CASE LOG

Case study · Bookkeeping & document tool

Quoted £10k. She built it herself in 6 weeks.

A UK tax lawyer turned her clients' worst bookkeeping pain into a working tool — uploads in, transactions auto-categorised, twelve hours of manual sorting down to two. She built every bit of it. I just made sure she never had to stop.

Overview

She didn't need a developer. She needed to stop spinning.

She's a lawyer who helps UK firms stay tax-compliant. Her clients were paying £3,000+ a year for bookkeeping and still drowning — endless back-and-forth between her, the client, and the accountants just to get transaction categories right, and months of spending data sorted by hand, line by line.

She knew there had to be a better way, and she'd been looking for one for a while. So she tried to build it herself. She started with the big vision — a full banking app, everything at once — got overwhelmed, and went to the AI tools. Gemini, ChatGPT, Claude. Conversation after conversation, and every time she came back the product had shifted again. New suggestions, new scope, new direction. She wasn't building anything. She was just spinning.

Our first call wasn't about building anything. It was about stopping the spinning — deciding the one thing her clients needed first, and what she wanted to build, not what the AI kept suggesting. From there she ran every prompt, made every decision, and built it herself in six weeks — for about a third of what she'd been quoted.

Engagement Built with you

Build time 6 weeks

What got built Shared document system · smart transaction categorisation

She built it with

AI / vibe coding On-call guidance Weekly sessions

THE DIAGNOSIS

The challenge

She could build it. She just hadn't decided what to build.

"They start too big, get quoted too much, and hand the one decision that matters — what to build first — to the AI."

This is what I see with almost every business owner trying to vibe code an app. They start with the full vision. They get quoted thousands. They open Claude, get three different suggestions in three conversations, and suddenly they don't know what they're building anymore. Meanwhile the actual business keeps running — clients calling, work to do — so the app idea gets pushed. They never abandon it. They just stop making real progress.

The problem was never that she couldn't build it. It was that she'd never stopped to decide what to build first — and she'd handed that decision to the AI. So the first move was to take it back.

What we scoped

Not a banking app. One thing her clients needed today.

Her clients didn't need a full banking app — not yet. They needed somewhere to put their financial data and have it sorted for them. So that's what got built first, and nothing else.

upload_file

Upload & auto-analyse

Clients upload their old financial data and it's automatically analysed and categorised — instead of months of spending sorted by hand, line by line.

rule

Smart categorisation

The AI sorts transactions into categories first, so bookkeepers verify what's already done rather than starting from a blank sheet every time.

groups

One shared document

The lawyer, the bookkeepers and the accountants all work inside the same document — no more emailing versions back and forth to agree on categories.

BUILT WITH YOU

How we worked

She built it. I made sure she never had to stop.

Prioritise

The first call cut the full vision down to the one problem costing her clients time and money today — and what she actually wanted to build.

Unstick

Wrong output from the AI, unsure which tool to use, wondering if it could even build what she pictured — every time she got stuck, we figured it out together on a call.

Ship

No more courses, no more rabbit holes. Clear answers from someone who knew her build and the next move — so she kept going until it was live.

Results

What changed.

6 wksIdea → live, built by her

~⅓Of the £10k+ quote spent

12 → 2 hrsBookkeeper sorting time

LiveReal clients logging in now

THE SHIFT

The shift

The bottleneck wasn't the build. It was the decision.

full banking-app vision→one thing clients needed today

AI deciding the scope→she decided, AI executed

spinning, no progress→live tool in six weeks

£10,000+ developer quote→a third of that, built by her

She didn't need a £10,000 developer. She needed to prioritise and build one solid thing first — and someone who'd done this before to give her timely, relevant answers to the questions that kept coming up, so she didn't get overwhelmed and stop.

Next case study

Shipped fast, validated faster

DuaComposer

A focused MVP with AI at the core — idea to revenue in three weeks. 1.6K+ downloads and 29 paying subscribers.

Read case study →

Still
spinning
on it?

Sitting on an app idea you know your business needs but can't get across the finish line? Take the 2-minute diagnostic, or tell me what you're trying to build and where you keep getting stuck — I'll tell you exactly what's in the way.

Start the diagnostic → Tell me where you're stuck