I Don't Know HTML. I Build HTML Reports Anyway.

How AI helped me turn financial data into stories people can actually explore

Let me start with a confession: I don't know HTML. I learned a little about it in school, and I know websites are built with it, but that's about it. And yet, some of the reports I'm proudest of today are interactive HTML pages. My job isn't to write the code. My job is to know what the numbers need to say, and to guide AI until it says it clearly.

Where it started

I'd been hearing stories about people building HTML games as part of an internal innovation initiative, but I never really explored it. Then I got on a call with Shanbor, who wanted to visualise some data for a dashboard. Neither of us were designers, so we said, let's see what AI can do. The layout it gave us was really good. I remember thinking, wow, I didn't know we could do that. And then, honestly, it just stopped there. We didn't share it with anyone.

The real moment came when Carlos, our external consultant, wanted to see our financial data broken down, customer by customer, going deeper and deeper. I built an Excel file where clicking on revenue took you to another page showing where the money came from. When I looked at it, I thought: this is confusing, even for me. So I took the same data and tried it with AI. The first result stunned me. No Chillibreeze branding, default colours, default fonts, but in a matter of minutes it had understood the data, segregated it, and made it so much simpler. I got excited and shared it with everyone.

Why HTML works for finance

For years, the standard approach has been simple: compare budget versus actuals with two bars and use a small note to explain why they don't match. We kept following that format because that's how it was always done. I come from a finance background. I might have some creative visualisation in my head, but turning that into a concrete output has never come naturally to me. Even as a child, I'd ask my dad to do my drawings. It wasn't a lack of ideas. I just struggled to bring what I saw in my mind onto the page.

HTML changed that. In Excel or PowerPoint, you click a chart and maybe you see a data label. In HTML, you click and you go inside the detail. Go back, and you see the whole picture again. Hover over a name and a comment appears; move away and it's gone. It's basically a website. Sometimes I sketch an idea on my iPad or find a reference on Pinterest, send it over, and the AI tries to replicate it.

The taxi report

One report I worked on analyzed how our taxis were being used. The data showed patterns we hadn't noticed before, including time slots with consistently low occupancy and opportunities to make better use of each vehicle.

The insight wasn't really about taxis. It was a reminder that when you look beyond the numbers, you often find opportunities to improve how things work. In this case, the data gave us evidence to rethink existing practices and make better decisions.

AI builds it. I still have to think.

A good HTML report starts with human analysis, before I even write a prompt. I need to know what I'm trying to find. And AI can get things wrong, so we check the numbers, catch the errors, and prompt it to correct them before anyone relies on the report.

HTML gets you from point A to point B. What you do once you're at point B is still up to you.

This is what lets finance be a thinking partner. We have a lot of data, but before we couldn't really show it. Now I can show something like a purchase order raised for $100,000 where only $50,000 was spent, and that opens a conversation. Teams can go back and look for the root cause. What-if scenarios that used to mean changing numbers in an Excel file can now be edited right inside the report. In finance, we're trying to find constraints in our systems, and I don't think we could have done it without HTML and AI.

Getting faster

I've become so much faster. A recent report that already had a structure took me about an hour and a half with two or three iterations. A straightforward report takes roughly one to two hours; a deeper analysis for a presentation can take two or three days. We used Cowork at first, and now we use Copilot. I pin reports that still need work and keep refining them until they're ready.

My advice if you want to try this

Be patient. The first result is rarely what you imagined, so keep iterating.
Teach the AI your world. Once your logos, brand templates and regular requirements are embedded, it gets much faster.

Stay curious. It can do more than you expect.