Nine real-world templates for disclosing AI authority across public services. Each includes the decision-grade level, system description, AI and human roles, contestability paths, and oversight mechanisms.
Assistive
Medicaid Recertification
Summary
Auto-renews coverage when no changes detected. Flags complex cases for worker review. 60% auto-renewed, 40% reviewed by a human. Coverage continues during review.
Service
Medicaid Expansion Eligibility Recertification
Agency
[State] Medicaid Program
What This System Does
Automated recertification every 6 months (federal requirement).
AI Role
Checks databases for income, employment, and address changes
Auto-renews if no changes and still eligible
Flags data mismatches for worker review
Human Role
Workers review all flagged cases
Workers determine what additional information is needed
Workers make the final determination
For Members
If auto-renewed: Confirmation notice, no action needed
If flagged: Letter explaining what is needed, 30 days to respond
Contestability
Call your worker: [phone]
Submit reconsideration if terminated
90-day appeal window
Oversight
Monthly monitoring of auto-renewal and termination rates
Quarterly review for unnecessary documentation requests
Bias testing for disproportionate flagging
Advisory
Unemployment Fraud Detection
Summary
Generates risk scores (0–100). An investigator must find independent evidence before any action. The score alone cannot deny benefits. Claimants are notified before review begins.
Service
Unemployment Insurance Fraud Detection
Agency
[State] Employment Security Department
What This System Does
Flags claims for potential fraud investigation.
AI Role (Advisory Only)
Generates risk score based on application patterns and data inconsistencies
Score alone cannot deny benefits or trigger penalties
Human Role
All high-risk scores reviewed by a fraud investigator
Investigator must document independent evidence
Action requires investigator findings, never the score alone
If Flagged
Written notice within 3 business days
Right to provide documentation before determination
Determination must cite investigator findings
Oversight
Independent audit every 6 months
False positive rate published quarterly (target <5%)
System disabled if false positive rate exceeds 10%
Safeguards from Michigan MiDAS Failure
Michigan's MiDAS automated fraud detection produced a 93% false positive rate, wrongly accusing tens of thousands of claimants. These safeguards directly address that failure:
The system flags; it does not decide. There are no automated denials.
Every flagged case gets a trained investigator. Human review is mandatory.
Claimants know before any adverse action. Notification precedes investigation.
Advisory
SNAP Documentation Check
Summary
Scans case files for missing documents. Worker decides if additional info is needed. The system does not determine eligibility, calculate benefits, or auto-deny applications.
Service
SNAP Error Rate Reduction
Agency
[State] Department of Human Services
What This System Does
Helps workers identify missing documentation before federal audits.
AI Role
Scans for missing income verification, address docs, and signatures
Generates a checklist for worker review
Human Role
Worker reviews alerts and decides if additional docs are needed
Worker can mark alerts as inapplicable
Worker makes the final eligibility determination
For Applicants
What may change: You may receive requests for additional documentation
What stays the same: Eligibility criteria, benefit amounts, and timelines
Why This Exists
Federal law (HR1) penalizes states for insufficient documentation. This helps workers complete case files to protect state SNAP funding.
Limitations
Does not calculate benefit amounts
Does not determine eligibility
Does not auto-deny applications
Does not share data with immigration enforcement
Advisory + Limited Determinative
Child Welfare Hotline Screening
Summary
Generates risk scores (0–20). Scores of 18–20 trigger mandatory investigation. Below 18, the screener decides. The score supplements clinical judgment for 95% of calls.
Service
Child Welfare Hotline Screening
Agency
[County] Department of Human Services
What This System Does
Generates risk assessment scores for incoming hotline calls.
