AI has made it ridiculously cheap to build software. It hasn't made it cheap to own. The moment 30 people depend on the app you vibe coded, you're running a software company. Here are ten reasons to build, ten reasons to buy, and not one of them is "it's cheaper." Then 25 checks to run before you build anything your company can't live without.
Deciding on one tool right now? Find its kind in Three Kinds, then paste The Prompt into Claude. Already built it? Go straight to The Checklist and see how you score.
Building internal tools with AI is fun, and it gets addictive. But a founder rebuilding their CRM instead of calling the people in it is doing Building Theater. It feels productive and it makes $0.
I'm building a place where you learn to build products that work like a real company's on your own, and get them in front of people who pay, with a CTO who's built two seven-figure software companies in your corner.
Build What Sells.
Free to join the waitlist. Opens at $49/mo, and the price goes up as seats fill, so first in pays the least.
Five years ago this was easy: just buy it. Now I can build a useful internal tool in a weekend, and I think companies should build way more of them. But the weekend is the cheap part. Before you build, ask who's going to own it for the next five years.
Every company has that one critical spreadsheet only Janet understands, and God forbid Janet ever quits. AI just gave Janet's spreadsheet a React front end, a database and a login screen. Same problem. Now it has a dashboard.
AI has made building software almost free. It hasn't made running software free.
Build or buy is the wrong first question. First work out which kind of software it is. That decides most of the answer.
AI changed the math here completely. If it breaks, someone's annoyed for an afternoon and does it by hand. Build as many of these as you can.
A quote calculator · a dashboard over your Stripe data · a script that cleans the weekly export
Off-the-shelf software is built for the average customer. If this is part of why customers pick you, the average won't fit, and you'll fight the tool forever.
Your pricing engine · your onboarding flow · how you schedule the work
Buy it, unless you'll commit to owning it for years: someone on call, tests, backups you've restored, audits. That's a team, not a weekend.
Accounting · payroll · HR · your main CRM · anything that moves money
The kind isn't about how hard it is to build. It's about what happens when it's down. Tools also move between kinds without anyone deciding: the dashboard becomes the place orders get approved, and now it's a system of record.
Skip "it saves on licenses." Everybody knows that one, and it's usually wrong once you count the time. These are the reasons that actually matter.
Per-seat pricing quietly decides who gets access. The warehouse lead who needs one report a month doesn't get a login, so they ask someone, or a login gets shared. Build it, and everyone who should see the numbers can.
Ask: Who goes without access today because a seat costs too much?
Every tool ships with opinions: pipeline stages, field names, the approval flow. Six months in, those defaults are your process. If how you work is part of why customers pick you, don't let a vendor flatten it into the industry average.
Ask: Is this workflow part of why customers choose you?
An agent is only as useful as the data it can reach. In your own database, it can query anything. In a bought tool, it goes through a rate-limited API, a higher plan or the vendor's own AI add-on, if the vendor lets it in at all.
Ask: Will you want an agent working on this data next year?
Vendors get acquired, reprice at renewal, retire the plan you're on or shut down. Your own tool can't triple its price or sunset the one feature your team depends on.
Ask: What happens if this vendor doubles its price next year?
Bought software changes when the vendor decides. A redesign lands on a Tuesday and twenty people can't find the export button. Yours only changes when you change it.
Ask: How much did the last vendor redesign cost your team?
Bought tools do 90% of the job. The other 10% lives in a spreadsheet next to it, and that 10% is usually the part that's specific to you. Build it, and the tool does the weird thing too.
Ask: Will there still be a spreadsheet next to the tool you'd buy?
Need a field, a report or a button? On a vendor's roadmap, that's a forum thread with 400 upvotes and no reply. On yours, it's an afternoon with Claude Code.
Ask: How often will this need to change?
A tool for one launch, one audit or one migration can be built in a weekend and deleted in a month. No annual contract, no auto-renew, no 60-day cancellation window. Disposable software didn't make sense before AI. Now it does.
Ask: Do you need this for months, or for years?
Your first internal tool means setting up hosting, login, a database and deploys. The second one reuses all of it. Building is a skill your team gets better at. Buying never teaches it.
Ask: Is this your first internal tool, or your fifth?
Every tool you buy is another company storing your customer list, another security review to fill in, and another breach email you might get one day. Build it, and the data stays where the rest of your data already lives.
Ask: Would you rather trust this vendor's security, or your own?
Not maintenance, scaling or "it might break." These are the costs that don't show up until year two, and they're the ones that hurt.
Your ops manager vibe codes a great tool, then takes a new job. Nobody knows how it works or how to change it, and the AI that wrote it doesn't remember either. A vendor's team doesn't hand in its notice.
