The AI-Enabled Organization

Using AI, Data Analytics and Continuous Improvement to Drive Organizational Excellence and prepare for a SPQA Evaluation process.

Pension and investment organizations face growing expectations for accuracy, fiduciary stewardship, responsive member service, operational efficiency and transparency. The scale and public importance of the sector make this especially consequential: the 22 largest pension markets held an estimated $55.7 trillion in aggregate assets, equal to 69% of the GDP of those economies.^[1] At the same time, aging populations, inflation, inequality and the continuing shift in many markets from defined-benefit to defined-contribution arrangements are increasing the pressure on pension systems to operate efficiently while protecting retirement security.^[1]

Artificial intelligence (AI), data analytics and continuous improvement can help leadership meet these expectations—but only when they are applied systematically and tied to measurable organizational outcomes. Within an organizational excellence framework such as the Senate Productivity and Quality Award (SPQA), AI should be viewed not as a stand-alone technology initiative, but as an enabler of leadership, strategy, workforce effectiveness, operational excellence and results.

Leadership: Start with Mission, Governance and Measurable Value

Leadership must first define why AI is being adopted. The objective should be to improve mission outcomes: administer benefits accurately and timely, safeguard assets, strengthen investment operations, improve service, increase employee capacity and support sound decision-making. Every AI initiative should identify the business problems, expected member or organizational benefit, measurable performance target, risks and required human oversight.

Real-world context: Pension-industry research from CFA Institute identifies practical AI opportunities across the value chain, including member onboarding and communication, pension reporting, investment analysis, actuarial risk analysis, governance support and predictive analytics.^[1] This supports a portfolio approach: begin with specific, measurable use cases rather than a broad, undefined “AI transformation.”

Responsible governance is essential. Pension organizations manage sensitive personal, financial and investment information. Leadership should establish enterprise standards for data privacy, cybersecurity, acceptable AI use, output validation, records management and accountability. NIST’s voluntary AI Risk Management Framework (AI RMF) was created to help organizations incorporate trustworthiness considerations into the design, development, use and evaluation of AI systems; NIST also issued a Generative AI Profile in July 2024 to address risks particular to generative AI.^[2]

For pension organizations, a practical governance model should require that every AI use case has:

• A named business owner, data owner, risk/compliance reviewer and technology owner.
• A documented purpose, approved data sources and explicit prohibited uses.
• Pre-deployment testing for accuracy, privacy, security, bias and operational resilience.
• Human review thresholds for benefit, investment, fiduciary, employment or member-impacting decisions.
• Monitoring, incident escalation, periodic revalidation and a documented ability to suspend the system.

AI can assist professionals, but consequential benefit, investment and fiduciary decisions should remain subject to authorized human judgment.

Improve Pension Administration Through Process Intelligence

AI and analytics can transform pension administration from transaction processing to exception-based management. Across the retirement lifecycle—application, documentation, eligibility, service verification, calculation, review, approval and payroll—analytics can identify bottlenecks, recurring errors and unusual transactions. Instead of knowing only that retirement processing averages 40 days, leadership can determine whether delays arise from incomplete applications, employer data discrepancies, service-credit issues, manual calculations or approval queues.

Illustrative operating model: A plan may combine workflow timestamps, case codes and document-completeness data to classify applications by risk and complexity. Straightforward cases continue through standard controls; incomplete or unusual cases are routed to an experienced examiner with the precise exception identified. The AI does not calculate or alter the benefit on its own—it prioritizes the reviewer’s work and gives the reviewer an auditable explanation of why the record was flagged.

AI can also flag unusual calculations or records for review without independently changing a member’s benefit. This approach can improve processing time, accuracy, backlog management and quality while preserving appropriate controls. It is aligned with a broader industry view that AI can improve pension efficiency and accuracy when its application is targeted to the particular needs of each fund.^[1]

Measures to establish before a pilot:

• Median and 90th-percentile end-to-end processing time.
• First-pass quality and post-payment correction rate.
• Backlog aging by case type and complexity.
• Percentage of AI-flagged cases confirmed as valid exceptions.
• Reviewer override rate and reasons for override.
• Member inquiries or complaints attributable to the process.

Strengthening Investment Operations and Reporting

Investment teams receive large volumes of information from managers, custodians, consultants and market data providers. AI-supported analytics can reconcile market values, returns, benchmarks, fees and exposures across sources and direct staff attention to material exceptions. AI can also compare current and prior manager reports to identify significant changes in performance, strategy, personnel, portfolio positioning and risk.

A particularly valuable use is investment-report quality assurance. Before materials reach executives, investment committees or boards, AI can compare narrative statements with underlying tables and source data, flagging inconsistent values, mathematical errors, unexplained benchmark changes, stale commentary or contradictory conclusions. Professionals remain responsible for interpretation and approval, while AI performs high-volume comparison and validation.

This use case is consistent with industry research indicating that AI and machine-learning tools can expand portfolio managers’ analytical capacity, support actuarial analysis of pension risk and help keep market-trend assessments current. CFA Institute also notes possible value in analyzing private-market and sustainable-investment data, where information can be extensive and unstructured.^[1]

Guardrails for investment use:

• Preserve source data, prompts, model versions, outputs and approvals for records and audit purposes.
• Treat AI summaries as drafts and require analyst verification of all performance, benchmark, exposure and attribution figures.
• Separate factual extraction and reconciliation from investment recommendation and fiduciary decision-making.
• Apply heightened review to external, unstructured or confidential data and to any model output used in board materials.

