Documentation / QueryTek Review glossary / Tier 3 / Embedded Human Evaluation
Embedded human evaluation places a human review step inside another product's workflow through an API, rather than as a separate process users leave to perform. The host application requests evaluation, work continues elsewhere, and the outcome returns to the flow that requested it.
The distinction from ordinary review is architectural. Standalone review is a destination: someone exports material, reviews it in another system, and brings the result back. Embedded review is a step: the host application submits material through an API boundary, a human evaluation tier does the work, and the outcome returns to the workflow that requested it.
Three properties make this workable.
It is asynchronous by nature. Human evaluation takes minutes to days, so the host application cannot block on it. The integration submits, receives an identifier, and is notified on completion — which means the host must be designed for material in a pending state.
Context travels with the request. The evaluator needs to know which standard applies and what the material is for. An integration that submits content without context gets a generic assessment, which is rarely what the workflow needed.
Identity and tenancy come along. Submitted material belongs to a customer, and the evaluation happens within that tenant boundary. An embedded integration that loses tenant context has moved material outside the boundary its owner assumed, which is a more serious failure than a slow response.
The value is that evaluation becomes part of the product rather than an operational process around it. An organization can offer human-verified output as a feature without building a review operation, and users encounter verification as a state of their work rather than a separate errand.
The corresponding cost is that a workflow now depends on human throughput, so capacity, turnaround, and degradation behaviour become product concerns rather than vendor concerns.
Products increasingly need a human judgement they cannot automate — verifying generated output, assessing sensitive material, or applying a standard requiring expertise. Building that operation in-house is a considerable undertaking.
Consuming it through an API keeps the capability inside the product experience, provided the integration carries context and tenancy rather than only content.
QueryTek Review exposes human evaluation through versioned API surfaces so a host workflow can request review and receive outcomes, with each request resolving to a tenant. Endpoint detail and integration authorization are handled through onboarding.
