Case Study 02: The COBOL Skills Crisis at Federal Insurance Corp

Background

Federal Insurance Corp (FIC) is one of the largest property and casualty insurance companies in the United States, providing homeowners, auto, and commercial insurance to over 5 million policyholders across 48 states. Founded in 1961, FIC processes approximately 2.1 million claims annually, manages $18 billion in annual premium revenue, and maintains a workforce of 12,000 employees.

Like most large insurers, FIC's core operations run on a mainframe. The company's policy administration system, claims processing engine, billing platform, and actuarial calculation programs are all written in COBOL. The system comprises 22 million lines of COBOL code distributed across approximately 6,800 programs, supported by a DB2 database with over 5 billion records and roughly 1,200 VSAM datasets.

The system is the beating heart of FIC's business. Every policy issued, every premium calculated, every claim adjudicated, and every payment disbursed flows through these COBOL programs. The system runs on two IBM z15 mainframes in a primary/disaster-recovery configuration, processing an average of 12 million transactions per day.

For sixty years, this system has been reliable, accurate, and stable. But now, FIC faces a crisis that no amount of hardware upgrades can solve: the people who understand the system are disappearing.

The Workforce Profile

FIC's mainframe technology group consists of 52 employees organized into four teams:

COBOL Development Team (24 developers)

Database Administration Team (8 DBAs)

• 3 senior DBAs (ages 55-64), specializing in DB2 and VSAM
• 3 mid-level DBAs (ages 42-51)
• 2 junior DBAs (ages 30-35)

Systems Programming Team (10 systems programmers)

• Responsible for z/OS, CICS, and infrastructure
• Average age: 56; youngest member is 44

Operations Team (10 operators)

• Manage batch scheduling, monitoring, and incident response
• Average age: 52

The total average age of FIC's mainframe workforce is 53. Within five years, 19 of the 52 team members will be eligible for retirement. Within ten years, that number rises to 34 -- nearly two-thirds of the entire team.

The Triggering Event

The crisis became impossible to ignore in March 2025, when Harold Whitfield, FIC's Chief Mainframe Architect and the most knowledgeable person in the organization about the core insurance processing system, suffered a heart attack and was forced into immediate medical retirement. Harold, 64, had been with FIC for 41 years. He had personally designed or supervised the design of the claims processing engine, the rating algorithm, and the reinsurance calculation programs.

Harold's departure was sudden and unplanned. There was no transition period, no knowledge transfer, and no designated successor. Within weeks, the team discovered the scope of what they had lost:

• Harold was the only person who fully understood the reinsurance calculation programs, a suite of 47 COBOL programs that determine how much risk FIC retains and how much it transfers to reinsurance partners. These calculations directly affect the company's financial statements.

• Harold had maintained a personal notebook (a physical, paper notebook) containing key business rules, system dependencies, and workarounds for known issues. The notebook was in his office, but much of it was written in a personal shorthand that others found difficult to interpret.

• Three active projects that Harold was leading -- a regulatory compliance update, a new product launch, and a performance optimization initiative -- were left without technical leadership.

The CIO, Sandra Williams, called an emergency meeting. "We have been talking about this risk for years," she told her leadership team. "Now it is no longer a risk. It is a reality. We need a plan, and we need it yesterday."

Assessing the Damage

Sandra commissioned a comprehensive assessment of the mainframe team's knowledge distribution. The results were alarming.

Knowledge Concentration Analysis

The assessment team mapped each of the 6,800 COBOL programs to the developers who could maintain them. The results revealed dangerous concentrations:

• 312 programs (4.6%) were understood by only one person
• 1,847 programs (27.2%) were understood by only two people
• 2,340 programs (34.4%) were understood by three or more people
• 2,301 programs (33.8%) had no clearly identified owner or maintainer

The programs in the first two categories -- understood by one or two people -- included some of the most critical systems: claims adjudication, premium rating, regulatory reporting, and financial close processing.

