How does gotoHuman's human-in-the-loop review process work step by step?

gotoHuman is a platform for inserting human review steps into AI workflows: you define a review type once, your workflow sends a review request when it needs a decision, a reviewer approves or edits the output in an inbox or via notification, and the result returns to the workflow through a webhook. This four-step loop fits event-driven or long-running AI workflows where a human must approve, correct, or regenerate AI output before it moves forward. You'll need a gotoHuman account, a review type, and a workflow that can call an integration and receive a webhook.

Step 1: Create a review type with the no-code editor

A review type defines what the reviewer sees and what input gets captured. In the no-code editor, you choose the content fields (text, images, markdown, JSON) and the input controls (buttons, checkboxes, selections) appropriate to the decision.

Design choices worth making here:

• Select vs. edit — if reviewers should only pick an option, offer selections; if they should fix AI-generated copy, turn on editing for that field.
• Retry controls — enable retry on fields so a reviewer can loop back to the AI step and regenerate the whole artifact instead of hand-editing it.
• Prompt editing — where a one-click retry isn't enough, pass the prompt along in the request so reviewers can change it.

The expected result is a reusable review type your workflow can reference by name.

Step 2: Send a review request from your AI workflow

From the workflow, request a human review using one of gotoHuman's integrations. The request references the review type and includes all relevant content the reviewer needs — the generated text, the source data, or the artifact version.

This is the point where the workflow pauses and waits. Because the platform is described as platform-agnostic, the integration you use depends on how your workflow is built; the constant is that the request carries the review type reference plus the content payload.

Step 3: A reviewer handles the item

Reviewers find pending items in the Agent Inbox, which collects everything needing input — whether assigned to them personally or to the whole team. Alternatively, they can open a review directly from a notification.

Notification options:

• Email digests for batched review
• Slack notifications for in-channel alerts
• Short-lived public links, available if you opt in, for reviewers without direct platform access

Inside the review, the reviewer approves, selects, or edits content in place. For multi-step workflows, reviewers get navigation across the review steps of a single run, and artifact versioning lets them compare generated versions before approving.

Step 4: The completed review returns via webhook

When the review is completed, gotoHuman sends the result to your workflow via webhook, including the feedback the reviewer provided. The workflow resumes and continues with that input — approved content moves on, edited content replaces the original, or a retry signal sends the step back to the AI.

What the loop looks like end to end

Common sticking points

• Webhook not resuming the workflow — confirm your workflow is actually listening for the completion event and that the review type referenced in the request matches the one you built.
• Reviewer can't edit — editing is a per-field setting; if a field was configured as selection-only, reviewers can only choose.
• Retries regenerate too much — retry loops regenerate the entire artifact, so scope which fields carry retry controls.
• Notification links expiring — public links are intentionally short-lived; for recurring reviewers, the Agent Inbox or Slack is more reliable.

For teams deciding whether this fits: gotoHuman is aimed at AI builders who need approval gates on customer-facing messages, brand-sensitive content, or high-stakes decisions. If your workflow can call an integration and accept a webhook, the loop is four steps and no custom review UI to build.