Synthetic / technical evidence

Same decision core. Two separate configurations. No shared decision thresholds.

This public example shows one e-commerce exception-routing core tested under two isolated configurations. No customer data was used. Every result belongs to a frozen synthetic acceptance set.

Frozen cases12

6 cases for configuration A + 6 for configuration B.

Expected route matches12/12

Evaluated with the same deterministic control sequence.

Live customer measurement—

No production usage or second-pilot evidence is claimed.

Core

The decision sequence does not change

Configuration changes thresholds and permissions. It does not bypass the shared control sequence.

1Required fields
2Untrusted instruction
3Source conflict
4Source freshness
5Authority / threshold
6Approved route
Separate configuration

One core, different operating boundaries

Example configuration A

Narrower currency scope and lower human-approval threshold

Data namespacespec-a
Allowed currencyTRY
Source freshness20 minutes
Human approvalAbove 500 units
Autonomous actionQueue only
Example configuration B

Broader currency scope and a different approval threshold

Data namespacespec-b
Allowed currencyTRY + EUR
Source freshness45 minutes
Human approvalAbove 1,500 units
Autonomous actionQueue + tag
Evidence

Different configuration, different route; same decision core

Example: the same 780 TRY queue case requires human approval in A but may auto-queue in B. EUR is out of scope for A and allowed for B.

CaseConfigCheckExpected route
A-01ANormal / 150 TRYAuto-queue
A-02A780 TRY > thresholdHuman approval
A-03ARequired field missingStop: missing field
A-04AStale sourceStop: stale source
A-05AEUR out of scopeStop: currency scope
A-06AEmbedded instructionStop: untrusted instruction
B-01BNormal / 780 TRYAuto-queue
B-02B1,900 TRY > thresholdHuman approval
B-03BSource conflictStop: source conflict
B-04BEUR allowedAuto-queue
B-05BRefund action not authorizedHuman approval
B-06BStale sourceStop: stale source

Open the frozen synthetic case file →

Pilot measurement

If there is no live usage, we do not enter “0”.

These are production-pilot measurement fields. They remain blank here because no live pilot is being claimed.

Activation—

First run against the agreed acceptance criterion.

Time to first value—

Time from activation to the first accepted business outcome.

Usage / accepted output—

Total usage, human review and accepted outcomes are tracked separately.

Readiness

10 controls before a paid pilot starts

1One workflow and one accountable owner are written down.
2Baseline and acceptance criterion are explicit.
3Data sources and minimum fields are listed.
4Configuration identity and data namespace are separate.
5External-action permissions are explicitly bounded.
6Human approval / escalation owner is known.
7Fallback and stop paths are testable.
8A frozen acceptance set exists.
9Activation, usage and time-to-value fields are defined.
10Scope, price and change boundaries are written.

Evidence boundary

This is not a customer result, production-accuracy claim, multi-tenant security validation, SLA, revenue-uplift claim or ROI claim. The isolated-configuration example demonstrates a technical design pattern and frozen synthetic acceptance test only. Real-environment isolation must be validated separately.

Same commercial path

Freeze the workflow scope and baseline before building.

A Controlled Pilot is offered only when Process Analysis shows a suitable workflow. No free custom development. International prices below are separate market list prices, not FX conversions of Turkish prices.

Process Analysis
$149 starting price
Controlled Pilot
$750 starting price
Monthly Management
$449/month starting price
Request Process Analysis