Agentic Enterprise Patterns¶

Overview¶

Agentic Enterprise Patterns are reusable architectural blueprints for how AI agents participate in enterprise operations. Each pattern describes a proven way to integrate agents into a specific type of enterprise work, with defined human touchpoints, autonomy levels, and governance requirements.

These patterns are technology-agnostic. They describe what the agent does and how it relates to humans — not which platform or model implements it.

Pattern Structure¶

Each pattern is documented with:

Description — What this pattern does
Applicable ZORBA Layers — Where in the architecture this pattern operates
Agent Type — Which agent type (from the Workforce Model) this pattern uses
Human Touchpoints — Where and how humans interact with this pattern
Autonomy Level — The typical autonomy level (A0–A4)
Governance Requirements — What governance must be in place
Example Applications — Concrete enterprise use cases

Pattern 1: Strategy Formulation Assistant¶

Description: An agent that continuously monitors the external and internal environment, synthesises information, generates strategic insights, and prepares materials for human strategic decision-making. It does not make strategic decisions — it ensures that the humans who do are comprehensively informed.

Applicable Layers: L1 (Strategy), L2 (Objectives)

Agent Type: Co-Pilot

Human Touchpoints: - Human defines strategic questions and areas of focus - Human reviews and interrogates agent-generated analyses - Human makes all strategic decisions - Human provides feedback on relevance and quality of insights

Autonomy Level: A1 (Supervised) — all outputs reviewed before influencing decisions

Governance Requirements: - Source attribution for all intelligence gathered - Bias disclosure — agent must flag when data sources are limited or skewed - No autonomous external communications or commitments - Audit trail of all analyses provided to decision-makers - Periodic review of the agent's information sources and analytical methods

Example Applications: - Board meeting preparation — scanning market, competitor, and regulatory developments - Strategic planning cycles — generating scenario analyses and strategic options - M&A opportunity screening — monitoring and evaluating potential targets - Strategic risk radar — continuous monitoring of threats and emerging risks

Pattern 2: Autonomous Process Monitor¶

Description: An agent that continuously observes business processes in real-time, detects deviations from expected behaviour, identifies bottlenecks and quality issues, and either resolves them autonomously (within authority) or escalates to humans.

Applicable Layers: L4 (Processes), L5 (Activities), L6 (Work)

Agent Type: Autonomous Agent

Human Touchpoints: - Human defines monitoring parameters and alert thresholds - Human receives and acts on escalations - Human reviews periodic performance summaries - Human adjusts parameters based on changing business needs

Autonomy Level: A3 (Trusted) — operates autonomously, escalates by exception

Governance Requirements: - Defined escalation triggers (cannot be modified by the agent) - Complete monitoring logs retained for audit - False positive / false negative tracking - Human review of auto-resolved issues on a sampling basis - Clear authority boundaries for autonomous remediation actions

Example Applications: - Order fulfilment monitoring — detecting delays, stock issues, routing problems - SLA compliance monitoring — tracking service levels across customer contracts - Financial control monitoring — detecting anomalous transactions or approval violations - IT operations monitoring — infrastructure health, performance, and security

Pattern 3: Decision Support Analyst¶

Description: An agent that gathers data, performs analysis, models scenarios, and presents structured recommendations to a human decision-maker. The agent prepares the decision; the human makes it.

Applicable Layers: L2 (Objectives), L3 (Capabilities), L4 (Processes)

Agent Type: Specialist Agent

Human Touchpoints: - Human frames the decision and defines the question - Human reviews analysis and recommendations - Human makes the final decision - Human provides feedback on the quality and relevance of analysis

Autonomy Level: A1 (Supervised) for high-stakes decisions; A2 (Guided) for operational decisions

Governance Requirements: - Methodology transparency — the agent must explain its analytical approach - Confidence levels on all recommendations - Sensitivity analysis showing how conclusions change with different assumptions - No cherry-picking — the agent must present countervailing evidence - Decision log linking recommendation to final decision and outcome

Example Applications: - Pricing decisions — market analysis, elasticity modelling, competitive positioning - Resource allocation — capacity analysis, ROI projections, scenario comparison - Vendor selection — multi-criteria evaluation with weighted scoring - Investment decisions — financial modelling, risk assessment, portfolio impact

Pattern 4: Execution Orchestrator¶

Description: An agent that manages the end-to-end execution of a business process, routing work to the appropriate performers (human or agent), managing dependencies and sequencing, monitoring progress, and ensuring completion.

