Write-back

What it is

Integration claims come in three grades. Read-only: the agent can look things up but not change them. Notification: the agent sends an email, a text or a dashboard entry that a person must act on. Write-back: the agent itself creates or updates the record, with the right fields, in your system, before the call ends. Only the third grade removes the human step.

Write-back is done through the target system's API, through a vendor-to-vendor partner integration, or, in weaker products, through screen automation that breaks when the target's interface changes.

Why it matters when buying

The business case usually assumes write-back and the product often delivers notification. The gap shows up after go-live as a staff member re-keying bookings from an inbox. Write-back also carries the risk: a wrong write is worse than no write, which is why the details must be read back and confirmed before the call fires, and why permissions should be scoped to the fields the agent needs.

What to ask

In the demo, make a booking or an update and watch it appear in a sandbox of your own system while the call is still running. Ask which fields are written and which are left for staff. Ask how the agent behaves when the write fails (the system is down, the slot was taken a second earlier). Ask how writes are audited and reversed. A vendor who can only show their own dashboard has shown you a message, not an integration.