Buy when every business like yours does the job the same way, or when time is short. Build when no tool fits how your team works. Do it yourself with AI when you have someone in house who will own the result. AI can build real tools now. Whichever you pick, decide who keeps it running.
We build custom software for a living. Sometimes we tell a client not to build it. Lately some owners tell us they can build it themselves with an AI tool, and they are right that it works. All three are fair options. Here is when each one fits, what it costs you, and who is on the hook when it breaks.
Option 1: Buy
When it fits. The job is one every business like yours does. Booking, payroll, invoicing and email are examples. Someone else has built it, tested it and keeps it updated. Buying also fits when time is short. A ready-made tool works this week.
What it costs you. A subscription, and your team learns the tool's way of working. You rent it, so check you can take your data with you.
Who is accountable when it breaks. The vendor keeps the tool running. You still answer for whether it fits your process.
Example. A booking-based service business asked us whether to build its own booking system. We said no. Booking is a standard job. They needed to show the schedule, take the booking and take the payment, and ready-made platforms do that. They also had a deadline, and a custom build would have cost money and weeks they needed elsewhere. Our part is to connect the platform to their website and set it up.
We wrote before about why owning your system matters. This post covers the other half: when renting is the right call. The rule we use in both is the same. Own the parts that make your business different, and rent the parts every business has.
Option 2: Build
When it fits. No tool fits the way your team works, and every tool asks you to change your process first. The process is part of what makes you different.
What it costs you. Money up front and weeks before it works. Afterwards someone has to keep it documented and updated.
Who is accountable when it breaks. The team that built it, if the contract says so. Ask who holds the code and the servers, and whether a different team could continue the work.
Example. A Singapore civil engineering firm raised permits to work on paper. One project raised 10 to 15 a day, and the firm runs 5 to 8 projects. Each permit took about 30 minutes. A permit process follows how a firm works, so nothing off the shelf fit. We built a system around their existing sign-off chain. Each permit now takes 3 minutes. Read the case study →
A second example, in progress. A Singapore service business asked us to quote on a system that drafts its customer quotations with AI. Scoping showed us where the work sits. Writing the draft is the easy part. The hard part is turning what the crew sees on site into clean information, and writing down the judgment their most experienced coordinator applies. In the design, a crew member fills in a short form on site and adds a photo. The system finds the closest past job and drafts a quote from it, and shows which job it used. A coordinator reviews every draft before it goes out. Later it could connect to their field service system and send customers their updates.
Option 3: Do it yourself with AI
When it fits. You have someone in house who can build with AI and who will own the result. Small tools and one-off tasks fit easily. Owners now tell us they have built their own ERP for a team of 100 or more this way. That is real, and it deserves to be taken seriously.
What it costs you. More than the build. The AI usage alone can run to several thousand dollars or more, before you count the builder's weeks. The AI is the smaller part of the work. The larger part is the information it works from: your price list, your past jobs, and the judgment of your most experienced person. Someone has to collect that and keep it current. Then come testing, security checks, fixes, and cover for the days that person is away. If you run more than one system, someone also has to connect them, host them, keep them secure and plan for more users. AI-built systems are new, so there is little track record yet.
Who is accountable when it breaks. You, and the one person who built it. Ask three questions before your staff depend on it:
- Who keeps it running when it breaks, when the tool it relies on changes, or when it has to work with your other systems?
- Who answers when it fails on site or in front of a customer?
- Who controls access and records? Who can see the data, change it, and find it a year later?
If the honest answer is "whoever made it, if they remember", treat it as a prototype.
Example. Say a business with over 100 staff builds its own ERP with AI tools. The bill you see is the AI usage: several thousand dollars or more. The bills you do not see are the builder's weeks, the testing and security checks, the fixes after launch, and the cover when that person is away. None of this means the idea is wrong. It means the price is bigger than the token invoice.
A plain rule for deciding
Own what makes you different. Rent what every business has. Then check three things:
- Does every business like yours do this job the same way? If yes, buy.
- Would your team have to change how it works to fit the tool? If yes, look at building.
- Will people depend on it every day, and does it hold records or access? If yes, it needs a named owner, whether you buy it, build it with us or build it yourself with AI.
Questions owners ask
Can I just use an AI tool like ChatGPT to build our software?
Yes, far more often than a year ago. People now build real internal tools this way, even for large teams. The AI usage is a visible cost and often a large one. Before your staff depend on it, settle three things: who keeps it running, who answers when it fails, and who controls access and records.
When should a small business buy software instead of building it?
Buy when every business like yours does the job the same way, or when you need it working soon. Booking, payroll, invoicing and email are common examples.
Does buying ready-made software mean giving up control?
Not by default. Before you sign, check that you can export your data, what happens if the vendor changes its terms, and who you call when something breaks.
Can you connect a ready-made tool to my website?
Yes. If a ready-made tool fits, we say so, then connect it to your website and the rest of your business.
Not sure which one you need?
Tell us the process that slows your team down. If a ready-made tool fits, we will say so. If you can build it yourself with AI and have someone to own it, we will say that. If you have already started building with AI, tell us what it does and we will tell you what to check before more people rely on it. If nothing fits, we will say that too. Whichever way you go, ask the ownership questions first. We list them in You Don't Own Your System, You're Renting It.