Applicable Layers: L4 (Processes), L5 (Activities)

Agent Type: Orchestrator Agent

Human Touchpoints: - Human defines process design and rules - Human performs assigned decision/judgement steps within the process - Human handles escalations from the orchestrator - Human reviews process performance metrics

Autonomy Level: A3 (Trusted) — manages flow autonomously, escalates exceptions

Governance Requirements: - Process definition is human-approved and version-controlled - Routing logic is transparent and auditable - Cannot modify process definitions without human approval - Full process execution audit trail - SLA tracking with automated alerting - Defined behaviour for process failures and exceptions

Example Applications: - Employee onboarding — coordinating IT setup, HR paperwork, manager introductions, training - Incident management — coordinating detection, triage, investigation, resolution, post-mortem - Procurement cycle — requisition through approval, sourcing, PO, receipt, payment - Campaign launch — coordinating creative, compliance review, channel setup, activation

Pattern 5: Continuous Compliance Guardian¶

Description: An agent that continuously monitors enterprise activities for compliance with regulations, policies, and standards. Operates as an always-on compliance function that catches issues in real-time rather than in periodic audits.

Applicable Layers: L4 (Processes), L5 (Activities), L6 (Work)

Agent Type: Autonomous Agent

Human Touchpoints: - Human defines compliance rules and regulatory interpretations - Human receives and investigates flagged violations - Human makes judgement calls on ambiguous compliance situations - Human reviews false positive rates and adjusts rules

Autonomy Level: A3 (Trusted) for monitoring and flagging; A1 (Supervised) for enforcement actions

Governance Requirements: - Compliance rule definitions require legal/compliance human approval - All flags and enforcement actions logged and auditable - False positive tracking with regular calibration - Cannot autonomously impose sanctions or penalties - Regular testing against known compliance scenarios - Regulatory change monitoring to keep rules current

Example Applications: - Data privacy compliance — monitoring data handling against GDPR/privacy regulations - Financial compliance — transaction monitoring for AML, fraud, sanctions screening - HR compliance — monitoring employment practices against labour regulations - Information security — monitoring access patterns, data flows, and policy adherence

Pattern 6: Intelligent Triage Engine¶

Description: An agent that receives incoming requests, enquiries, or issues, classifies them, assesses priority and urgency, enriches them with context, and routes them to the appropriate handler (human or agent).

Applicable Layers: L4 (Processes), L5 (Activities)

Agent Type: Autonomous Agent

Human Touchpoints: - Human defines classification taxonomy and routing rules - Human handles escalated or ambiguous cases - Human reviews triage accuracy periodically - Human adjusts rules based on changing patterns

Autonomy Level: A3–A4 (Trusted to Autonomous) — high-volume, pattern-based work

Governance Requirements: - Classification accuracy metrics with minimum thresholds - Misrouting tracking and correction protocols - Priority override capability for humans - Full audit trail of triage decisions - Regular retraining/recalibration based on outcomes

Example Applications: - Customer support triage — classifying and routing customer enquiries - IT service desk — categorising and prioritising incidents and requests - Lead qualification — scoring and routing sales leads - Document processing — classifying, extracting, and routing incoming documents

Pattern 7: Knowledge Curator¶

Description: An agent that continuously maintains, updates, and improves the organisation's knowledge base. It ingests new information, identifies outdated content, resolves contradictions, and ensures knowledge is accessible to both humans and agents.

Applicable Layers: L3 (Capabilities), L5 (Activities)

Agent Type: Autonomous Agent

Human Touchpoints: - Human subject matter experts validate significant knowledge updates - Human defines knowledge quality standards - Human reviews agent-proposed changes to critical knowledge articles - Human provides feedback on knowledge gaps

Autonomy Level: A3 (Trusted) for routine maintenance; A2 (Guided) for significant content changes

Governance Requirements: - Version control on all knowledge artefacts - Change approval workflow for sensitive or regulated content - Source attribution for all knowledge claims - Contradiction detection and resolution logging - Knowledge usage analytics to prioritise maintenance

Example Applications: - Customer-facing knowledge base maintenance - Internal policy and procedure documentation - Technical documentation and runbook management - Regulatory and compliance knowledge management

Pattern 8: Predictive Advisor¶

Description: An agent that analyses historical and real-time data to predict future states, risks, or opportunities, and proactively advises humans on actions to take.

