I Tried the Small Tools, the Big Brands, and My Excel. All Half-Right.

I'm a Hong Kong financial advisor. Over the years I've tried both ends of the market — small indie CRMs I found on the net, the big brands every advisor in my city ends up trying, even the ones with Instagram hype. They're well-designed. They're polished. They're trying to consolidate the mess of every insurance company's separate platform. I used them. I also kept an Excel sheet running on the side, because every time I opened one of those tools I noticed I'd already built a better version of it in Excel. Less pretty. More function. All of them half-right. Then Claude Code let me build the other half.

That's the post.

The Multi-CRM Mess

Here's what most people outside my industry don't know.

A Hong Kong financial advisor's client usually has insurance policies across multiple companies. Life cover with one insurer. Critical illness with another. Savings plan with a third. ILAS with a fourth. MPF with whoever the employer chose, which is a fifth. The combinations are infinite. Managing them has never been easy.

Every insurance company has its own CRM. Different login. Different layout. Different export format. To get one unified picture of one client's full coverage, I log into five separate platforms and assemble the picture by hand.

That's the daily pain. Not a bug, not a software failure — the structural shape of the industry.

What the Existing Tools Do (And What They Don't)

I'm not naming names — most of them are good companies trying to do good work. The small indie tools cover one slice well and skip the rest. The big-brand HK platforms try to be the unified layer that sits above all those separate insurer CRMs. The UI is clean. The AI assistants have nice touches. They're shipping.

I tried them. I still respect what they're building.

But every time I tried to use one of them for the work I actually do, I noticed the same thing: I already had a better version of it in Excel. Less polished, no question. More function, no question. Tracking the things the off-the-shelf products don't track because those things are too specific to my workflow, my client base, my team's process.

That's not a knock on any of them. The market is small enough that no commercial product can afford to be deeply opinionated about every workflow. They have to ship the average. My workflow isn't average. Yours probably isn't either.

The Option Problem

Here's the part the rest of the industry needs to hear.

The usual builder-vs-buyer question is "is there a tool to buy?" That's the wrong question for a niche this small. The real question is "is there even a tool for us to buy?"

Sometimes the answer is no. Not because the market hasn't gotten around to it. Because the market is too small and too specific to ever get around to it properly.

So the option set was:

The big-brand HK platforms (polished, ~70% of what I need)
The small indie tools (each one good at one slice, none cover the whole job)
Whatever the insurance companies hand me (each one different, each one only covers their own products)
Excel (95% of what I need, ugly, manual, mine)

Pick your compromise. Pretty without function. Function without pretty. Or the company tools, which are neither.

Pre-Claude Code, that was the whole option set. There was no fourth choice.

Post-Claude Code, the option set has one more entry: build it yourself.

What I Actually Built

The internal name is the financial-crm. The brand is Financial Planning Pro. It does the things my Excel did, with the polish my Excel didn't have:

A 10-step wizard that takes a new client from "what do you earn" to "here's your retirement gap" in about 30 minutes
Bilingual everything (EN/繁中) — chat, form, PDF, buttons
Branded reports per agency — every advisor on the platform gets their own logo, colours, footer
Stripe billing with grandfathered pricing for early agencies
Multi-tenant from day one — every agency's client data is walled off from every other agency's, including from me

Every bullet up there has an Excel ancestor. Five tabs of interactive formulas. A medical package calculator that pulled real plan prices and re-ran them against the client's age. The bilingual outputs were the newest piece I'd added — and the most fragile thing in the workbook. Claude Code didn't invent the spec. It built the app version of what was already in my .xlsx.

None of it is innovative software. All of it is software that wasn't being built for my exact workflow because the niche is too small to deserve it commercially. So I'm building it.

The Mindset Shift

Here's what changed in my head, specifically.

Before Claude Code: "There's no perfect tool. So I'll pick the closest one and Excel around the gaps." That sentence ran on loop for years. The real cost wasn't the subscription. It was the silent acceptance that nothing better was coming.

After Claude Code: "There's no perfect tool. So I'll build it." Same starting sentence. Different second half.

The Excel version was already evidence the build was possible. I'd been quietly maintaining the spec for years — every column, every formula, every edge case my industry has. Claude Code didn't teach me what to build. It removed the gap between knowing what to build and shipping it.

That's the actual rewiring. The Excel sheet is the product requirements doc. The AI is the dev team. Knowing what your industry needs is the part that took years to learn — and that part nobody else can copy.

The Test (Today)

Want to know how locked-in this brain change is?

Today I sat down to do nothing technical. I just wanted to write down what the financial-crm IS — the kind of doc you'd hand a new team member on day one. Old me would have written one paragraph and called it done.

Builder me wrote a 9-section product requirements doc, made it the workspace template so every future project starts the same way, and propagated the pattern to four other projects in the same afternoon. Then I noticed one of those projects was sitting in the wrong folder. So I wrote a separate plan to move it without breaking the GitHub link or the Vercel deploy, ran the plan, verified six different gates, and committed it.

None of that is coding. It's all builder thinking. Plan first. Write the doc the future-me will read. Ship in small pieces. Verify before committing. The mindset is the work. The code is the easy part.

I covered the team-side MPF tool in an earlier post — that was the first internal proof. This one is what came after, when the same brain change went from "tool for my team" to "the niche product nobody else was going to build."

Takeaways

The right question isn't "is there a tool to buy?" It's "is there even a tool FOR US to buy?" Niches too small for the average commercial product to deserve are exactly where builders win.
If you've been Excel-ing around the same gap for years, that's the spec. Your Excel sheet is the product requirements doc. You've been writing it the whole time without calling it that.
Pretty without function and function without pretty can both be replaced by one app you build. Pretty AND function isn't a luxury anymore. It's the new default once you can build.
You don't replace the dominant tool by competing on polish. You replace it by being more opinionated about a workflow it can't afford to be opinionated about. The narrower your niche, the bigger your edge.

Extra: How to start this week

Open the Excel sheet you've been quietly maintaining for years. The one you wrote because the official tool didn't have the column you needed.

That's the spec. Show it to Claude Code. Ask: "What would it take to build a small app that does just this Excel sheet, but in a browser?"

Read the answer. Sit with it for a day. If it sounds achievable, you've found your project. If it doesn't, ask the same question again in 6 weeks — the models keep getting better fast, the answer will be smaller, and your brain will have moved a little closer to "yes."

If you haven't gotten the tool working yet, start here.

I'll keep showing the financial-crm builds as they ship.

What's the Excel sheet you've been quietly maintaining for years?