evidoria

← Back to browse

Good practice Imported

SSA's AI Phone Bot and Claims-Support Tools — Automating Social Security's 800 Number, With Documented Accuracy Gaps

United States of America · Woodlawn · See the United States of America profile

Evidence: Observational / pre–post Top 24% 73/100 · Ask Evidence Copilot about this practice

SSA's AI phone bot now handles a large share of the 5 million monthly calls to its national helpline and cut average wait times from 30 to 7 minutes, but independent testing found it can misstate benefit rules and mishandle vulnerable callers.

~5.1 million/month
Monthly calls to SSA helpline (Sept 2025)
~1.6 million/month
Calls resolved via automated self-service (Sept 2025)
~30 minutes
Average speed of answer, before (Jan 2025)
~7 minutes
Average speed of answer, after (Sept 2025)
49 consecutive weeks
Disability-claims backlog decline streak (as of May 2025)
25
AARP test calls placed

Details

Maturity
Scaling
Promoter
U.S. Social Security Administration
Period
2025–2026
Keywords
social security, benefits administration, contact centers, disability claims, public sector AI

Context

The U.S. Social Security Administration (SSA) operates a national 1-800-772-1213 helpline serving roughly 5.1 million calls a month. In January 2025 average speed of answer was about 30 minutes.

Objectives

To automate resolution of routine helpline requests (e.g. benefit-verification letters, address changes) via a voice bot, reduce wait times, and help caseworkers process disability-claim medical evidence, while retaining a human-agent option for complex cases.

Activities

In April 2025 SSA launched an AI-powered «Phone Bot» on its national helpline that greets callers with «How can I help you today?» and resolves routine requests through automated self-service, letting callers say «agent» to reach a human. Separately, AI tools help caseworkers organise medical evidence submitted with disability claims, some running over 1,000 pages. SSA situates the deployment within a formal Enterprise Artificial Intelligence Strategy, a compliance plan for OMB Memorandum M-25-21, and a public AI-use inventory.

Results

By September 2025, roughly 1.6 million of the approximately 5.1 million monthly calls were resolved through automated self-service, and SSA's average speed of answer improved from about 30 minutes (January 2025) to about 7 minutes (September 2025). SSA's disability-claims backlog fell for 49 consecutive weeks as of May 2025. However, AARP's 25 test calls found the bot struggled to correctly explain spousal and survivor benefits and sometimes confused Social Security retirement benefits with Supplemental Security Income (SSI). Community Legal Services of Philadelphia documented a case where a client asking about a disability-overpayment reconsideration was instead given unrelated information about federal railroad retirement and domestic-partnership rules.

Conclusions

The accuracy problems surfaced through outside testing (AARP) and congressional oversight (Senator Elizabeth Warren's office) rather than through SSA's own public error-rate reporting; legal-aid advocates warn that callers with cognitive difficulties or mental-health conditions may be poorly served by an automated channel.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Short (< 1 year)
Staffing & skills
SSA operates the phone bot on its existing national 800-number infrastructure, Disability caseworkers use companion AI tools to organise medical evidence, Governance sits under SSA's Enterprise Artificial Intelligence Strategy team

Conditions for success

  • Built-in human-agent escape hatch ('say agent') preserved in the call flow
  • Formal governance and compliance framework under OMB Memorandum M-25-21
  • Deployment on existing, already-scaled national helpline infrastructure

Common failure modes

  • Misstates benefit rules for complex cases such as spousal and survivor benefits
  • Confuses Social Security retirement benefits with Supplemental Security Income (SSI)
  • Legal-aid groups documented cases of the bot misdirecting vulnerable callers on active disability matters
  • Accuracy problems surfaced via outside testing and congressional pressure, not SSA's own reporting

Where it fits

Governance type
national government agency
Scale
national (~5 million calls/month)
Income level
high income (USA)

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful