By Raphael Schwirtlich · Published Tue Nov 17 2026 · Last updated Sat Oct 20 2018
Risky Vibes pt. 01
Build VS. Buy
Disclaimer
This article is general information and my personal opinion, not legal, tax or insurance advice. Laws differ by country and state/canton, change over time, and depend on the facts of each case. Before you replace a business system, ask a qualified lawyer, tax adviser or data protection specialist about your own situation.
The Subscription Math
Every week another post tells me that SaaS is finally dead, that developers and software companies are effectively out of business, because they are being replaced by AI.
A CEO cancels their CRM and builds a new one over a weekend. A company announces that it saved six figures by replacing subscriptions with vibe coded in-house tools.
I build business automation systems for a living, so I'm not allergic to the idea of creating something that meets your individual requirements. But after I recently reviewed a vibe coded business app for a client, I realized the entire magnitude of exposure vibe coded apps create for a business. And I couldn't stop wondering: What could possibly go wrong with this approach?
This 10 piece series is my attempt at answering this question. It is not a verdict or a call to stop vibing. It is a list of considerations worth making before you cancel your business subscriptions and replace them with home made tools.
The pull is real
In a 2026 survey by Retool (about 800 respondents), 35% of teams said they had already replaced at least one SaaS tool with custom software, and 78% expected to build more internal tools this year. Retool sells tools for building internal software, so read this as a sign of the trend, not as a neutral measurement. The most replaced categories were workflow automation, internal admin tools and dashboards. Accounting and payroll were not at the top of that list, which is worth remembering.
The appeal is obviously very easy to understand. Per-seat pricing quickly adds up to a huge monthly bill, and an AI coding tool can produce something that looks like your CRM in an afternoon.
What the subscription fee actually pays for
A subscription is not just access to software.
It usually also covers:
- hosting, uptime and backups that someone else monitors
- security patches and monitoring
- updates when rules change, such as tax formats, payroll tables or invoice standards
- support when something breaks
- legal paperwork: data-processing agreements, lists of sub-processors, defined data locations
- evidence you can show customers, insurers or auditors, such as certifications and audit reports
- a counterparty with contractual obligations (often limited, but still someone)
A commentator in my feed put it plainly:
SaaS fees cover ongoing maintenance, while a custom build shifts that burden to you.
What the homemade version costs
Build time is only the visible part. The rest of the effort is easy to overlook:
- upkeep,
- security review,
- testing that backups really restore,
- keeping up with rule changes,
- and the time of whoever understands how it works.
A company that publicly runs production vibe-coded apps reports spending roughly 30 to 60 minutes a day maintaining each one. This is an anecdote, not a benchmark, but multiply it by five replaced tools and the savings look different.
The cost of being wrong
An investor commentary on this trend made a point I found useful:
AI lowers the cost of creating software but "does not change the cost of being wrong".
So think through the possible failure modes. Invoices with wrong VAT. A customer list that leaks. A payroll mistake. With a subscription, the vendor carries part of that risk. With a homemade tool, you carry the burden by yourself.
Where it can make sense
Low-stakes tools used by one team, where you control the data and the worst case is some inefficiency. Internal trackers, forms and small automations are examples. Some observers point out that building your own also brings real ownership. Article 9 will come back to this.
The effort at a glance
| Line | Subscription | Homemade |
|---|---|---|
| Yearly fee | The price you pay today | None |
| Build time | None | Hours times your hourly cost |
| Upkeep | Included | Hours per month times 12 |
| Security review and testing | Included | Yours |
| Hosting, backups, restore tests | Included | Yours |
| Rule changes (tax, payroll, invoice formats) | Included | Yours |
| Cost of an incident | Shared with the vendor | All yours |
| Cost of going back | Export and cancel | Rebuild or migrate |
What the rest of this series covers
The kinds of data these tools hold (article 2), then customer data (3), financial records (4), HR and pension data (5), what security reviews tend to find (6), what you lose when the vendor goes (7), insurance and responsibility (8), a hybrid approach (9), and a plan for the first 90 days (10).
Questions to bring to your adviser
- Which of my tools hold personal data, and what does that mean for me if I run them myself?
- If this tool failed or leaked, what would my exposure be, and what does my insurance say?
- Do any of my customers or contracts require particular systems or security evidence?