How to start carrier portal automation: five practical steps

March 11, 2026 · Updated September 5, 2026

Start carrier automation with a task your agency can describe and verify. A useful first scope includes a carrier, line, state, transaction, starting information, and expected result. “Automate all quoting” is too broad to tell the team what success looks like.

This guide describes a practical rollout method. Timing and support depend on the product and workflow; a website demonstration does not establish that your full carrier list is ready.

1. Choose a frequent, understandable task

Review a sample of the team's current work. Look for repeated preparation or entry that produces a clear deliverable. Include the time spent gathering information, checking results, and handling exceptions.

Choose a case the team understands well enough to validate. The highest-volume carrier is useful only if the intended workflow is supported and the output can be reviewed. A smaller, well-defined task may be a better starting point.

Write a short task brief: “Prepare this supported homeowners quote from reviewed client information and return the available quote details and outstanding questions.” Keep the expected result precise.

2. Establish the source information

Use the documents and data the agency is permitted to provide. A dec page, application, or schedule may be the starting point. Identify changed exposures and unanswered carrier questions before treating that information as complete.

If a connection to the AMS is proposed, ask what it actually transfers, in which direction, and how changes or failures are handled. An export or document upload is a different handoff from live synchronization. Confirm the path you will use rather than assuming an integration.

Keep source information available for review. The person inspecting the result should be able to establish where important values came from.

3. Confirm the approved access arrangement

Agree on the users and permissions needed for the task. Review credential handling, authentication, session expiry, and what happens when MFA requires a person. Do not disable carrier security controls to make a demonstration easier.

Clarify who can start, pause, resume, and review the workflow. A working login is only one part of authorization. The actions performed must fit the agency's carrier arrangement and internal authority.

Assign a contact for access problems and a process for removing permissions when responsibilities change.

4. Validate with a permitted example

Use an approved test environment or a permitted representative example with the provider. Protect client information and avoid creating duplicate submissions or unintended external actions. A real-looking demonstration should not require an unnecessary production transaction.

Check more than whether the browser moved:

• Important fields match the reviewed source and the intended period.
• Missing answers are surfaced instead of guessed.
• The result has a clear status and preserves relevant terms.
• An interruption identifies the last confirmed step.
• The team understands how to correct or resume the task.
• The reviewed outcome reaches the right person and record.

Include an incomplete example. It reveals how the process behaves when the agency cannot finish immediately.

5. Expand from observed results

Measure active preparation, review, correction, and completed handoffs on comparable work. Track failures and the time needed to recover. Establish who maintains the workflow when a carrier changes its process.

Add another carrier or task only after confirming its own support. A successful path for one line or state does not prove another. Keep a small internal record of the scope validated and the exceptions the team knows how to handle.

Starting with Relay

Use the AI Quoting page to review the current product workflow and confirm your intended scope during setup. The document parsing and proposal workflows may also be useful before or after the browser task.

For the business case, use your agency's observed work in the quoting cost worksheet. A benchmark measures its stated test conditions; it does not establish a guaranteed outcome on your book.