notes for months.

Three frustrations. One product. That is how elloAI began.

The Problem I Was Trying to Solve

The AI tools landscape in 2023 was, and still is, even today, genuinely overwhelming.

New tools are launched every week. Some were serious products with real teams behind them. Many were thin wrappers around OpenAI’s API with a landing page and a Twitter account. A few were genuinely brilliant products that most people never discovered because they launched without a marketing budget or distribution.

For professionals trying to evaluate AI tools for specific use cases (writing, research, code generation, image creation, audio, video, data analysis), finding the right tool required visiting five or six different directories, none of which agreed on categories or quality standards. Most were ad-heavy. Several mixed free and paid tools without a clear distinction. A few listed tools had shut down months earlier.

The problem was not a shortage of AI tool directories. It was a shortage of ones built with the user’s decision-making process in mind.

A person looking for an AI writing tool does not need a list of 400 options. They need a curated, organised, up-to-date list of tools that have been vetted to at least the level of “this actually exists and works as described.” They need categories that map to how they actually work, not how a developer decided to tag things in a database. They need enough information about each tool to make an informed shortlist.

That was the gap I was building into.

What elloAI Is

elloAI is a curated directory of AI tools, organised by category, with a fully automated listing process.

Builders and companies who want their tool listed pay a one-time fee. Standard listings go live automatically. Featured listings get priority placement in their category, more visibility, and a dedicated profile. Neither requires any manual processing on my end once the pipeline is in place.

The categories are structured around how professionals actually use AI: writing, research, code generation, image and video creation, productivity, marketing, and a handful of others. Each tool gets a profile with a description, use case, pricing model, and a direct link.

For visitors, elloAI is a starting point. Not a comprehensive database of everything that exists, but a reliable first stop when you are trying to figure out which AI tool to evaluate for a specific job.

How I Built It

Year One: Architecture and the pipeline problem

The most important decision I made early was that I would not manually process any listing submissions. That decision shaped everything that followed.

If every listing required me to review a form, upload content, make a judgment call, and hit publish, then I had built myself a job, not a product. The value I was creating would be capped by my own time and attention. That is not a business model. That is freelancing with extra steps.

So before I designed a single page, I mapped out the automated pipeline. A builder finds elloAI and wants their tool listed. They fill out a submission form. Stripe handles the payment. A webhook fires on successful payment. GitHub receives the listing data. Vercel rebuilds the relevant pages. The tool goes live.

That pipeline took most of the first week to get right. Each piece worked individually, but making them communicate reliably (handling edge cases, failed webhooks, payment errors, duplicate submissions) required more iteration than I expected. The technical complexity was not in any single component. It was in the handoffs between them.

Year Two: The part I underestimated

The hardest part of building elloAI was not the automation. It was designing the category structure and the UI around it.

I had assumed this would be the easy part. It turned out to be the most consequential decision in the whole project.

The category structure determines how useful the directory is. If you put a tool in the wrong category, the right user never finds it. If your categories are too broad, the directory loses the signal that makes it valuable. If they are too narrow, you end up with dozens of near-empty sections that make the product feel underdeveloped.

I went through four or five different versions of the category architecture before landing on one that felt honest to how people actually work. The final structure is neither alphabetical nor technical. It is task-based. If you are trying to do a specific job, you go to the category that matches that job. That sounds obvious, but getting there required throwing away a lot of work.

The UI problem was related. Most AI tool directories I had used were either visually cluttered or so minimal that they felt like they had no opinion. I wanted elloAI to feel opinionated, like someone who had thought carefully about what a user actually needs when evaluating tools, and had made deliberate choices about what to show and what to leave out.

Getting that balance right took longer than the technical work. Design problems often do.

The first listing

The first paying listing on elloAI was a tool I almost did not list because I was not sure the pipeline was stable enough.

When that first Stripe payment came through, fired the webhook, updated the repo, triggered the Vercel build, and produced a live page with the tool’s details, without me touching anything, I sat with that for a moment.

