It is nine on a Monday morning, and the finance lead at a restaurant and hotel group is staring at two reports that should say the same thing. The point-of-sale system says the weekend's food sales were strong. The inventory spreadsheet says the kitchens barely used any stock. The booking system shows a full house on Saturday, but the staffing sheet was built from last week's forecast. Someone has already copied numbers from one screen into another three times, and the figures still do not agree.
Down the hall, a general manager is on the phone with a supplier who says an order was never received. It was placed, but only in the purchasing tool, which does not talk to the kitchen's stock list. Nobody did anything wrong. Every system works. They simply do not work together, and the people in the middle have become the integration.
If this sounds like your business, you are not short of software. You are short of connections between it.
In short: to integrate disconnected business systems, first map which data has to move between them and who owns it, then connect the systems through a shared layer, a central place where data is validated and passed on, rather than a growing web of one-off links. Start with the one flow that hurts most and extend from there.
Why do disconnected business systems keep appearing?
Picture how the group got here. The restaurants bought a point-of-sale system years ago. The hotel brought its own booking platform. Finance chose accounting software. Purchasing started with a spreadsheet because it was quick, and the spreadsheet stayed. Each choice made sense on the day it was made.
Over time each team added what it needed, and each tool kept its own version of the truth: its own supplier names, its own item codes, its own idea of what counts as a "sale". Nobody planned the gaps. They grew in the spaces between decisions.
The cost shows up in small, constant ways. Rekeying data. Reports that need a day of reconciling before anyone trusts them. Questions that should take a minute and take an afternoon. And errors that nobody can trace, because the number passed through four hands and two spreadsheets on its way to the report.
What happens when people become the integration?
The finance lead, call her Maya, starts keeping a private workbook that matches the point-of-sale totals to the bank and the accounting entries. It works, as long as she is in. When she takes leave, month-end slips by days, because only she knows which columns to adjust and why.
Meanwhile the operations manager asks for a shortcut. IT sets up an export from the booking system that lands in a shared folder every night. Then another for purchasing. Then one for payroll. Each is a direct link between two systems, and each seemed small. But when the booking vendor changes a field, two of the links quietly break, and nobody finds out until a report looks odd.
This is the pattern to recognise. Point-to-point connections are fast to build and fine when only a few systems are involved. As the number of systems grows, the number of links grows faster, and every change in one system can ripple into several others. The workaround that saved a week becomes the thing that costs a week every month.
How do you integrate disconnected business systems properly?
The group's leadership decides to stop patching and to fix it once. The sequence they follow is the one that tends to work in any industry.
1. Name the flows, not the systems
They do not begin by listing software. They list the questions the business cannot answer easily: what did we really sell and use, what did we order and receive, what is each venue's labour against its bookings. Each question is a flow of data between two or more systems. Ranking those by pain shows where to start.
2. Decide who owns each piece of data
A supplier, a menu item, a room type, a guest: each should have exactly one place where it is created and corrected. Once ownership is agreed, the other systems read from that source instead of keeping their own copies. Most reconciliation work disappears at this step, before any code is written.
3. Connect through a shared layer
Instead of wiring each system to every other, they build a central integration layer. Each system connects to it once, using the system's own interface for exchanging data. The layer validates what arrives, translates between formats, routes it to wherever it is needed and records what happened. When the booking vendor changes a field, one mapping is updated, not four links.
4. Make failures visible
An integration that fails silently is worse than none. The layer needs clear alerts and a simple log so that a failed order or a mismatched total shows up the same day, with enough detail for someone to fix it. This is what turns a fragile setup into something people trust.
5. Deliver one flow at a time
They begin with purchasing to stock to finance, because that is where Maya loses the most hours. It runs alongside the old process for a short time so everyone can compare results. Once it is trusted, the next flow is added. Each step is useful on its own, which keeps the project from becoming a long wait for a big finish.
Should you buy an integration tool or build one?
It is a fair question, and the honest answer depends on your systems. Ready-made integration platforms can be a good fit when your systems are common and your flows are standard. Many of the guides on this topic are written by integration vendors, so weigh their advice against your own situation.
Custom work earns its place when your processes are unusual, when an older system has no good interface, or when the rules about how data should move are specific to your business. The group in our story has all three: an older purchasing spreadsheet, a booking platform with limited export options, and approval rules that differ by venue. A layer built around those realities fits better than forcing the business into someone else's template.
Whichever route you take, the principles do not change: clear ownership, one shared layer, visible failures and small steps.
What changes when the systems finally talk to each other?
Months later, Monday morning looks different. The weekend's sales, stock use and bookings reach the finance lead's screen already matched. Where something does not agree, it is flagged with the reason, so she spends her time on the exceptions instead of the whole pile. The general manager's supplier order appears in the kitchen's stock list the moment it is placed. Staffing is planned against the actual bookings, not last week's forecast.
Nothing dramatic happened. The tools are the same ones the group already had. What changed is that data now moves on its own, and the people who used to carry it can use their judgment instead.
Frequently asked questions
What does it mean to integrate business systems?
It means setting up automatic, reliable exchange of data between the software your teams already use, so information is entered once and appears wherever it is needed. The aim is a single trusted version of each fact, such as a sale, an order or a supplier.
Is point-to-point integration always a bad idea?
No. Connecting two stable systems directly can be perfectly sensible and quick. It becomes a problem as more systems and flows are added, because each new link raises the effort of every later change.
Where should we start?
Start with the flow that costs your people the most manual effort or causes the most errors, and prove it end to end before adding more. A small, working connection teaches you more about your data than any plan on paper.
Which solution fits
In the story, the group's real problem is not any single tool. It is that its workflows, approvals and data rules are spread across systems that were never designed to cooperate. The best fit is Custom Enterprise Software: a purpose-built integration layer and the screens around it, shaped to how the venues actually run. If the biggest pain is manual matching and exception handling, AI & Intelligent Automation can add automatic reconciliation and flagging on top.
Polynode builds custom, AI-powered software around how each business actually runs, scoped first and delivered in weekly iterations, so the first connected flow can be working and trusted before the next one begins. If your Mondays look like the one above, talk to our team.
Cover photo by Olena Kamenetska on Unsplash.