agentic-ai

name	agentic-ai
description	Agentic AI and intelligent automation knowledge for consulting engagements. Use when working on process automation, multi-agent systems, RPA/IPA implementation, workflow automation, or custom AI solutions. Covers automation assessment, agent architecture, and implementation approaches.

Agentic AI & Intelligent Automation Skill

Overview

Provides specialized knowledge for automation and agentic AI engagements including opportunity assessment, agent design patterns, platform selection, and implementation approaches.

---

Automation Assessment Framework

Process Scoring Criteria

Criterion	Weight	Score (1-5)	Description
Volume	20%	-	Transaction frequency
Standardization	20%	-	Rule-based vs. judgment
Data Structure	15%	-	Structured vs. unstructured
Stability	15%	-	Process change frequency
Error Rate	15%	-	Current quality issues
Strategic Value	15%	-	Business criticality

Automation Suitability Matrix

                    High Volume
                        │
    ┌───────────────────┼───────────────────┐
    │   AUTOMATE NOW    │   OPTIMIZE FIRST  │
    │   (Quick Wins)    │   (Standardize)   │
High├───────────────────┼───────────────────┤
Rule│                   │                   │
Base│   EVALUATE ROI    │   LOW PRIORITY    │
    │   (Case-by-case)  │   (Manual OK)     │
    └───────────────────┼───────────────────┘
                    Low Volume

ROI Calculation Template

Annual Benefit = (Hours Saved × Hourly Cost) + Error Reduction Value + Speed Value

Implementation Cost = License + Development + Training + Change Management

ROI = (Annual Benefit - Annual Cost) / Implementation Cost × 100

Payback Period = Implementation Cost / Annual Benefit

---

Agent Design Patterns

Single Agent Architecture

User Request
    │
    ▼
┌─────────────────────────────────────┐
│           AGENT                      │
│  ┌─────────────────────────────┐    │
│  │      Reasoning Engine        │    │
│  │   (LLM + System Prompt)      │    │
│  └─────────────────────────────┘    │
│              │                       │
│  ┌───────────┴───────────┐          │
│  ▼           ▼           ▼          │
│ Tool 1    Tool 2      Tool 3        │
│ (API)    (Database)  (Search)       │
└─────────────────────────────────────┘
    │
    ▼
Response

Multi-Agent Patterns

Supervisor Pattern:

                 ┌──────────────┐
                 │  Supervisor  │
                 │    Agent     │
                 └──────┬───────┘
           ┌───────────┼───────────┐
           ▼           ▼           ▼
      ┌────────┐  ┌────────┐  ┌────────┐
      │ Worker │  │ Worker │  │ Worker │
      │ Agent 1│  │ Agent 2│  │ Agent 3│
      └────────┘  └────────┘  └────────┘

Pipeline Pattern:

Input → Agent 1 → Agent 2 → Agent 3 → Output
        (Extract)  (Transform) (Validate)

Collaborative Pattern:

┌────────────────────────────────────┐
│         Shared Context             │
└────────────────────────────────────┘
     ▲              ▲              ▲
     │              │              │
┌────────┐    ┌────────┐    ┌────────┐
│ Agent 1│◄──►│ Agent 2│◄──►│ Agent 3│
└────────┘    └────────┘    └────────┘

Agent Component Checklist

• Clear goal definition
• Appropriate tool selection
• Memory strategy (short-term, long-term)
• Guardrails and constraints
• Error handling approach
• Human escalation triggers
• Logging and observability
• Testing strategy

---

Automation Technology Stack

By Automation Type

Type	Technologies	Best For
Task Automation	UiPath, Automation Anywhere, Power Automate	Desktop/web tasks
Workflow Automation	Camunda, Temporal, n8n	Process orchestration
Document Processing	ABBYY, Kofax, Azure Form Recognizer	Unstructured data
Integration	MuleSoft, Workato, Zapier	System connectivity
AI/ML	OpenAI, Anthropic, Hugging Face	Intelligent automation

Agent Framework Comparison

Framework	Strengths	Best For
LangChain	Ecosystem, flexibility	General purpose
Claude Agent SDK	Anthropic integration, simplicity	Claude-based agents
AutoGen	Multi-agent, Microsoft	Complex workflows
CrewAI	Role-based, simple	Team simulations
Semantic Kernel	Microsoft stack	Enterprise .NET

---

Implementation Approach

Phase 1: Discovery (2-4 weeks)

• Process inventory and documentation
• Stakeholder interviews
• Volume and complexity analysis
• Technology assessment
• Quick win identification

Phase 2: Design (3-6 weeks)

• Solution architecture
• Integration design
• Exception handling
• Security requirements
• Success metrics

Phase 3: Build (4-12 weeks)

• Development sprints
• Unit testing
• Integration testing
• UAT preparation
• Documentation

Phase 4: Deploy (2-4 weeks)

• Pilot deployment
• User training
• Hypercare support
• Performance monitoring
• Optimization

Phase 5: Scale (Ongoing)

• Additional process automation
• Capability expansion
• COE maturation
• Continuous improvement

---

Common Use Cases

High-Value Automation Candidates

Category	Examples
Finance	Invoice processing, reconciliation, reporting
HR	Onboarding, benefits administration, reporting
IT	Ticket routing, user provisioning, monitoring
Operations	Order processing, inventory updates, shipping
Customer Service	Query routing, FAQs, status updates
Compliance	Data validation, reporting, audit prep

Agent Use Cases

Category	Examples
Research	Information gathering, summarization
Analysis	Data analysis, insight generation
Content	Document creation, communication drafts
Support	Customer inquiries, internal helpdesk
Coding	Code generation, review, debugging
Operations	Monitoring, alerting, remediation

---

Risk Mitigation

Common Automation Risks

Risk	Mitigation
Scope creep	Fixed scope per phase, change control
Integration failures	Thorough API testing, fallback procedures
Data quality	Input validation, exception handling
User adoption	Change management, training, champions
Maintenance burden	Documentation, COE ownership

Agent Safety Controls

Control	Implementation
Output filtering	Content moderation, format validation
Action limits	Scope restrictions, approval gates
Rate limiting	Request throttling, cost controls
Audit logging	Complete action history
Human escalation	Confidence thresholds, sensitive triggers

---

Success Metrics

Automation Metrics

Metric	Target
Process time reduction	>70%
Error rate reduction	>90%
Cost per transaction	>50% reduction
Employee satisfaction	Increase
Automation coverage	% of eligible processes

Agent Metrics

Metric	Target
Task completion rate	>95%
Accuracy	>98%
Response time	<X seconds
Escalation rate	<5%
User satisfaction	>4.0/5.0

---

References

See references/ folder for:

• Automation platform comparison
• Agent architecture examples
• ROI calculation templates
• Implementation checklists