Documentation Audit

A review of documentation revealed:

• Only 35% of COBOL programs had up-to-date technical documentation
• Only 20% had business rule documentation that a new developer could use to understand the program's purpose
• 15% had inline comments that were considered adequate
• Many comments that did exist were outdated, referencing business rules or system configurations that had changed years ago

Skills Inventory

A skills assessment of the remaining team found:

• Only 6 developers rated themselves as "expert" in the core claims processing system
• Only 4 developers had significant CICS online programming experience
• Only 3 developers understood the actuarial calculation programs
• Only 2 people (both over 60) understood the interface programs that connect FIC's mainframe to external partners and regulatory agencies

The Recruitment Challenge

Even before Harold's departure, FIC had been struggling to hire COBOL developers. The HR department's data told a stark story:

Hiring History (Last 3 Years)

Over three years, FIC posted 18 COBOL developer positions and managed to fill only 7 of them. The primary reasons candidates cited for declining offers were:

1. Perception of the technology: Candidates feared that COBOL skills would not be transferable and that the technology was "dying"
2. Compensation: FIC's initial salary offers were below market rate for COBOL developers, reflecting the company's general pay scales rather than the premium that COBOL skills command
3. Location: FIC's headquarters is in a mid-sized Midwestern city that some candidates found less attractive than major tech hubs
4. Career growth concerns: Candidates worried about limited advancement opportunities within a specialized COBOL team

The Response: A Multi-Pronged Strategy

Sandra Williams and her team developed a comprehensive strategy to address the skills crisis, organized into five workstreams.

Workstream 1: Emergency Knowledge Preservation

Objective: Capture critical knowledge before more retirements occur.

Actions taken: - Identified the 20 most critical and least-documented program suites - Assigned each senior developer a set of programs to document using a standardized template covering business rules, data flows, system dependencies, and known issues - Contracted with a mainframe consulting firm to conduct "knowledge extraction" interviews with senior developers, producing video recordings and written summaries - Created a searchable internal wiki for mainframe documentation - Implemented mandatory code review requirements so that at least two developers are familiar with every program change

Investment: $1.2 million (consulting fees, tools, and dedicated documentation time)

Workstream 2: Compensation and Retention

Objective: Retain existing COBOL developers and make FIC competitive in the hiring market.

Actions taken: - Conducted a market salary analysis for COBOL developers, which revealed FIC was paying 18-25% below market rates - Implemented immediate salary adjustments for all mainframe developers, bringing compensation to the 75th percentile of the market - Created a "Mainframe Specialist" career track with clear promotion criteria and salary bands comparable to the company's digital technology track - Introduced retention bonuses for senior developers who commit to remaining for at least three more years - Offered phased retirement options that allow senior developers to transition to part-time mentoring roles rather than leaving abruptly - Added remote work options (3 days remote, 2 days on-site) to improve work-life balance

Investment: $3.8 million annually in increased compensation and bonuses

Workstream 3: Training and Development Pipeline

Objective: Build a sustainable pipeline of new COBOL developers.

Actions taken:

University Partnerships: - Partnered with three regional universities to introduce COBOL and mainframe computing into their computer science curricula - Funded a "Mainframe Center of Excellence" at one university, equipped with a Hercules-based z/OS emulation environment and IBM's z/OS development tools - Sponsored a "Mainframe Challenge" competition for CS students, with prizes including paid internships and scholarship funds - Created a 12-week summer internship program that pairs university students with senior FIC developers

Internal Training Program -- "COBOL Academy": - Designed a 16-week intensive COBOL training program for existing FIC employees with programming experience in other languages (Java, Python, C#) - Curriculum covers COBOL fundamentals, JCL, CICS, DB2, and FIC's specific systems - Participants receive full salary during training and a completion bonus - Senior developers serve as mentors, spending 4 hours per week coaching trainees - First cohort: 8 trainees from FIC's distributed development team; 6 completed the program and transitioned to the mainframe team

External Bootcamp Partnerships: - Partnered with a workforce development organization to sponsor COBOL training bootcamps in the local community, targeting career changers and veterans - Offers guaranteed interviews to bootcamp graduates who meet qualification standards

Investment: $2.5 million in the first year (university funding, training program development, instructor costs, intern salaries)

Workstream 4: Modernization to Reduce Complexity

Objective: Reduce the maintenance burden by modernizing the most problematic parts of the system.