That is what automation is supposed to feel like. Not magic. Just a machine working the way you designed it to work.

What elloAI Looks Like Today

elloAI is a live, indexed AI tools directory with listings across multiple categories.

The submission pipeline works exactly as designed. Builders who want to list their tools go through the process without any involvement from me between payment and publication. Standard listings and featured listings have different price points and visibility levels, and the distinction is clear to anyone going through the submission flow.

The site is built on a stack I use consistently across my projects: the frontend is fast, the hosting is on Vercel, and the content is version-controlled in GitHub. That means updates are trivial to deploy, and the infrastructure cost is low.

What I am still working on is distribution and growth. Building a product is the first problem. Getting the right people to find it is the ongoing one. elloAI has organic traffic from search, but the directory model works better with more listings, and more listings come with more visibility. That flywheel is still spinning up.

What I Learned Building This

Automation is a design problem before it is a technical problem.

The hardest part of making the listing pipeline work was not writing the webhook handler or configuring the Stripe integration. It was deciding what should happen in every possible state of the system, what happens if payment succeeds but the webhook fails? What happens if a submission has missing fields? What does the builder see at each step?

Those are design questions. The code answers them. If you have not answered them first, the code will make arbitrary decisions for you, and those decisions will create support problems later.

Categories are your product.

In a directory, the category structure is not a navigation decision. It is a product decision. It determines whether the right person finds the right tool.

Getting that wrong at the start would have been hard to recover from, because your early listings would have been placed under the wrong structure, and changing it later would have broken URLs and confused returning users.

I spent more time on categories than on almost any other part of the build. It was the right call.

The business model has to be in the product architecture from day one.

I did not add the payment layer after building the directory. The payment layer was the foundation around which the directory was built. Because of that, monetisation was never an afterthought. The product made sense economically from its first user interaction. That is the right order to build things.

What Is Next for elloAI

The immediate priority is distribution. elloAI needs more traffic and more listings, and both come from visibility.

I am working on a content strategy that brings in organic search traffic for specific AI tool categories (people searching for “best AI tools for market research” or “AI writing tools for SaaS marketing”) and converts that traffic into both visitors and listing submissions.

The longer-term direction involves improving the quality signal. Right now, the directory is curated in the sense that I am building it thoughtfully, and listings go through a submission process. But I want to add a layer of community-driven quality signals (upvotes, verified reviews, usage indicators) that make the curation more visible and more useful to visitors making decisions.

I also want to add a free tier with different limits, so that newer, earlier-stage tools can get listed without a barrier, while still making the paid tiers clearly valuable for tools that want more visibility.

The AI tools space will keep expanding. The problem elloAI solves (helping professionals find the right tool for a specific job without wading through noise) is not going away. If anything, it becomes more valuable as the number of tools grows.

What This Project Says About How I Work

elloAI was built to solve a real problem I experienced as a user, to fill a market gap I identified through research, and to test a business model I had been thinking about for months.

All three of those inputs were necessary. A product built only from personal frustration can miss what the market actually needs. A product built only from market analysis can lack the authentic understanding of the user’s experience that makes it resonate. A product built only to test a business model can end up optimising for revenue mechanics before it has earned the right to ask for money.

The combination (personal experience, market research, and deliberate monetisation design) is how I approach most of the products I build. It is not a framework I follow. It is just how my brain organises the problem before I write any code or design any screens.

I built elloAI in about three years from idea to first live version. The automation pipeline, the category architecture, the UI, the submission flow, and the first listings. Three years is not slow for a solo builder working without a team, a budget, or a deadline from anyone other than myself.

It is, however, fast enough to test whether an idea deserves more investment. elloAI passed that test. The pipeline worked. The first listings came in. The product did what it was supposed to do.

That is what I look for in the early stages of any build. Not perfection. Evidence.

Adeyemi Adetilewa is a product builder, content strategist, and digital marketer. He has built digital products across AI, fintech, SaaS, healthtech, and content publishing. You can find elloAI at elloai.com and read more about his work at adeyemiadetilewa.com.