USA-003

Social Security Administration (SSA) – Insight Decision-Quality Support System

Download PDF
United States North America High income Operational Deployment (Limited Rollout) Confirmed

U.S. Social Security Administration (SSA)

At a Glance

What it does Perception and extraction from unstructured inputs — Operational and process automation
Who runs it U.S. Social Security Administration (SSA)
Programme SSA disability adjudication quality-review process (Insight subsystem)
Confidence Confirmed
Deployment Status Operational Deployment (Limited Rollout)
Key Risks Model-related risks
Key Outcomes Available sources state that internal SSA studies associated Insight with improved work quality, improved remediation of quality issues during drafting, improved recognition of quality issues on appeal, and more efficient case processing.
Source Quality 3 sources — Working paper / technical note, Report (multilateral / development partner), Legal document / regulation

The U.S. Social Security Administration (SSA) has developed and deployed the Insight software suite to support decision-quality review in disability adjudication at the hearings and appeals levels. Insight is the best-documented SSA AI subsystem in the retained source base and is therefore treated here as a narrower case than the agency's broader disability-AI portfolio.

Insight was devised by SSA attorney Kurt Glaze and developed within SSA's Office of Appellate Operations (OAO). The tool applies natural language processing to written hearing decisions, extracts information about findings and rationale, and combines that with structured case information from workload systems. Using this combined picture, Insight applies rule-based and probabilistic machine-learning methods to identify potential quality issues in adjudicative decisions across roughly 30 issue areas. In other words, it is designed to read draft or completed decisions as text and help surface patterns or omissions that matter for internal quality review.

The system is explicitly assistive rather than determinative. It does not decide eligibility, order outcomes, or prescribe remedies. Instead, it flags possible quality issues for adjudicators and reviewers, who remain responsible for evaluating the case record and making any resulting determination. Insight was fully deployed to adjudicative staff at the appeals level by late 2017 and at the hearings level by late 2018. That deployment history matters because it shows the system moved beyond a small experiment and into routine use inside a major federal benefits-adjudication environment.

The retained sources associate Insight with improvements in work quality, remediation of quality issues during drafting, recognition of quality issues on appeal, and more efficient case processing. However, those performance statements come from internal studies described in secondary technical and policy sources rather than from a full public operational evaluation released by SSA itself. The case therefore rests on strong documentation of the tool's existence, purpose, and organisational adoption, but only more limited public evidence on its measurable downstream effects.

The broader SSA disability-adjudication environment is relevant context. The agency handles millions of disability claims and faces persistent backlog, staffing, and evidence-processing challenges. Decisions are legally consequential and often depend on large volumes of structured and unstructured evidence. Those pressures help explain why assistive AI tools such as Insight emerged. But other SSA tools, including IMAGEN and QDD, are no longer bundled into this record because they have different purposes, maturity levels, and evidence depth.

Insight operates within a human-in-the-loop oversight framework. Final benefit decisions remain with human adjudicators, and the main documented risks concern automation bias, transparency, explainability, and the possibility that a quality-support tool could still shape outcomes in a rights-impacting domain if staff over-rely on it. Even though Insight is framed as quality assurance rather than direct adjudication, a tool that systematically influences how reviewers identify deficiencies can still affect claimant experience, remand patterns, and the consistency of disability decision-making across the agency.

That is why the decision criticality for the case remains high. The software does not itself award or deny benefits, but it operates close to the core of a rights-affecting adjudication process. Public documentation is also limited relative to the significance of that setting: external observers still lack full visibility into evaluation design, production monitoring, subgroup effects, and contestability mechanisms specific to the tool. Insight is therefore best understood as a mature and real SSA assistive-AI deployment, but one embedded in a domain where even support tools require careful scrutiny.

Classifications follow the DCI AI Hub Taxonomy. Hover over field labels for definitions.

