06 — Process Twin
The business, as it actually runs.
Mined from real event logs and continuously improved through Six Sigma, TQM and TRIZ — every path, queue, rework loop and exception is diagnosed, simulated and optimized before any change is funded.
North Star
Throughput · Touch time per case · Sigma level
The stream
Automation → Capability → Value driver → Financial value
Automation
Continuous process mining and conformance
Capability
See the real process, not the diagram
Value driver
Operational efficiency
Financial value
Lower cost per transaction
Automation
Six Sigma / DMAIC optimization loop
Capability
Define, measure, analyze, improve and control with statistical rigor
Value driver
Defect reduction
Financial value
Cost of poor quality falls
Automation
TQM voice-of-customer and continuous improvement
Capability
Customer-critical quality embedded into every step
Value driver
First-time-right
Financial value
Less rework, higher retention
Automation
TRIZ contradiction resolution and inventive problem solving
Capability
Break trade-offs without adding cost or complexity
Value driver
Innovation yield
Financial value
Throughput up, resources flat
Automation
Exception handling agents
Capability
Resolve the tail without new headcount
Value driver
Capacity
Financial value
Deferred hiring, higher throughput
Automation
Orchestration across systems
Capability
Straight-through processing
Value driver
Cycle time
Financial value
Faster revenue and service delivery
Value system
Value ≠ North Star ≠ KPI ≠ driver ≠ process metric.
The translation layer that connects a corporate objective to something an operating team can actually move on Monday.
Value
Operational throughput and quality
Economic outcome
Cost-to-serve reduction at constant or improving quality
North Star
Cost per transaction at target service level and sigma level
Outcome KPIs
Driver metrics
Process metrics
Causal model
Value leakage
Where the value goes missing today.
Leakage is multiplicative. Every gate that stays leaky discounts everything built upstream of it.
Intake
Incomplete cases enter the flow and bounce.
Rework
Routing
Work assigned by availability, not fit.
Cycle time
Handoffs
Each transfer adds wait and context loss.
Throughput
Exceptions
The 12% of odd cases consume 60% of effort.
Cost-to-serve
Variation
Special-cause variation goes unnoticed until it becomes a defect.
Quality cost
Trade-offs
Throughput and quality treated as conflicting goals.
Innovation yield
Digital twin
Simulate the interventions before anyone funds them.
Instead of implementing ten recommendations and learning the result in six months, the twin stacks them first.
Cost per transaction
Guardrails
- Service level maintained
- Quality/first-time-right not reduced
- Sigma level holds or improves
- No unattended action on high-value cases
Constraints
- Union and labour agreements
- System change windows
- Peak-season capacity
- Quality standards and certification requirements
Automation
What runs without asking.
Bottleneck detection
Ranks friction by dollars lost, not case count.
Process Mining Agent
DMAIC project runner
Guides each improvement through Define, Measure, Analyze, Improve, Control with live data and control charts.
Six Sigma / DMAIC Agent
Voice-of-customer prioritization
Links customer pain to process steps and ranks improvement priorities.
TQM Continuous Improvement Agent
Contradiction solver
Applies TRIZ principles to resolve throughput-vs-quality or speed-vs-cost trade-offs.
TRIZ Innovation Agent
Statistical process control
Control charts flag special-cause variation before it becomes a defect.
Six Sigma / DMAIC Agent
Exception resolution
Handles the non-standard cases with human escalation.
Exception Handling Agent
Cross-system orchestration
Executes steps across ERP, ITSM and workflow tools.
Orchestration Agent
Process Mining Agent
Keeps the twin conformant with reality.
Process mining + conformance checking
Six Sigma / DMAIC Agent
Runs DMAIC cycles, control charts and capability studies.
Statistical process control + optimization solver
TQM Continuous Improvement Agent
Prioritizes VoC feedback and drives closed-loop quality improvement.
SLM sentiment + RAG on quality standards
TRIZ Innovation Agent
Maps contradictions and proposes inventive solutions that avoid trade-offs.