Actions taken: - Implemented IBM Wazi Analyze to perform automated analysis of the COBOL codebase, identifying dead code, unused variables, and circular dependencies - Removed 1.8 million lines of dead code (programs and code paths that were no longer executed), reducing the codebase from 22 million to 20.2 million lines - Consolidated 340 redundant copybooks (shared data definitions) into 95 standardized versions - Implemented automated regression testing for the 200 most critical programs, reducing the manual testing burden on the development team - Began exposing key COBOL programs as APIs, allowing new functionality to be built in modern languages while leveraging existing COBOL business logic

Investment: $4.2 million (tools, consulting, and developer time)

Workstream 5: Organizational and Cultural Change

Objective: Change how the organization perceives and values mainframe work.

Actions taken: - Renamed the "Legacy Systems" team to "Core Insurance Platform" -- a seemingly small change that had a significant psychological impact - Included mainframe developers in agile ceremonies, architecture reviews, and innovation discussions alongside the distributed development team - Created cross-functional teams where mainframe and cloud developers work together on integration projects - Established a "Mainframe Innovation Lab" where developers can experiment with new z/OS features, cloud integration patterns, and AI-assisted code analysis tools - The CIO began regularly highlighting mainframe achievements in company-wide communications, counteracting the perception that mainframe work is unglamorous

Investment: $200,000 (team events, communication materials, lab infrastructure)

The Business Case

Sandra Williams presented the total investment to FIC's executive committee as follows:

Total 3-Year Investment

Cost of Inaction

Sandra also presented the estimated cost of not addressing the crisis:

• Unplanned system outage due to inability to maintain critical programs: A single day of downtime in the claims processing system would cost an estimated $15-25 million in delayed claims payments, regulatory penalties, and reputational damage
• Failed regulatory audit: If FIC cannot demonstrate adequate control over its core systems, regulators could impose restrictions on the company's operations
• Key person departures: If two or three more senior developers leave without knowledge transfer, FIC estimates it would take 18-24 months and $5-8 million in consulting fees to rebuild the lost expertise -- if it could be rebuilt at all
• Competitive disadvantage: Without the ability to modify and enhance its core systems, FIC would lose the ability to launch new insurance products, adjust pricing, and respond to market changes

The executive committee approved the investment unanimously.

Results After Two Years

Two years into the program, FIC has seen measurable improvements:

Staffing

• Net headcount in the mainframe team has increased from 52 to 58
• 12 new developers have been hired or transitioned from other teams
• Only 4 retirements have occurred (versus 8 that were projected), due to retention incentives and phased retirement options
• The COBOL Academy has graduated 14 trainees across three cohorts, with a 75% completion rate
• 6 university interns have been converted to full-time employees

Knowledge

• Documentation coverage of critical programs has increased from 35% to 71%
• The number of programs understood by only one person has decreased from 312 to 89
• The knowledge wiki contains over 2,000 articles and is actively used by the team

System Health

• The codebase has been reduced from 22 million to 19.8 million lines through dead code removal
• Automated test coverage for critical programs has reached 68%
• 42 COBOL programs have been exposed as APIs, enabling the digital team to build three new customer-facing features without modifying mainframe code
• Average time to deploy a change has decreased from 3 weeks to 5 days

Culture

• Employee satisfaction scores for the mainframe team have increased by 23 points
• Internal transfer requests into the mainframe team have increased by 400% (from 2 per year to 10 per year)
• The team has presented at two industry conferences about their modernization journey

Remaining Challenges

Despite the progress, significant challenges remain:

1. The reinsurance programs: Harold Whitfield's specialty remains the most fragile area. Two senior developers have made progress in understanding these programs, but they estimate it will take another 12-18 months before the knowledge gap is adequately closed.