Social Protection Functions

Implementation/delivery chain
Assessment of needs/conditions + enrolment primaryAccountability mechanisms Case management
SP Pillar (Primary) The social protection branch: social assistance, social insurance, or labour market programmes. Social insurance
Programme Name SSA disability adjudication quality-review process (Insight subsystem)
Programme Type The type of social protection programme, classified under social assistance, social insurance, or labour market programmes. View in glossary Old age, survivors and disability pensions
System Level Where in the social protection system the AI is applied: policy level, programme design, or implementation/delivery chain. View in glossary Implementation/delivery chain
Automation Subtype For operational automation cases: (a) document processing and generative staff assistance, or (b) workload and resource forecasting. (a) Document processing and generative staff assistance
Programme Description Insight is an SSA decision-quality support system used within the disability adjudication workflow at the hearings and appeals levels. It assists review of SSDI and SSI disability decisions by flagging possible quality issues in written adjudicative decisions.
Implementation Type How the AI output is produced: Classical ML, Deep learning, Foundation model, or Hybrid. Affects validation, compute requirements, and governance profile. View in glossary Hybrid
Lifecycle Stage Current stage in the AI lifecycle, from problem identification through to monitoring, maintenance and decommissioning. View in glossary Integration and Deployment
Model Provenance Origin of the AI model: developed in-house, adapted from open-source, commercial/proprietary, or accessed via third-party API. View in glossary Developed in-house
Compute Environment Where the AI system runs: on-premise, government cloud, commercial cloud, or edge/device. View in glossary Not documented
Sovereignty Quadrant Classification of data and compute sovereignty: I (Sovereign), II (Federated/Hybrid), III (Cloud with safeguards), or IV (Shared Innovation Zone). View in glossary Not assessed
Hybrid Components Insight uses NLP extraction combined with rule-based and probabilistic machine-learning methods for quality checking of adjudicative decisions.
Data Residency Where the data used by the AI system is stored: domestic, regional, or international. View in glossary Not documented
Cross-Border Transfer Whether data crosses national borders, and if so, whether documented safeguards are in place. View in glossary Not documented
Decision Criticality The rights impact of the decision the AI supports. High criticality requires HITL oversight; moderate requires HOTL; low may operate HOOTL. View in glossary High
Human Oversight Type Level of human involvement: Human-in-the-Loop (active review), Human-on-the-Loop (monitoring), or Human-out-of-the-Loop (periodic audit). View in glossary HITL
Development Process Whether the AI system was developed fully in-house, through a mix of in-house and third-party, or fully by an external provider. View in glossary Mix of in-house and third-party
Highest Risk Category The most significant structural risk source identified: data, model, operational, governance, or market/sovereignty risks. View in glossary Model-related risks
Risk Assessment Status Whether a formal risk assessment, informal assessment, or independent audit has been conducted for this system. Formal assessment
  • Bias audit
  • Human oversight protocol
  • Independent evaluation
CategorySensitivityCross-System LinkageAvailabilityKey Constraints
Administrative data from other sectorsSensitiveLinks data across multiple systemsCurrently available and usedMedical evidence from healthcare providers collected via Health Information Technology processes; written hearing decisions and ALJ rationales
Beneficiary registries and MISSpecial categoryLinks data across multiple systemsCurrently available and usedDisability claims records including structured administrative data from eCMS and ACAT systems; claimant biographical and claim history data
Unstructured and text-based contentSpecial categorySingle source (no linkage)Currently available and usedUnstructured medical documentation; records can exceed 1,000 pages per claimant with ~80% unstructured; high redundancy (estimated half from prior visits); SSA spends ~$500 million/year collecting medical evidence

Glaze, K., Ho, D.E., Ray, G.K. and Tsang, C. (2024). Artificial Intelligence for Adjudication: The Social Security Administration and AI Governance. Stanford, CA: Stanford Digital Government Hub. Available at: https://dho.stanford.edu/wp-content/uploads/SSA.pdf (Accessed: 31 October 2025).

View source Working paper / technical note

National Academy of Social Insurance (2025). Phase One Report: Task Force on Artificial Intelligence, Emerging Technology, and Disability Benefits. Washington, DC: NASI. Available at: https://www.nasi.org/wp-content/uploads/2025/04/Phase-One-Report-Task-Force-on-Artificial-Intelligence-Emerging-Technology-and-Disability-Benefits.pdf (Accessed: 31 October 2025).

View source Report (multilateral / development partner)

The White House (2023). Executive Order 14110 -- Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. Washington, DC: Executive Office of the President. Available at: https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/ (Accessed: 31 October 2025).

View source Legal document / regulation
Deployment Status How far the system has progressed into real-world operational use, from concept/exploration through to scaled and institutionalised. View in glossary Operational Deployment (Limited Rollout)
Year Initiated The year the AI system was first initiated or development began. 2015
Scale / Coverage The scale and geographic or population coverage of the deployment. Insight was deployed to SSA hearings and appeals staff, with full deployment at the appeals level by late 2017 and at the hearings level by late 2018.
Funding Source The source(s) of funding for the AI system development and deployment. U.S. federal government appropriations (SSA administrative budget)
Technical Partners External technology vendors, academic partners, or development partners involved. Insight was developed within SSA's Office of Appellate Operations with blended legal and technical expertise; SSA later invested in software-development staff and contractors to scale it to an enterprise system.
Outcomes / Results Available sources state that internal SSA studies associated Insight with improved work quality, improved remediation of quality issues during drafting, improved recognition of quality issues on appeal, and more efficient case processing. No fully public SSA evaluation with detailed external performance reporting was identified in the retained source pack.
Challenges Public documentation of Insight remains limited relative to its importance in a rights-impacting adjudication environment. Key concerns include automation bias, limited external visibility into evaluation methods, and the broader difficulty of assessing bias in SSA systems because the agency has not collected race and ethnicity data for decades.

How to Cite

DCI AI Hub (2026). 'Social Security Administration (SSA) – Insight Decision-Quality Support System', AI Hub AI Tracker, case USA-003. Digital Convergence Initiative. Available at: https://socialprotectionai.org/use-case/USA-003 [Accessed: 1 April 2026].

Change History

Created 30 Mar 2026, 08:41
by v2-import (import)