ops-automation-opportunity-finder - SKILL.md Agent Skill

name: ops-automation-opportunity-finder description: Identify and evaluate automation opportunities in banking operations using structured assessment frameworks. Use when analyzing processes for RPA, intelligent automation, AI/ML, or straight-through processing potential across payments, lending, account servicing, compliance, and back-office functions.

Ops Automation Opportunity Finder

Overview

This skill produces structured automation opportunity assessments for banking operations. It evaluates processes against automation readiness criteria including volume, standardization, rule-based logic, error rates, and ROI potential. It covers RPA (Robotic Process Automation), intelligent document processing (IDP), AI/ML-driven decisioning, straight-through processing (STP), and workflow automation. Output supports business cases, technology roadmaps, and operational transformation programs.

When to Use

Conducting automation opportunity assessments across banking operations
Evaluating specific processes for RPA, AI, or intelligent automation suitability
Building business cases for automation investments with ROI projections
Prioritizing an automation pipeline based on value, feasibility, and risk
Assessing automation readiness (data quality, process maturity, system landscape)
Supporting digital transformation and operations modernization programs
Identifying quick wins vs. strategic automation investments

Required Inputs

Input	Description	Format
Process inventory	List of operational processes with descriptions	Process catalog
Volume data	Transaction/task volumes by process	Operations metrics
Effort data	FTE effort, time per task, manual steps	Time study/workforce data
Error data	Error rates, rework rates, exception frequencies	Quality metrics
System landscape	Applications used, integration capabilities, APIs	IT architecture
Cost data	Labor costs, error costs, processing costs	Finance data
Compliance constraints	Regulatory requirements affecting automation	Compliance mapping

Methodology

Step 1: Catalog Candidate Processes

Build a comprehensive process inventory across operations:

Domain	Process	Volume/Month	FTEs	Manual Steps	Systems	Error Rate
Payments	Wire initiation and release	[N]	[N]	[N]	[List]	[X%]
Payments	ACH return processing	[N]	[N]	[N]	[List]	[X%]
Payments	Check exception handling	[N]	[N]	[N]	[List]	[X%]
Lending	Loan document review	[N]	[N]	[N]	[List]	[X%]
Lending	Condition clearing	[N]	[N]	[N]	[List]	[X%]
Account services	Account opening data entry	[N]	[N]	[N]	[List]	[X%]
Account services	Address change processing	[N]	[N]	[N]	[List]	[X%]
Compliance	SAR narrative preparation	[N]	[N]	[N]	[List]	[X%]
Compliance	KYC document verification	[N]	[N]	[N]	[List]	[X%]
Reconciliation	GL reconciliation	[N]	[N]	[N]	[List]	[X%]
Reconciliation	Nostro/Vostro reconciliation	[N]	[N]	[N]	[List]	[X%]

Step 2: Assess Automation Suitability

Score each process against automation readiness criteria:

Criterion	Weight	Score 1 (Low)	Score 3 (Medium)	Score 5 (High)
Volume	20%	<100/month	100-1,000/month	>1,000/month
Standardization	20%	Highly variable, many exceptions	Mostly standard, some exceptions	Highly standardized, few exceptions
Rule-based logic	20%	Requires significant judgment	Mix of rules and judgment	Clearly defined business rules
Digital inputs	15%	Paper-based, unstructured	Mix of digital and paper	Fully digital, structured data
System stability	10%	Frequent changes, unstable	Occasional changes	Stable, well-documented
Error impact	15%	Low impact errors	Moderate financial/customer impact	High financial/regulatory impact

Automation suitability score = Σ (Weight × Score)

Score Range	Suitability	Recommended Approach
4.0-5.0	High — Immediate candidate	RPA or STP; fast implementation
3.0-3.9	Medium — Good candidate with preparation	Intelligent automation; process redesign first
2.0-2.9	Low-Medium — Requires significant investment	AI/ML for unstructured; phased approach
1.0-1.9	Low — Not ready for automation	Process maturation needed before automation

Step 3: Select the Right Automation Technology

Match process characteristics to automation technology:

Technology	Best For	Characteristics	Typical ROI Timeline
RPA	High-volume, rule-based, multi-system data entry	Structured data, defined steps, stable UI	6-12 months
Intelligent Document Processing (IDP)	Document-heavy processes (loans, KYC, correspondence)	Unstructured/semi-structured documents	9-18 months
Workflow automation	Multi-step processes with approvals and routing	Sequential/parallel tasks, rule-based routing	3-9 months
AI/ML decisioning	Pattern recognition, prediction, classification	Historical data, probabilistic outcomes	12-24 months
Straight-through processing (STP)	End-to-end elimination of manual intervention	API integration, event-driven architecture	12-24 months
Chatbot/Virtual assistant	Customer and employee inquiry resolution	FAQ, guided workflows, NLP	6-12 months
Process mining	Process discovery, conformance checking, optimization	Event logs, process execution data	3-6 months

