Your Prompt
I'm considering building a custom solution for my business.
Before recommending whether I should build, buy, or customize, help me figure out:
What outcome am I actually trying to achieve?
Do any existing software already solve 80-90% of this problem? Are they close to closing the last 10%?
What responsibilities would I be taking on if I owned this solution?
Ask follow-up questions until you understand both the business problem and the economics of solving it.
Only after you've gathered enough information should you give your assessment of whether I should buy or build.
I get excited and easily distracted by new projects, so keep me focused on the outcome.
Against The Grain
A few years ago, I bought a 3D printer.
Like a lot of people, I was fascinated by the idea.
Need some pantry organizers? Print it.
Need a replacement part for something around the house? Print it.
Need a tiny plastic Baby Yoda?
Who doesn't? Print it.
For a while, it was a fun hobby.
Searching for fun STL files. Experimenting with different tensile strengths. Learning enough to be dangerous.
And every successful print felt like magic.
Then reality showed up.
What was supposed to be a fun hobby with my son soon turned into dad fixing things by himself, and him returning to his Minecraft world where everything snaps together reliably.
Prints failed halfway through. The bed needed recalibration. The nozzle clogged.
The settings that worked perfectly last weekend suddenly didn't work this time.
An eleven-hour print would fail during hour ten, leaving me with a chunk of stringy plastic.
Eventually, I sold the printer once I realized the lesson.
I didn't actually want a 3D printing hobby.
I wanted the thing.
And if I wanted the thing, Amazon could get it to my house in two days.
If I needed something unique, Etsy was full of people who had already solved the problem. If I could find an STL file for the thing, there was someone else already printing and selling it.
The economics worked for them.
Not for me.
I'm starting to see the same pattern with the current wave of vibe coding.
Claude Code is convincing accountants they should build their own accounting systems.
Their own workpaper platforms.
Their own invoicing software.
Their own CRMs.
Their own client portals.
Can AI help you build those things?
Absolutely.
And I'll admit... it's fun.
There's something deeply satisfying about describing a problem in plain English and watching software appear.
Not that different from watching that first successful 3D print slide off the build plate.
But a weekend hobby is not the same as a business model.
Building software is the easy part. Owning software is the hard part.
The first version works. You're so stoked.
Then comes version two. Bug fixes. Security holes. Permission settings. Documentation. Training your team.
Then the employee who built most of it just accepted a job somewhere else. Or the AI platform you rely on jacked up its usage rates.
Now you're not an accounting firm with custom software.
You're an accounting firm that tried to become a software company. And software companies tend to spend most of their time maintaining software.
The irony is that the AI tools making software easier to create are also making software companies more capable of customizing their products.
The last ten percent that used to require an internal development team is increasingly becoming part of the product itself.
Which means many firms are racing to build bespoke tools from scratch while software vendors race towards configuration from a nearly-there baseline.
This is not an argument against experimentation.
We should all spend time pushing on these tools.
You learn by building.
You discover opportunities by playing.
You become a better buyer when you see what's possible.
And I've certainly known plenty of unhappy accountants who finally found their exit by becoming software companies.
But I love accounting. I'm where I want to be.
And for most of us who love our professions, learning how to build something and deciding to own it forever are two very different decisions.
The question isn't whether AI can piece together an accounting system.
The question is whether maintaining an accounting system is the highest and best use of an accountant.
Most people don't buy a 3D printer because we wanted to become experts in nozzle maintenance.
We just wanted the thing.
