Most talk about AI in business is about chatbots: drafting an email, summarising a PDF, writing a social post. Those help, but they’re the small end. The bigger change is that AI has made custom software much cheaper and faster to build. An established business can now have its systems built around how it actually works, with every app it already uses connected to one central hub, and AI agents doing the reading, sorting and drafting inside it.
That matters because of how most businesses end up with their software. This guide covers why off-the-shelf apps never quite fit, how the pile of disconnected tools builds up, what one hub looks like in practice, and where AI does and doesn’t belong in it.
No two businesses run the same way
Every business has its own fingerprint. Two plumbing companies in the same suburb will quote differently, schedule differently, chase payments differently and report to the owner differently. A blinds factory has its own rules for turning a window measurement into a cut list. A clinic runs its own programs and payment plans. Those differences are usually why customers pick one business over another.
Off-the-shelf software is built for the average business in a category. It covers what most businesses in that category have in common and leaves you to work around the rest. So it’s always missing something: a field you need, a step your team does that the app doesn’t know about, a report it can’t produce. Staff fill the gap with a spreadsheet, a sticky note or a memory.
That’s fine for a while. It stops being fine as the business grows.
How the software pile builds up
Most businesses start with one, two or three people. Everyone knows every job, and the system is a spreadsheet and a shared inbox.
As you grow, you add people, and you add software one problem at a time:
- an accounting package, such as Xero, MYOB or QuickBooks Online
- a quoting and invoicing tool
- a CRM for leads and customers
- a booking or scheduling app
- an online store, a payments provider, a shipping tool
- and spreadsheets to cover whatever none of them do
Each one made sense when you bought it. But by the time you have four, five or six of them, none of them talk to each other. A customer’s details are typed into the CRM, then again into the quoting tool, then again into the accounts. Someone exports a report from one app and pastes it into another every Monday. When the owner wants to know how the business is going, someone rebuilds the numbers by hand.
Your staff become the connection between the systems. That’s where the hours go, and it’s where the mistakes come from: the changed colour that never reached the factory, the deposit that was paid and never matched, the order that shipped without an invoice.
One hub that knows what every department is doing
The fix is to take the data from each of those systems and bring it into one central place. One hub, where each record is entered once and every department works from the same version of it. Sales, the office, the floor and the accounts all see the same job.
The hub doesn’t have to replace everything. Where an app does its job well, like Xero for the books or Shopify for the store, the hub connects to it and keeps it in sync. Where an app doesn’t fit, or where staff are running a spreadsheet beside it, the hub does that work instead, built to your rules.
Three examples from our own work:
- Raw Blend runs six Shopify stores, 26 suppliers and one QuickBooks Online company from one hub with 52 automations. Orders become invoices, tracking numbers reach the right invoice, and supplier price lists update every store.
- A Brisbane blinds and shutters manufacturer was typing every job into an old quoting program, MYOB and spreadsheets. Their system carries one record from the rep’s home visit through the factory floor to the installer’s sign-off. It does the work of a CRM, a quoting tool, a production system, an installer app and a wholesale portal. It’s about 62,000 lines of code, written for that one factory, and it’s in acceptance testing with their team.
- A Brisbane clinical nutrition practice had client records spread across old practice software, Notion boards and email. One hub now runs bookings, Stripe payment plans, Xero, forms and session notes. It was live with real clients after about four weeks.
None of these is a small fix. Each one changes how the whole business runs day to day.
What AI changes about building software
Until recently, a system like the blinds manufacturer’s meant a development team and a budget most $1–5M businesses would never approve. That’s why so many of them stay with the pile of apps.
AI coding tools have changed the cost. An experienced engineer working with AI can write, test and change code far faster than before. The AI drafts code, writes tests, reads old data exports and works through the boring parts. The engineer decides what gets built, checks what the AI writes and is responsible for whether it works. The clinic’s hub, for example, has 905 automated tests that run on every change.
The engineer still has to understand your business. The hard part of the blinds system was never typing code. It was working out the factory’s cut rules from about 25,300 past openings in the old program, then proving them against recent jobs. AI made that analysis faster. It didn’t replace the need to do it properly.
What this means for you: software built around your business, owned by you, at a price and timeline that used to be out of reach. A single automation starts from about $1,000 and can run in about a week. A hub across every department is a bigger build, and our guide to custom software costs explains what drives the price.
What AI does inside the hub
Once your data lives in one place, AI agents can work on it. They’re good at the jobs where someone reads something messy and decides what to do with it:
- Sorting and drafting email. At Raw Blend, Claude sorts the shared inbox and drafts replies. Staff read and send every one.
- Writing notes. At the clinic, Claude turns a session transcript into notes and keeps a running summary for each client. Names are swapped for a code before anything reaches the model, and a practitioner approves every draft.
- Explaining numbers. In a reporting dashboard we built as a prototype for a 150-room resort hotel in Brazil, the code works out every figure and the AI only explains them. Its monthly revenue report lands within 0.05% of the hotel’s own figures.
- Checking work before it costs money. In the blinds system, AI explains the pre-cut check that compares what was quoted, what the customer signed and what’s about to be made. It never sees customer names, addresses or phone numbers.
Two rules run through all of these. Code does the maths, because a language model can produce a total that looks right and isn’t. And a person approves anything that leaves the business, whether it’s a customer reply, a clinical note or an invoice.
A chatbot on its own can’t do any of this, because it only knows what someone pastes into it. It can’t see the order in Shopify or the invoice in Xero. The agents above work because the hub gives them the real record.
When you don’t need any of this
Custom software isn’t the answer for every business, and it’s better to know that before you spend money.
- If you’re two or three people, off-the-shelf apps and a spreadsheet are usually right. The cost of the gaps is small, and you’re still working out how you run.
- If your apps already talk to each other through their built-in connections and nobody re-types anything, leave them alone.
- If the job is one person drafting the odd letter, a ChatGPT, Claude or Copilot subscription is enough. Check its data terms before anyone pastes in customer details.
The signs you’ve outgrown the pile: the same details are typed into more than one system, someone spends part of every week rebuilding reports, and errors keep surfacing between departments. Those are the jobs a hub takes away.
Where to start
Draw your business on one page. Write down every app and spreadsheet you use, and draw a line wherever information moves from one to another. Mark each line where a person does the moving: re-typing, exporting, copying, chasing.
Then time the worst two or three for a week. That map and those hours tell you where a hub would earn its keep, and they’re the first thing any good developer should ask you for. Our guide on what to automate first covers how to rank them.