2. Compiler version migration: FIC is running an older version of IBM's Enterprise COBOL compiler. Upgrading to the latest version would unlock performance improvements and new language features but requires regression testing of thousands of programs.

3. The next generation: While the COBOL Academy has been successful, the question remains whether the pipeline can sustain itself. Will enough new developers choose COBOL careers to replace the continuing wave of retirements over the next decade?

4. Vendor dependency: FIC's heavy reliance on IBM for hardware, software, and tools creates concentration risk. The company is exploring Micro Focus COBOL and GnuCOBOL as supplementary platforms for development and testing.

Discussion Questions

1. Prevention vs. Cure: FIC's crisis was years in the making. What early warning signs should management have recognized? At what point should they have begun investing in knowledge preservation and pipeline development?

2. The COBOL Academy Model: Evaluate FIC's approach of retraining existing Java/Python developers as COBOL programmers. What are the advantages and disadvantages of this approach compared to hiring experienced COBOL developers from the market? What makes a good candidate for retraining?

3. Compensation Strategy: FIC raised salaries by 18-25% and introduced retention bonuses. Is compensation the primary factor in retaining COBOL developers, or are other factors (meaningful work, career growth, culture) equally important? How would you structure a total compensation package to maximize retention?

4. The Dead Code Problem: FIC discovered 1.8 million lines of dead code (approximately 8% of the total). How does dead code accumulate over decades? What risks does it create? What processes should be in place to prevent dead code accumulation in the future?

5. University Partnerships: FIC invested in university partnerships to create a COBOL talent pipeline. What challenges do you anticipate in convincing computer science students to study a language that many perceive as outdated? How would you pitch COBOL to a college sophomore?

6. Naming Matters: FIC renamed their "Legacy Systems" team to "Core Insurance Platform." Why is this more than just a cosmetic change? How do labels and terminology affect developer recruitment, retention, and morale?

7. Business Case Construction: Sandra Williams had to justify a $30.7 million investment to the executive committee. If you were a member of that committee, what questions would you ask? What additional data would you want to see before approving the investment?

8. The Harold Whitfield Scenario: Harold's sudden departure created an immediate crisis. Design a "bus factor" mitigation plan that ensures no single developer's departure can create a similar crisis. What specific policies, practices, and tools would you implement?

9. Cross-Training Economics: FIC's COBOL Academy takes 16 weeks and costs approximately $180,000 per trainee (salary, instruction, lost productivity). A newly trained developer is productive at roughly 40% of a senior developer's level after completing the program. Is this a good investment? At what point does the investment break even?

10. Industry Implications: FIC's situation is not unique -- it is representative of thousands of organizations worldwide. What role should industry organizations, governments, and educational institutions play in addressing the COBOL skills gap at a systemic level? Should there be public policy incentives for COBOL workforce development?

Key Lessons

• The COBOL skills crisis is not a future problem -- it is a present reality that requires immediate action.
• Knowledge preservation must be treated as a strategic priority, not an administrative task. The cost of capturing knowledge proactively is a fraction of the cost of reconstructing it after key people leave.
• Competitive compensation is necessary but not sufficient. Developers also need career paths, modern tools, meaningful work, and a culture that values their contributions.
• Retraining existing developers from other language backgrounds is a viable and cost-effective approach to building COBOL capacity, but it requires sustained investment in curriculum, mentoring, and support.
• Reducing system complexity through dead code removal, automated testing, and API enablement makes the system easier to maintain and reduces the knowledge burden on the team.
• Organizational culture and perception matter. How an organization talks about its mainframe team directly affects its ability to recruit and retain talent.
• The business case for investing in COBOL workforce development is compelling when the cost of inaction -- system failures, regulatory penalties, competitive disadvantage -- is quantified honestly.