Connect Disconnected Business Software
Last month's client report took your senior consultant most of a Friday. She pulled hours from the time tracker, matched them to projects in the project tool, checked contract terms in a shared folder, then copied the totals into an invoice. Along the way she found that a client's name was spelled three ways and that one project had been closed in one tool and left open in another. Nobody did anything wrong. You simply own nine software subscriptions, and they have never met each other.
If you run a professional services firm, this is probably your week. This article compares the main ways to connect those tools, explains the technical choices in plain language, and gives you a decision method. You do not need to write code to follow it, though you will see where code earns its place.
First, draw the map you already have
Before choosing a technology, write down what you own. Use one row per tool and four columns.
| Tool | What it holds | Who uses it | What it should hand to others |
|---|---|---|---|
| Time tracker | Hours by person and project | Everyone | Hours to billing |
| CRM | Clients, contacts, deals | Partners, sales | Won deals to project tool |
| Project tool | Tasks, status, deadlines | Delivery team | Status to client reports |
| Billing | Invoices, payments | Finance | Payment status to CRM |
| Document storage | Contracts, deliverables | Everyone | Links to projects |
Add your remaining four tools. Then draw arrows for every place a human currently carries information from one tool to another. Those arrows are your real list of integration candidates. Most firms discover between ten and twenty, and that two or three of them consume most of the effort.
Four ways to connect software
Connection methods differ in effort, control, and fragility. Here are the four you will meet.
Native integrations
Many tools ship with a built-in connection to another popular tool. You click a button, authorize access, and pick a few settings. These are the cheapest option and often the best one for simple needs. Their limits are that you get only what the vendor decided to support, field mapping is usually shallow, and if the vendor changes the connection you have little say. Use native integrations first when they cover the whole job.
Point-to-point links
A point-to-point link joins exactly two tools, often through a no-code service or a small script. With three tools you may need three links. With nine, the theoretical count is 36, because each pair could need its own connection. Real firms do not need all of them, but the trend is clear. Every new tool adds links, each link has its own mapping and failure behavior, and nobody remembers how they all work. This approach is fine for a small number of high-value connections. It becomes a tangle past about five or six tools.
A hub
A hub sits in the middle. Each tool connects once, to the hub, and the hub holds the rules for what moves where. Nine tools means nine connections instead of up to 36. The hub can be an integration platform, a small database with some automation around it, or a purpose-built service. The benefits are one place to monitor, one place to define what a "client" is, and an easier time swapping a tool out. The cost is that the hub itself must be maintained, and it becomes something you depend on.
Shared database or warehouse for reporting
Sometimes you do not need tools to talk to each other at all. You need to see everything together. In that case you copy data from every tool into one reporting store on a schedule and build reports there. This does not fix retyping between tools, but it can fix the Friday report. We will return to this in the decision method.
Webhooks versus polling
Two basic mechanisms decide how one tool learns that something changed in another.
Polling means asking on a schedule. Every five minutes, a connector asks the CRM whether anything is new. It is simple and works almost everywhere. Its downsides are delay, because changes wait for the next check, and wasted effort, since most checks find nothing. Many tools also limit how often you can ask, which is called a rate limit.
Webhooks work the other way. The source tool sends a message to a web address of yours the moment something happens. They are fast and efficient, but they bring responsibilities.
- Verify the sender. A web address that accepts messages from anyone is an open door. Good vendors sign each message with a shared secret, a webhook signature, and your receiver checks the signature before trusting the content.
- Expect duplicates and gaps. Senders retry when they do not get a reply, so the same event can arrive twice. Receivers must handle that safely, which is the idea of idempotency, using an event ID to ignore repeats. Delivery can also fail altogether, so you want a periodic catch-up check.
- Be reachable. If your receiving service is down during a storm or an outage, you need the sender to retry, or a queue that holds messages.
A sensible pattern for a small firm is webhooks for speed with a slow polling pass as a safety net. If a tool offers only polling, accept the delay and design around it.
| Question | Webhooks | Polling |
|---|---|---|
| How fast does it learn of a change | Seconds | Up to the polling interval |
| How hard to build | Needs a receiver and verification | Simple |
| What if you are offline | Depends on sender retries | Catches up next poll |
| Risk of missing changes | Moderate without a safety net | Low if you track a last-seen marker |
No-code platforms versus code
The other big choice is who builds the connection.