LLM + TRIZ contradiction matrix + patent/standard knowledge
Exception Handling Agent
Reasons through the tail cases.
LLM + RAG on SOPs
Orchestration Agent
Executes multi-system workflows safely.
SLM routing + tool contracts
Integrations
Where it plugs into the domain.
Vendor-agnostic by design. The twin reads and writes through whatever stack the domain already runs on.
Systems of record
- ERP
- ITSM / service management
- Core industry systems
Process excellence
- Celonis
- SAP Signavio
- UiPath Process Mining
- Microsoft Process Mining
- Apromore
- Minitab
- JMP
Signals
- Event logs / CDC streams
- Workflow & RPA platforms
- Telephony and case notes
- Quality management system (QMS)
- Customer feedback / VoC platforms
Data & runtime
- Graph + time-series store
- SPC / control-chart engine
- Simulation engine
- MCP connectors
Domain platforms we connect
Process & task mining
Automation & orchestration
Systems of record
Service & case management
Workforce & scheduling
Data & runtime
Plus anything else with an API — connectors are added per engagement, not sold as a platform lock-in.
Value
The levers, and what they move.
Cost
Remove rework and manual touches.
Cost per case · Touch time
Quality
Reduce defects and variation with Six Sigma and SPC.
DPMO · Sigma level · Cost of poor quality
Speed
Cut queue and wait time end to end.
Lead time · Straight-through rate
Experience
Fewer failures reaching the customer.
First-time-right · CSAT
Innovation
Break throughput-vs-quality trade-offs with TRIZ.
Ideas implemented · Constraint removal rate
Risk
What could go wrong, and what stops it.
Every risk in this domain has a named containment in the runtime — not a slide.
Automation amplifying a broken process
Process mining + DMAIC Define/Measure establish the as-is before any agent is built
Medium
Workforce impact and change resistance
Redeployment plan agreed before deployment; humans stay on exceptions and improvement cycles
Medium
Silent failure in an unattended flow
Outcome monitors with automatic pause and rollback
High
Statistical overfitting on small samples
Control charts and capability studies require minimum sample sizes and special-cause rules
Medium
Compliance · Security
Built into the runtime, not bolted on.
EU AI Act
Limited risk; high-risk controls where processes affect essential services or worker management.
- Human oversight on any action affecting a customer entitlement
- Logging of automated decisions with reversal path
- Impact assessment before removing a human step
- Statistical models affecting workers reviewed for bias and proportionality
GDPR
- Event logs pseudonymized before mining
- No worker-level performance profiling from process data
- Purpose limitation between operational and analytical use
- Customer feedback (VoC) processed on documented lawful basis
Security
- Scoped service accounts per tool contract
- Rate limits and blast-radius caps on write actions
- Full trace and replay of agent runs
Controls & oversight
- Risk-tiered human-in-the-loop
- Rollback procedure per automated action
- Change control on orchestration flows
- DMAIC gate reviews before production rollout
Agentic execution
Agents earn authority. They are not given it.
No agent in this twin controls anything it has not first proven in replay, evaluation, simulation and shadow.
01
Historical replay
Re-run the last 12 months. Would the agent have been right?
02
Offline evaluation
Scored against held-out outcomes, not opinion.
03
Digital twin
Simulated against the causal model under stress.
04
Shadow mode
Runs live, decides nothing. Divergence is logged.
05
Human recommendation
Proposes; a person executes and rates it.
06
Bounded pilot
One segment, capped exposure, hard rollback.
07
Human-supervised execution
Acts inside thresholds, humans approve exceptions.
08
Progressive autonomy
Authority widens only where evidence widened.
The unit of value
Don't buy transformation. Buy measurable movement.
This twin is contracted the way it is engineered: a baseline, a target, a guardrail, and an attribution method agreed in advance.
Baseline
Cost per transaction $14.20, 3.2 sigma
Target
$7.50 at unchanged SLA with sigma level ≥ 4.0
Guardrail
First-time-right rate and sigma level hold or improve
Proof / attribution
Process-mining event logs and SPC charts against a matched control cohort
Next twin
AI Twin →