Step 4: Calculate ROI and Business Case

For each automation candidate, quantify the business case:

Cost savings calculation:

Component	Current State	Automated State	Savings
Labor (FTE equivalent)	[N FTEs × $X]	[N FTEs × $X]	[$X/yr]
Error/rework costs	[$X/yr]	[$X/yr]	[$X/yr]
Processing time	[X hrs/item]	[X hrs/item]	[X hrs saved]
Overtime/temp staff	[$X/yr]	[$X/yr]	[$X/yr]
Total annual savings			[$X/yr]

Investment required:

Component	Cost
Software licensing	[$X/yr]
Implementation (partner/internal)	[$X one-time]
Integration development	[$X one-time]
Change management/training	[$X one-time]
Ongoing maintenance	[$X/yr]
Total first-year cost	[$X]
Total ongoing annual cost	[$X/yr]

ROI metrics:

Net annual benefit: [Annual savings - ongoing cost]
Payback period: [Total investment / net annual benefit] months
3-year NPV: [Calculated at institution's hurdle rate]
IRR: [Internal rate of return]
FTE capacity freed: [N FTEs redeployed to higher-value activities]

Step 5: Assess Risk and Compliance Considerations

Evaluate automation risks specific to financial services:

Risk Category	Considerations	Mitigation
Regulatory	Does the process have regulatory requirements for human review?	Identify required human-in-the-loop checkpoints
Model risk	Does AI/ML automation create SR 11-7 model risk obligations?	Assess model risk classification, validation requirements
Operational	What happens when the automation fails?	Design fallback procedures, monitoring, alerting
Data privacy	Does the automation process PII or restricted data?	Apply data handling controls, encryption, access limits
Audit trail	Can automated decisions be explained and audited?	Ensure logging, explainability, record retention
Change management	How will staff and processes adapt?	Training, role redesign, communication plan
Vendor risk	Does the automation depend on third-party platforms?	Vendor due diligence, contractual protections, exit strategy

Step 6: Prioritize the Automation Pipeline

Rank opportunities using a value-feasibility matrix:

Process	Value Score	Feasibility Score	Combined	Priority
[Process 1]	[1-5]	[1-5]	[Average]	[Rank]
[Process 2]	[1-5]	[1-5]	[Average]	[Rank]

Value score factors: Annual savings, error reduction, customer experience improvement, strategic alignment Feasibility score factors: Technical complexity, process maturity, data availability, organizational readiness, regulatory constraints

Pipeline categorization:

Quick wins (High feasibility, moderate value): Implement in 0-6 months
Strategic bets (High value, moderate feasibility): Plan and implement in 6-18 months
Low-hanging fruit (Moderate both): Implement as capacity allows
Long-term vision (High value, low feasibility): Requires process maturation first

Step 7: Design the Implementation Roadmap

Structure the automation program in waves:

Wave 1 — Foundation (0-6 months):

Quick wins with proven RPA technology
Process documentation and standardization
Center of Excellence (CoE) establishment
Governance framework and change management

Wave 2 — Expansion (6-18 months):

Intelligent automation (IDP, workflow)
Cross-functional process automation
Analytics and process mining integration
Scaling infrastructure and monitoring

Wave 3 — Transformation (18-36 months):

AI/ML-driven decisioning and prediction
End-to-end STP for target processes
Customer-facing automation (onboarding, servicing)
Continuous improvement and optimization

Output Specification

# Automation Opportunity Assessment: [Scope]

## Executive Summary
[Key findings: number of opportunities, total savings potential, recommended priorities]

## Process Inventory
[Catalog of evaluated processes with volumes, effort, and error rates]

## Automation Suitability Scores
| Process | Volume | Standardization | Rule-Based | Digital Input | System Stability | Error Impact | Total | Suitability |
|---------|--------|-----------------|------------|---------------|------------------|-------------|-------|-------------|
| [Process] | [1-5] | [1-5] | [1-5] | [1-5] | [1-5] | [1-5] | [X.X] | [High/Med/Low] |

