Integration

When systems do not talk to each other, people become the connection. They copy figures between screens, reconcile spreadsheets and answer questions that one system could answer if it knew what the other knew. That costs hours, adds typing mistakes and leaves each team looking at a different version of a customer.

Integration is what makes a single source of truth possible. It also underpins workflow automation, since an automation can only act on data it can reach.

• Map the data flows that matter before connecting anything: what must move, from where, to where.
• Start with the biggest gap, such as the task with the most manual entry.
• Prefer a native integration when it covers the need. It needs the least upkeep.
• Decide the direction of sync. One-way is simpler. Two-way can create conflicts when both sides change the same record.
• Store access keys in the tool's secure settings, never in a shared document or a chat.
• Monitor it. An integration that fails quietly leaves stale data for weeks.

An integration links two tools so that something happening in one shows up in the other. A prospect fills in a form and appears in the CRM. A customer pays an invoice and the deal is marked as won. A support ticket opens and the account owner is notified.

There are three common ways to build one:

• Native: the two products offer a ready-made connection that is switched on in settings.
• Through a connector tool: a service such as Zapier or Make sits between the two and passes data along.
• Custom: someone writes code against the tools' APIs, the interfaces software offers to other software. This is often called an API integration. Many integrations also rely on a webhook, where one tool sends a message the moment something changes.

• Connecting before mapping. Without a clear flow, two tools end up with duplicate or conflicting records. Decide which tool owns each field first.
• Two-way sync by default. If both sides can change the same record, changes can collide. Start one-way unless a second direction is truly needed.
• No monitoring. A connection that fails quietly leaves stale data for weeks. Alert on failures and check a sample of records each month.
• Building custom when native exists. A native integration is maintained by the vendor. Custom code is yours to fix whenever an API changes.
• One person who knows how it works. Write down what each integration does, who owns it and where its keys are stored.
• Treating integration and automation as the same. An integration moves data between tools. A workflow automation acts on that data, with rules and steps.