Applicable Layers: L2 (Objectives), L3 (Capabilities)

Agent Type: Specialist Agent

Human Touchpoints: - Human defines prediction domains and acceptable risk thresholds - Human receives and evaluates predictions - Human decides on actions based on predictions - Human provides outcome feedback to improve predictions

Autonomy Level: A2 (Guided) — predictions are advisory, not action-triggering

Governance Requirements: - Model transparency and explainability requirements - Confidence intervals on all predictions - Backtesting results published and reviewed - Bias auditing on a regular schedule - No autonomous action based on predictions alone (unless explicitly authorised)

Example Applications: - Customer churn prediction with retention recommendations - Demand forecasting with inventory and staffing recommendations - Equipment failure prediction with maintenance scheduling - Cash flow forecasting with liquidity management recommendations

Pattern 9: Workforce Coordinator¶

Description: An agent that manages the allocation and coordination of work across a blended workforce — assigning tasks to the most appropriate performer (human or agent) based on capability, availability, autonomy level, and workload.

Applicable Layers: L4 (Processes), L5 (Activities), L6 (Work)

Agent Type: Orchestrator Agent

Human Touchpoints: - Human defines workforce allocation policies - Human handles override requests and special assignments - Human reviews workload balance and performance - Human makes staffing and capacity decisions

Autonomy Level: A3 (Trusted) — allocates work autonomously within defined policies

Governance Requirements: - Allocation logic transparency - Fairness monitoring (for human workers) - Workload caps and wellbeing considerations (for human workers) - Agent capacity and performance monitoring - Escalation when no suitable performer is available

Example Applications: - Shared services team coordination — routing work to available humans or agents - Contact centre workforce management — balancing load across human and AI agents - Project resource management — matching tasks to available capability - Field service coordination — scheduling and dispatching human and robotic workers

Pattern 10: Autonomous Executor¶

Description: An agent that independently executes a defined class of work from receipt to completion, with no human involvement in the standard flow. The most autonomous pattern — used for high-volume, well-defined, low-risk work.

Applicable Layers: L5 (Activities), L6 (Work)

Agent Type: Autonomous Agent

Human Touchpoints: - Human defines execution rules and quality standards - Human reviews exception reports - Human performs periodic quality audits - Human adjusts rules and parameters as needed

Autonomy Level: A4 (Autonomous) — full authority within defined scope

Governance Requirements: - Strict boundary definition — what the agent can and cannot do - Continuous quality monitoring with automated alerts - Statistical process control on outputs - Exception handling protocols with human escalation - Regular performance reviews and scope assessments - Kill switch / pause capability for emergencies

Example Applications: - Invoice processing — receive, validate, match, approve, pay (within thresholds) - Data migration and transformation — moving and transforming data between systems - Report generation — automated creation and distribution of standard reports - Email classification and response — handling routine enquiries end-to-end

Pattern Selection Guide¶

Pattern	Best For	Autonomy	Risk Tolerance
Strategy Formulation Assistant	Strategic planning support	Low	Low
Autonomous Process Monitor	Operational oversight	High	Medium
Decision Support Analyst	Complex decision preparation	Low	Low
Execution Orchestrator	Multi-step process management	Medium-High	Medium
Continuous Compliance Guardian	Regulatory adherence	Medium-High	Low
Intelligent Triage Engine	High-volume intake and routing	High	Medium
Knowledge Curator	Knowledge management	Medium	Low-Medium
Predictive Advisor	Forecasting and early warning	Low-Medium	Low
Workforce Coordinator	Blended team management	Medium-High	Medium
Autonomous Executor	High-volume routine work	Maximum	Medium-High

Combining Patterns¶

Patterns are composable. A typical enterprise process might combine:

An Intelligent Triage Engine to receive and classify incoming work
An Execution Orchestrator to manage the process flow
Multiple Autonomous Executors for routine steps
A Decision Support Analyst for judgement-required steps
A Continuous Compliance Guardian monitoring the entire flow
A Knowledge Curator maintaining the knowledge that agents and humans rely on

This composition of patterns — not any single pattern — is what characterises the agentic enterprise.

Previous: ← Domain Reference Model | Next: Governance & Trust Framework →