## Top Opportunities
### [Opportunity 1]
- **Process**: [Name]
- **Technology**: [RPA/IDP/AI/STP]
- **Annual Savings**: [$X]
- **Investment**: [$X]
- **Payback**: [X months]
- **FTEs Freed**: [N]
- **Risk Level**: [Low/Medium/High]

## Prioritized Pipeline
[Value-feasibility matrix with categorization]

## Implementation Roadmap
[Three-wave implementation plan with milestones]

## Risk Assessment
[Regulatory, operational, and technology risks with mitigations]

## Recommendations
[Top 3-5 recommendations with supporting rationale]

Analysis Framework

Automation Maturity Assessment

Evaluate the institution's automation maturity:

Level 1 — Ad hoc: Individual macros and scripts, no governance
Level 2 — Opportunistic: Piloting RPA, initial CoE formation
Level 3 — Systematic: Established CoE, pipeline management, scaling RPA
Level 4 — Intelligent: Integrated intelligent automation, AI/ML in production
Level 5 — Autonomous: Self-optimizing processes, minimal human intervention

Process Mining Application

Before automating, mine the actual process execution:

Discover the true process (not the documented process) from system event logs
Identify process variants, rework loops, and bottlenecks
Quantify the proportion of straight-through vs. exception processing
Use conformance checking to identify deviations from standard process
Prioritize automation of the dominant process variant (80% path)

Human-in-the-Loop Design

For regulated processes requiring human oversight:

Define which steps can be fully automated vs. human-reviewed
Design exception routing for items outside automation confidence thresholds
Implement sampling-based quality assurance of automated decisions
Ensure explainability for AI/ML-driven decisions
Maintain regulatory audit trail with clear attribution (human vs. automated)

Examples

Example 1 — RPA Quick Win: "Account maintenance address change process: 2,400 requests/month, currently requiring manual data entry across 3 systems (core banking, CRM, card system) taking an average of 8 minutes per request. 3.2 FTEs dedicated to this task. Error rate: 4.5% (wrong field, incomplete update). Automation suitability score: 4.6/5.0 (high volume, highly standardized, rule-based, digital input from online banking). Recommended technology: RPA bot with structured data extraction from the online banking request form. Expected results: 95% straight-through processing (2,280 automated/month), 0.1% error rate, 2.8 FTE capacity freed. Investment: $85K implementation + $24K/year licensing. Annual savings: $196K labor + $18K error remediation = $214K. Payback: 5.2 months."

Example 2 — Intelligent Automation: "Loan document review and condition clearing: 800 loans/month, average 12 documents per loan, 45 minutes per loan for initial review. 8 FTEs. Error rate: 6% (missed conditions, incorrect classification). Automation suitability: 3.2/5.0 (moderate — semi-structured documents, some judgment required). Recommended technology: Intelligent Document Processing (IDP) with ML-based document classification and data extraction, combined with rules-based condition matching. Human-in-the-loop for low-confidence extractions (<85% confidence score). Expected results: 60% of documents auto-classified and extracted, reducing average review time to 18 minutes. 3.6 FTE capacity freed. Investment: $350K implementation + $120K/year platform. Annual savings: $295K labor + $42K error reduction = $337K. Payback: 14 months. Regulatory note: final loan approval decision remains with human underwriter per SR 11-7 model risk requirements."

Guidelines

Automate the process as-is only if it's well-designed; redesign before automating when the process is fundamentally flawed
Start with high-volume, rule-based processes for initial automation to build confidence and capability
Always design for exceptions; no process is 100% automatable, and exception handling must be planned
Quantify both hard savings (FTE, error reduction) and soft benefits (speed, consistency, scalability)
Regulatory requirements may mandate human oversight for certain decisions; identify these constraints early
RPA is not a substitute for system integration; use APIs and STP for long-term architecture
Monitor bot performance continuously; automation can fail silently and accumulate errors
Consider the impact on staff (redeployment, upskilling, morale) in the business case
AI/ML automations may trigger SR 11-7 model risk management requirements
Maintain a centralized automation inventory with ownership, monitoring, and lifecycle management

Validation Checklist

Process inventory is comprehensive across all banking operations domains
Automation suitability scoring uses consistent, weighted criteria
Technology recommendation matches process characteristics
ROI calculation includes all cost components (labor, error, investment, ongoing)
Payback period and NPV are calculated at the institution's hurdle rate
Regulatory and compliance constraints are identified for each opportunity
Human-in-the-loop requirements are designed for regulated processes
Pipeline is prioritized using value-feasibility framework
Implementation roadmap is phased with realistic timelines
Risk assessment covers regulatory, operational, vendor, and change management dimensions