Skip to content
The Lean CPTO
Menu
Take the free check

Guides

Should my business build custom software, buy it, wait or stop?

Short answer: most of the time, don’t build first. Check what you already pay for, then look for a tool you can buy. Build only when nothing does it your way, it’s worth real money, and you’ve proved people will use it or pay for it. If it’s not worth it yet, wait. And if it’s not worth it at all, stop. Stop is a real answer. It’s the cheapest one you’ll ever get.

The expensive mistake isn’t the code

When owners ask me whether they should build something, they usually expect a conversation about code, developers and price. The code is rarely what goes wrong.

The expensive mistake isn’t using AI wrong. It’s building the wrong thing. Something your team won’t use, or your clients won’t pay for.

I spent two years selling offshore development to businesses in Australia and Japan. The problems I saw mostly came from decisions made before anyone wrote a line of code: what to build, how much of it, and whether anyone had checked it was needed.

So before you ask “how much will it cost?”, ask a different question. Which of the four answers is this?

Four answers: build, buy, wait or stop

Every software or AI idea for your business ends in one of these:

  • Buy it if something already does the job. That includes tools you already pay for.
  • Build it only if nothing does it the way you need, and it’s worth real money.
  • Wait if you need more proof first, or the timing is wrong.
  • Stop if it’s not worth it. That’s a good answer. It saves you a developer’s invoice.

Work down the questions below in order. Stop at the first one that fits.

Question 1: does something you already pay for do it?

Start here, every time. Microsoft 365, Google Workspace, your accounting software and your job or practice software have all been adding AI and automation. A lot of it sits there switched off.

Before you look at anything new, check the AI you’re already paying for. If it does 80% of what you need, turn it on and use it for a month. You’ll learn more about what you actually want from that month than from any planning meeting.

Question 2: is there a tool you could buy that does it?

If what you have doesn’t do it, look for a tool that does. Most business problems aren’t unique. Booking, quoting, reporting, chasing invoices and answering the same questions are solved problems with plenty of tools on the market.

Try it for a month before you commit. Watch whether your team keeps using it after the first week. A tool nobody uses costs more than the subscription.

There’s also a bigger reason to buy first. In MIT’s 2025 research on AI projects in large companies, bought tools and partnerships succeeded about 67% of the time, and internal builds only about a third as often (Fortune). That’s big companies, not Australian small business, and the research was early. But if building is hard for them, with their own tech teams, it’s worth taking seriously.

Question 3: nothing does it your way, and it’s worth real money?

This is where building starts to make sense. It has to pass three tests:

  1. Nothing does it the way you need. You’ve checked what you pay for and what you could buy, and the gap is real, not a nice-to-have.
  2. It’s worth real money. Put a number on it: hours saved, mistakes avoided, or what clients would pay.
  3. You’ve proved people will use it or pay for it. Your team uses a manual version every week, or clients have paid for the result done by hand.

Here’s a made-up example of test 2. Three people in an accounting practice each spend four hours a week on client reports. That’s about 600 hours a year. If a tool saved half of it, at a cost of A$60 an hour, that’s about A$18,000 a year. A A$40,000 build would take over two years to pay back, before running costs. Maybe that’s still worth it. Maybe a A$100 a month reporting tool gets you most of the way. The point is you can see it before you spend.

If it passes all three, build the smallest version first. The part people already use or pay for, nothing else. When paying customers need the same manual step every week, build that step, nothing else.

Question 4: not worth it yet, or you’re not sure?

Then wait, or stop.

Wait when the idea might be good but you don’t have the proof. Run it by hand, or on the tools you already have, until the same manual step keeps repeating. That repeat is your signal. Every manual step you repeat is a feature someone has already paid for.

Also wait, or buy, when the idea touches something regulated. If it gives financial, legal or health advice, holds client money, needs a licence, or stores sensitive personal information, an existing tool built for that may be safer than a custom build. Get advice for your industry.

Stop when the numbers don’t work, nobody will use it, or nobody will pay for it. Finding that out now, before a developer’s invoice, is a win. Plenty of ideas should stop. That’s not failure. It’s money kept in the business.

For tradies with a team

The same questions work for a trade business. The decisions just look different:

  • The cold-called AI receptionist. Before you sign up, check what your job software and phone plan already do for missed calls. Then try one for a month before you lock in.
  • Switching job software. Often the better answer is switching on the features you already pay for in the one you have.
  • A custom quoting app. Most of the time a good template and AI do the job. Start with AI for tradies: the first job to hand over and see how far that gets you.

If the answer is build

Then do it carefully:

  • Smallest version first. One page that says what it does, what’s left out, and what done looks like.
  • Own it. The code, the hosting and the accounts in your business’s name, from day one.
  • Check the quote before you sign. These 12 checks need no tech knowledge, and these questions to ask a developer help before the quote arrives.

I can introduce Sazae, a development partner. I’m paid a commission if you go with them, and you’re free to use anyone you like.

Where to go next

If you’d like a straight answer on your own idea, in writing, that’s what the Build Case is for. We talk it through on a 90-minute call, and you get a written answer: build, buy, wait or stop, what it costs, what it could return, the smallest version worth building, and a check of one developer quote if you have one.

If you’re earlier than that and want to get AI working on everyday jobs first, start with WTF is AI?. It’s six short videos and six cheat sheets, and video 6 is this same build, buy, wait or stop question.

FAQ

Should a small business build custom software or buy off the shelf? Buy first in most cases. Check what you already pay for, then look for a tool that does the job. Build only when nothing does it your way, it’s worth real money, and you’ve proved people will use it or pay for it.

How do I know if custom software is worth it? Put a number on it. Hours saved, mistakes avoided, or what clients would pay, against what it costs to build and run. If it takes years to pay back and a cheaper tool gets you most of the way, it probably isn’t worth building yet.

When should a business not build software? When something you already pay for does it, when you can buy a tool that does it, when nobody has used a manual version yet, or when the numbers don’t work. Stop is a good answer. It saves you a developer’s invoice.

Is an AI tool different from custom software? Not for this decision. The same four answers apply. Most AI tools for small business are things you buy or switch on, not things you build.

What should I do before I get a developer quote? Write one page: what it does, what’s left out, and what done looks like. Then check the quote for ownership, scope, payments and handover before you sign.

General guidance, not legal, financial or IT security advice. Get advice for your industry if your idea involves regulated advice, client money or sensitive personal information.

Free check · about 2 minutes

Where are you at with AI and software?

Tap a few answers and get your score, the first three things to do, and the one next step that fits. New to AI, using it a bit, holding a developer's quote, or thinking of building something. Free, and no sign-up to see your result.

If you need something built, I can introduce Sazae, a development partner. I'm paid a commission if you go with them, and you're free to use anyone you like.