Skip to main content

Finance Workflow Comparison

This walkthrough compares four implementations of the same high-risk finance action (reverse_wire_transfer).

Workflow Input​

Common input fields:

  • request id
  • amount and target account
  • runtime identity + attestation
  • evidence set (ticketLinked, ledgerSnapshot, beneficiaryValidated)
  • optional approval token

Reference source:

  • https://github.com/UAICP/uaicp-reference-impl/blob/main/src/examples/finance/workflow-comparison.ts

1. Manual Workflow​

Execution style:

  • human checklist and approval before action

Strengths:

  • explicit control and accountability

Limits:

  • slower
  • inconsistent across operators

2. Agentic Workflow (Framework-Native)​

Execution style:

  • orchestration runtime performs tool sequence
  • decision often driven by model confidence + task flow

Strengths:

  • fast and scalable

Limits:

  • incomplete governance guarantees unless explicit gates are added

3. Agent Implementation Without UAICP​

Execution style:

  • automated approval path based on confidence/task success

Observed risk:

  • can approve high-risk actions without complete evidence or explicit approval metadata

4. Agent Implementation With UAICP​

Execution style:

  • identity gate (IdentityValidator)
  • policy gate (PolicyEvaluator)
  • evidence gate (required evidence fields)
  • verification gate (confidence/verification constraints)

Outcome behavior:

  • allow only when all gates pass
  • needs_review for high-risk writes without approval
  • reject/fail_safe for missing evidence or invalid identity/policy conditions

Run the Example​

cd uaicp-reference-impl
npm install
npm run example:finance

Why This Matters​

The comparison makes one point clear: UAICP changes reliability outcomes from implicit best-effort behavior to deterministic, auditable gate semantics.