// decision · build log
Buy vs. Build: Should You Build Custom Software with AI in 2026?
AI made building cheap enough that 'just buy the SaaS' stopped being the automatic answer. Here's the framework I use with clients to decide, including the break-even math.
Not sure which side you're on?
Send your idea and I'll reply within 24 hours with what's buildable, how long it takes, and what it costs.
What changed
For twenty years the advice was simple: never build what you can buy. That held because custom software was slow and expensive. In 2026 a focused internal tool can be built in weeks for the price of one to two years of a mid-tier SaaS subscription — and it does exactly your process rather than the average of everyone's.
That doesn't flip the rule. It moves the line. Plenty of things are still obviously buy.
Six questions that decide it
If four or more point toward build, build:
01
Is this how you make money, or how you keep the lights on?
Payroll, email, accounting — buy. The workflow your customers actually pay you for, and that no competitor runs the same way — that's the one worth building.
02
How many tools are you stitching together?
If a job takes four subscriptions plus a spreadsheet plus manual copy-paste, the integration tax is already costing more than a build. Count the hours, not the licence fees.
03
Are you paying per seat for something ten people barely use?
Per-seat pricing punishes growth. At around $1,500/month in recurring tool spend for one workflow, a custom build usually pays back inside eighteen months.
04
Does the off-the-shelf tool need you to change how you work?
Bending your process to fit software is a hidden cost paid every day forever. Sometimes it's worth it. When the process is your advantage, it never is.
05
Who owns the data?
If the answer matters — regulated industry, sensitive client data, or a vendor who's also a competitor — build. Ownership is the one thing you can't buy back later.
06
Can you live with the roadmap you don't control?
Buying means waiting for features and absorbing price rises. That's fine for commodity software and painful for anything core.
Get a straight buy-or-build answer
Send your idea and I'll reply within 24 hours with what's buildable, how long it takes, and what it costs.
The break-even math
Take your annual spend on the tools covering the workflow, add the loaded cost of the manual hours spent working around them, and compare it to a build cost plus roughly 20% a year for maintenance.
A common shape: $1,800/month in subscriptions plus 15 hours a month of manual reconciliation is about $35K a year. A $20K custom build with $4K a year of upkeep breaks even before month nine — and from then on the process is yours, not rented.
If the numbers land close together, buy. Custom only wins clearly, and a marginal case usually means the workflow isn't core enough to justify owning.
The answer is often both
Buy the commodity layers — auth, payments, email, storage — and build only the thin layer that is genuinely yours. That's the cheapest version of building, and it's how most of the projects I take on are actually structured.
Get an honest buy-vs-build read
Tell me the workflow and what you're paying for it today. I'll reply within 24 hours with the numbers — including if the answer is 'keep the subscription'.
Get a build plan →