Ask: If the builder left tomorrow, who could change it?
The flip side: they stay, and they're stuck. They can't take the promotion, move teams or go off-grid for two weeks, because they're the only one who can fix it. You've handcuffed your best operator to a side project.
Ask: Could you promote the builder away from this tool?
Someone can't log in. Someone deleted a row. Someone can't find last month's report. None of those are bugs, and all of them are yours. A vendor has a support team, help docs and a status page.
Ask: Who answers "it's not working" at 9am on a Monday?
Salesforce retires an API version. Google changes a permission. Stripe changes a webhook. Your app breaks on a Tuesday because someone else shipped. Every integration is a relationship you now have to keep up.
Ask: How many outside systems does it connect to?
AI makes changes cheap, so you make lots of them. Who checks that this week's feature didn't quietly break last month's invoice export? It's worst anywhere money moves.
Ask: Does it touch money, payroll or anything a customer sees?
Time zones, partial refunds, a customer in two countries, a tax rule that changed in January. A mature vendor has been bitten by all of them, with thousands of customers paying for the fixes. Your tool meets each one for the first time, live.
Ask: Have a thousand other companies already solved this exact problem?
When someone leaves, a bought tool tied to your Google or Microsoft login shuts them out automatically. Yours has a users table someone has to remember. And the integrations often run on the builder's own account, so they die the day it's switched off.
Ask: When someone leaves, does their access leave with them?
Big customers and auditors ask about every system that touches their data: who has access, where the audit log is, how long you keep it, how you delete it. Your weekend project now needs answers to all of that.
Ask: Will customer data ever go into it?
Once building is easy, sales builds a CRM, ops builds a CRM and customer success builds a CRM. Now there are three, none of them agree, and nobody knows which one is right.
Ask: Has someone in another team already built a version of this?
When you raise money or sell the company, buyers ask what you run on, and a homegrown system of record goes on their risk list. And when you outgrow it, getting the data out in a shape another tool understands is all on you.
Ask: Could you move everything to another tool in a day?
Most of the time the answer is both. Buy the system of record and keep the data in it. Then build the small tools around it with AI. When something you built breaks, nothing important is lost, because the data never lived there.
The accountant gets the system they know. You get the view you want.
Your edge lives in the scoring, not in storing contacts.
Payroll law is their problem. Your first-day checklist is yours.
Run this before a small tool turns into something the company can't run without. Tick what's true today, not what you plan to do.
Not Ready To Own ItKeep it a small tool that nobody's job depends on, or buy something.
Paste this into Claude, ChatGPT or whatever you use. It interviews you first, then gives you a verdict and, if you're building, an ownership plan.
I'm deciding whether to build or buy [THE TOOL, e.g. a CRM for our sales team]. Interview me before you recommend anything.
Ask one question at a time and wait for my answer. Cover:
- What it does, who uses it, and how many people will depend on it
- What happens if it's down for an hour, a day, a week
- Whether it touches money, payroll, customer data or anything regulated
- Which outside systems it connects to
- Who would build it, who would own it, and what happens if they leave
- What we'd buy instead, what it costs, and what it can't do for us
Then give me:
1. The kind of software it is (a small tool, a workflow that's our edge, or a system of record) and why.
2. For each of these 20 points, one line on whether it pushes us toward build or buy:
Build: seat pricing limits who sees the data; we'd run on a vendor's defaults; our AI agents need the data; the vendor could reprice, be sold or shut down; forced redesigns; the workaround spreadsheet; waiting on a vendor's roadmap; we only need it for a while; we've built tools before; one less company holding our data.
Buy: the builder leaves; the builder can never move on; we become the help desk; outside APIs change; nobody re-tests old features; vendors have solved the edge cases; offboarding; customer security reviews and audits; other teams building their own; due diligence and getting the data out.
3. A recommendation: build, buy, or buy the core and build around it.
4. If it's build: a one-page ownership plan with two named owners, company accounts, backups we test, tests on anything touching money, a tripwire for when we'd switch to buying, and a review date.
Be blunt. If we should buy it, say so.
Swap in the tool you're deciding on. Leave the rest alone.
Answer the questions with whoever would build it and whoever would use it most. They'll disagree, and that's the useful part.
If it says build, work through The Checklist before more than a handful of people depend on it.
Anybody can build software now. Not everybody can build a business.
Reading about tools is a great way to feel busy. The Lab is for Founders who'd rather get customers: weekly office hours with a CTO who's built two seven-figure software companies, and a room of other Founders who care whether anybody buys what you're making.
Less Theater. More Customers.
Want my 30-Day Launch Plan? It's for Lab members only: validate, plan, build and launch, one week at a time.
See The Solo Founder Lab ▶