Should I Vibe It?
Do the math before you go down that rabbit hole, as luring as it might be
Heads up
this is vibe-coded slop and probably has bugs, verify the numbers, don't use this for actual business decisions
Grind it
Do it manually.
Yearly cost
- Run calculation to see itemized costs$0.00
Vibe it
Build and operate it yourself.
Yearly cost
- Run calculation to see itemized costs$0.00
Buy it
Use an off-the-shelf SaaS.
Yearly cost
- Run calculation to see itemized costs$0.00
Delegate it
Give it to a person.
Yearly cost
- Run calculation to see itemized costs$0.00
Should you automate it or just do it?
Building something with an AI coding assistant now takes an afternoon, which is exactly the problem. When the build is cheap enough to start on a whim, the decision stops being made on the numbers and starts being made on how interesting the task looks. A weekend script that saves ten minutes a week and then needs an hour of repair every month is a loss, and it is a loss you will not notice for a year.
The calculator above prices the same task four ways over twelve months and shows you which one is cheapest. It is the total cost of ownership, not the cost of getting to a first working version.
Four ways to get it done
1. Grind it
What does it cost to keep doing this by hand?
hours per month × your time value × 12 × team size
The baseline every other option has to beat. It is the only path with no build cost and no vendor, which is why it survives far longer than it should.
2. Vibe it
What does it cost to build and run it yourself?
build hours + maintenance × 12 + tokens + operating time + bug impact × 12
Five line items, and the one people forget is the last one. A build is a one-off number; maintenance, operating time, and the cost of it breaking are annual.
3. Buy it
What does an off-the-shelf tool actually cost once it is running?
(subscription × 12 × team size) + setup hours + operating hours × 12 × team size
The sticker price is per seat and the setup is a one-off, but the operating time is neither. Tools that need weekly babysitting rarely look cheap here.
4. Delegate it
What does it cost to hand it to a person?
onboarding + (supervision × your rate × 12) + (work hours × their rate × 12 × team size)
Supervision is priced at your rate, not theirs. That is usually what turns a cheap delegation into an expensive one.
Why the horizon is a year
Every recurring number is multiplied by twelve, and only the genuine one-offs (the build, the setup, and the onboarding) are counted once. Over a single month, building it yourself almost always wins, because the build has not started costing anything yet. Over a year, the ongoing items dominate, and a path that looked free at the start can end up as the most expensive of the four.
Team size matters for the same reason. Costs that repeat per person, such as a per-seat subscription and the time everyone spends operating the thing, scale with the team, while the build does not. A tool that is obviously too expensive for one person can be the cheapest option for twelve, and a script that pays for itself across a whole team is a waste of a weekend for one.
The inputs people get wrong
- Build time. Take your instinct and double it. The estimate you have in mind is for the version that works on your machine, not the one other people can use without you in the room.
- Maintenance. Anything that touches an external API, a login, or a file format will break on someone else's schedule. Two hours a month is a realistic floor, not a pessimistic guess.
- The cost of it being wrong. A script that silently produces a bad number is more expensive than one that fails loudly. Price the consequence, not the debugging.
- The manual hours. This is the input everyone guesses at, and the one the whole comparison rests on. Most people underestimate recurring admin badly, which is the thing DoneThat actually measures.
Common questions
When is it worth automating a task?
When the annual cost of doing it by hand is larger than the build, the maintenance, and the cost of it breaking, added together. The usual shortcut says to automate anything you do more than twice, but it ignores maintenance entirely. That is where most small automations quietly lose money.
Is vibe coding cheaper than buying software?
For something narrow, used by one person, and unlikely to change, usually yes. For anything that needs accounts, permissions, or reliability across a team, usually no. The crossover is not about the build being hard; it is about who fixes it at eight in the morning when it stops working.
What should I use as my time value?
Not your salary rate. Use what an hour is actually worth to you, which is generally higher. If you do not have a figure, the time value calculator works it out in five questions, and the team version does the same for a group.
What if two options come out close?
Then the money is not the deciding factor, and you should choose on something else: who owns it, how bad it is when it breaks, and whether you want to be responsible for it in six months. A margin under about twenty percent is well inside the error bars of your own estimates.