Platform Platform
System
ConceptsEnginePolicy as codeDeclarationsSafe changeGatewaysIntegrationsObservabilityAdministrationSecurityHuman reviewAudit and evidenceData retentionSecrets and data classification
Controls
Registries and documentationAuthentication and authorizationInjection detectionData redactionCode fingerprintingRole and judge checksContent classificationSpend and loop limitsBusiness rules
Solutions Solutions
By what you do
Sell into the enterpriseControl the AI you run
By industry
Financial servicesDigital assetsInsuranceHealthcareLegalUser-generated content
By discipline
AI governanceTrust and safetyRisk and compliance
Cases Cases Embedded control planeSource-code leakTrading agents over MCPLive firehoseRefund assistant
Compare Compare LiteLLMNVIDIA NeMo GuardrailsOPAROOSTAgent Governance Toolkit
Resources Resources
Guides
Enterprise review questionsPrompt injectionAgent and control layerAgent architecturesDecision system mapAI control maturity model
Standards
Standards OWASP Agent Control StandardEU AI ActPMI AI standardNIST AI RMFERC-8004
Book a demo
Platform · System

A rule opens a case. A person decides it.

Most decisions should run automatically. The few that should not are exactly the ones you will be asked about later.

Built into the Swiftward service you run.

The loop

  1. A rule flags itthe rule decides what goes to a person
  2. A case opensqueue, priority, owner
  3. A person decidesin a screen you declared
  4. Back into the engineas its own event, recorded like an automatic verdict, with the person and their note
rule flags it case opens queue, priority, owner a person decides back into the engine as its own event recorded exactly like an automatic verdict, with the person and their note attached

Because the reviewer's decision goes back into the engine as an event, "who approved this" is answered from the record, and nobody keeps a spreadsheet.

What a rule can open

A case with a queue, a priority and an owner. A timer is optional and off by default, because a queue nobody is watching should not approve things on its own at 3am. When the timer runs out, the case gets the answer you must declare in advance: reject, approve, or escalate.

Each deployment declares its own escalation path, such as a senior reviewer's queue. Once a case is decided it cannot be escalated: the server refuses, rather than trusting the screen to hide the button.

The reviewers are part of the design

The screen a reviewer works in is declared, so a queue carrying graphic material can protect the people who work it. An image blurred by default, or a field hidden until someone asks for it, is part of the screen rather than a habit you ask people to keep.

An appeal is a second queue you declare, with its own reviewers, so the person who reviews an appeal is not the person whose decision is being appealed.

Or send it to your own tracker

If your team already works in a tracker, a standard webhook integration hands the case to your system and brings the decision back when it is resolved.

Related: where the decision lands · moderation queues at volume
Book a demo