No-code and low-code platforms let you drag steps into a flow, for example "when a deal is won, create a project." They are quick to start, understandable by non-programmers, and often cover the popular tools out of the box. Their trade-offs are cost that grows with volume, since many charge per task, limited control over error handling, and flows that become hard to read once they have many branches. They also tie your logic to their platform.
Code gives you full control over mapping, retries, logging, and testing. It costs more up front and needs someone to maintain it, but the running cost per event is tiny and you own the logic. It suits connections that carry money, that must never duplicate, or that need unusual rules.
A fair rule of thumb follows.
- Use a no-code platform for notifications, simple record creation, and anything where a failure is merely annoying.
- Use code, or a hub with real engineering behind it, for billing, payments, client data that must stay consistent, and anything with many branches.
- Whatever you pick, insist on logs you can read and an alert when something fails.
A worked example, with illustrative numbers
This example is hypothetical. Consider a firm of fourteen people with nine tools. The firm decides that the two flows costing the most time are turning a won deal into a project, and turning approved hours into an invoice.
Assume the firm closes 6 deals a month and sends 40 invoices a month. Setting up a project by hand takes 45 minutes across three tools, and compiling an invoice from hours takes 30 minutes. That is 6 times 45, or 270 minutes, plus 40 times 30, or 1,200 minutes. Altogether about 24.5 hours a month. If a loaded staff hour costs $40 in this example, that is around $980 a month of internal time, before counting the cost of billing errors.
The invoice flow touches money, so the firm builds it carefully with idempotency and logs. The project setup flow is low risk, so a no-code platform handles it. The firm ends up with two connections, not 36, and a short document explaining both. The numbers are invented to show the method. Use your own measurements.
How to choose, a decision method
Work through these questions in order.
- Does a native integration cover the whole job? If yes, use it and stop.
- How many tools must share this information? Two tools suggests a point-to-point link. Three or more, with a shared notion of a client or a project, suggests a hub.
- What happens if it fails or duplicates? If the answer involves money or a client-facing error, favor code and tested retries. If it is only annoying, no-code is fine.
- How fast must it be? If minutes do not matter, polling is acceptable. If seconds matter, favor webhooks.
- Who will maintain it in two years? If nobody on staff can, choose the approach you can hand to a provider or a freelancer, and keep documentation.
- Is the real need reporting, not synchronization? Then a reporting copy may beat a web of live connections.
Failure modes to plan for
- Expired credentials. Connections rely on tokens or keys. When one expires or a staff member who authorized it leaves, the flow stops. Use a shared service account, not a person's login, and calendar the renewal.
- Excess permissions. When you authorize a connection, you grant it scopes, which define what it may read or change. Grant only what the flow needs.
- Vendor changes. A tool changes a field or retires a version of its interface, and your flow breaks quietly. Subscribe to vendor change notices.
- Silent drift. Nobody notices that records disagree. Schedule a monthly count comparison.
- Rate limits. A bulk import trips a limit and the flow fails halfway. Design for batching and resuming.
What to ask a vendor
- Does your integration use webhooks, polling, or both, and how often?
- How do you handle duplicates and retries?
- Which fields can I map, and can I add custom ones?
- What permissions will the connection request, and can they be narrowed?
- Where can I see logs, and how will I be alerted to failures?
- If I leave, can I export my data and my settings?
Monday morning
- Build the tool map with the four columns above.
- Mark every spot where a person copies data from one tool to another.
- Rank the arrows by minutes per month and by cost of an error.
- Pick one arrow and run it through the decision method.
- Name an owner for each connection, with a backup.
Puerto Rico adds its own wrinkles, such as storm season outages and bilingual client records, so test what happens to a flow when your connection drops, and make sure names with accents survive the trip between tools.
If you want help drawing the map or choosing between a hub and a few links, Ozcorp InfraOps does this kind of work for small firms. InfraOps Base starts at $500 per month, with vendor and license charges billed separately. Whether you do it with us or on your own, begin with the map. It is usually the most clarifying hour of the project.
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.