Increase Employee Productivity and Knowledge Access

The strongest workforce case for AI is augmentation rather than simple automation. Employees spend significant time searching policies, reviewing documents, comparing spreadsheets, drafting correspondence and preparing reports. A governed enterprise AI assistant can help staff search approved organizational knowledge, summarize information, compare documents, draft routine content and identify exceptions—with citations to authoritative sources where appropriate.

The productivity benefit should be reinvested in higher-value work: member counseling, analytical review, problem solving, investment due diligence, process improvement and strategic initiatives. Leadership should measure not only hours saved, but how released capacity improves organizational outcomes.

The U.S. Department of Labor’s 2024 AI best-practices roadmap emphasized several workforce principles relevant to pension and investment organizations: establish governance and review processes, maintain meaningful human oversight for significant employment decisions, communicate transparently with workers, involve workers in AI use, provide training and secure worker data.^[3] Although the Department notes that some older material may not reflect later policy changes, these remain useful operating principles for a workforce-centered implementation.^[3]

Practical workforce controls:

• Deploy a secure, approved assistant that is grounded only in authorized internal knowledge for sensitive work.
• Train employees to validate outputs, recognize hallucinations and protect confidential information.
• Publish clear rules on permitted data, prohibited uploads, retention and escalation of inaccurate or harmful output.
• Measure adoption, employee confidence, rework, time-to-information and quality—not merely prompt volume.

Create a Better Member Experience

AI can improve service without eliminating human interaction. Virtual assistants can provide timely access to approved general information, while AI tools can help member-service staff quickly locate policies and procedures during interactions. Analytics across calls, emails, CRM cases, surveys and complaints can identify recurring member concerns and their underlying causes.

CFA Institute identifies member onboarding, communications, reporting and retirement planning as areas where AI could strengthen engagement, financial literacy and support throughout the retirement lifecycle.^[1] However, member-facing tools should be deliberately limited to approved informational assistance unless the organization has separately validated a higher-risk use case and established the appropriate controls.

This enables a powerful continuous improvement cycle:

Member Feedback → Data → Analysis → Root Cause → Process Improvement → Measurement

Instead of repeatedly handling the same problem, leadership can use evidence to redesign the process that created it.

Example: If contact-center analysis shows repeated questions about a retirement estimate, the response should not be limited to improving the chatbot’s answer. The organization should examine the estimate letter, portal language, calculator assumptions, employee training and handoffs. The lasting result is fewer avoidable contacts and clearer member communications.

Move from Reporting to Predictive Management

Traditional dashboards explain what has already happened; AI-enabled analytics can help leadership anticipate what may happen next. Potential applications include forecasting retirement volumes, service demand and workload; identifying emerging project or vendor risks; detecting data-quality issues; and highlighting operational exceptions. The goal is not algorithmic management—it is earlier, better information for leadership decisions.

For example, a workforce and operations forecast could combine historical retirement application volumes, eligibility cohorts, seasonal patterns, staffing levels and document-completeness rates. The resulting forecast should be used to plan capacity and outreach, then evaluated against actual results. It should not be treated as a self-executing decision engine.

For predictive tools, leaders should define forecast accuracy targets, monitor performance drift and establish an escalation point when a model no longer produces reliable results. This approach fits NIST’s emphasis on incorporating trustworthiness into the entire AI lifecycle, including design, deployment, evaluation and use.^[2]

Measure Organizational Results

AI success should never be measured by the number of tools deployed. SPQA-oriented leadership should establish baselines and demonstrate sustained improvement in measures that matter.

A sound scorecard contains both outcome measures and control measures. For example, a member-service assistant may be evaluated through response time, first-contact resolution and satisfaction, but also through escalation rates, answer accuracy, privacy incidents and the rate at which staff correct its answers. A faster system that creates inaccurate benefit guidance is not an organizational improvement.

A Systematic Leadership Model

The most effective approach connects organizational-excellence disciplines into one management system:

Leadership → Strategy → Members & Stakeholders → Data & Knowledge → Workforce → Operations → Results

Leadership sets the direction; strategy prioritizes opportunities; member and stakeholder needs define value; trusted data provides evidence; employees use AI-enabled capabilities; operations are redesigned through continuous improvement; and performance measures demonstrate results.

The objective is not simply to become an AI-enabled organization. It is to become a high-performing pension and investment organization that systematically uses AI, trusted data, human expertise and continuous improvement to deliver better outcomes for members, employees, fiduciaries and stakeholders. That is where AI and the principles of SPQA organizational excellence can create sustainable and measurable value.

Sources

1. CFA Institute Research and Policy Center, “Pensions in the Age of Artificial Intelligence.” The report presents pension-sector use cases and notes seven representative case studies based on interviews with pension experts and industry professionals.
2. National Institute of Standards and Technology, “AI Risk Management Framework.” NIST describes the AI RMF as a voluntary framework for incorporating trustworthiness considerations throughout AI systems’ design, development, use and evaluation; it released its Generative AI Profile on July 26, 2024.
3. S. Department of Labor, “Department of Labor releases AI Best Practices roadmap for developers, employers,” October 16, 2024. The agency page includes a notice that some releases from before January 20, 2025, may be out of date or not reflect current policies.

Srikumar Bala

SPQA Board Member