AI Role
Scores 0–20 based on CPS history, child age, allegation type, and environmental factors
Scores 18–20: Automatic face-to-face investigation within 24 hours (state law)
Scores below 18: Score is one input among many; the screener decides
Human Role
Screeners review all calls
Screeners can upgrade response regardless of score
Supervisory review required to screen out calls scored above 15
Contestability
Request supervisor review during the call
Administrative review within 5 business days
Score and notes available through records request
Oversight
Weekly screener calibration sessions
Quarterly bias audit (race, zip code, reporter type)
Annual recalibration
Community advisory board reviews aggregate data
Advisory
911 Call Prioritization
Summary
Suggests priority level based on call keywords and location. Dispatcher makes the final determination and can override any suggestion. All life-threatening calls get immediate response regardless of system output.
Service
Emergency Call Dispatch
Agency
[City] 911 Communications
What This System Does
Assists dispatchers in prioritizing emergency response.
AI Role
Analyzes call keywords, location, and historical patterns
Suggests priority level (1–5, with 1 being highest)
Flags potential life-threatening situations
Human Role
Dispatcher makes all final determinations
Can override any suggestion with no documentation requirement
Life-threatening calls get immediate response regardless of system input
For Callers
The system helps speed up response. A trained dispatcher always decides what resources to send.
Oversight
Weekly review of override rates
Monthly audit of response times by priority level
Quarterly review of cases where system suggestion differed significantly from dispatcher decision
Limitations
Cannot assess caller tone or urgency beyond keywords
May miss regional dialects or uncommon emergency descriptions
Dispatcher training emphasizes professional judgment over system output
Assistive
Business License Application
Summary
Auto-approves when all requirements are verified. Flags incomplete or complex applications for reviewer. All edge cases get human review.
Service
Business License Application Processing
Agency
[City] Business Licensing Department
What This System Does
Processes routine business license applications.
AI Role
Verifies business address, zoning compliance, and fee payment
Auto-approves when all requirements are met and there are no complications
Flags applications with zoning questions, special permit needs, or incomplete info
Human Role
Licensing specialist reviews all flagged applications
Specialist determines what additional documentation is needed
Specialist makes the final determination on complex cases
For Applicants
Auto-approved: License issued within 24 hours
Flagged: Email within 2 business days explaining next steps
Contestability
Request manual review: [email/phone]
Appeal denial: [process link]
All denials include specific reasons and required corrections
Oversight
Sample 5% of auto-approvals monthly for accuracy
Track approval/denial rates by business type and neighborhood
Quarterly bias review
Advisory
Special Education Screening
Summary
Flags students who may need evaluation. The school team makes all decisions about assessment and services. The system identifies patterns; it does not diagnose or determine eligibility.
Service
Special Education Needs Identification
Agency
[District] Special Education Services
What This System Does
Helps identify students who may benefit from special education evaluation.
AI Role
Analyzes attendance patterns, grades, and behavioral referrals
Flags students showing patterns associated with learning challenges
Human Role
Teachers or parents can request evaluation regardless of system flag
School team (teachers, specialists, parents) reviews all flagged cases
Licensed professionals conduct assessments and determine eligibility
For Families
The system helps ensure students are identified early. Evaluation decisions involve you and the school team. Parent-requested evaluations proceed regardless of the system.
Oversight
Monthly review of flagged students by special education coordinator
Annual bias audit by demographic group
Parent feedback collected on referral process
Limitations
This is a screening tool only
Flags indicate patterns worth investigating, not eligibility
Cannot override parent-requested evaluations
Advisory
Building Inspection Prioritization
Summary
Ranks properties for inspection based on risk factors. Inspector decides the final schedule and findings. The system helps prioritize limited resources. It does not determine pass/fail or penalties.
Service
Building Safety Inspection Scheduling
Agency
[City] Building and Safety Department
What This System Does
Helps prioritize which properties get inspected when.
AI Role
Ranks properties based on building age, prior violations, complaint history, and neighborhood risk factors
Suggests inspection priority order
Human Role
Inspector can move urgent cases up in the queue
Inspector conducts all inspections and makes findings
Inspector determines violations and required corrections
For Property Owners
You can request an inspection regardless of system priority
System ranking does not affect inspection standards or outcomes
All violations and penalties are determined by a licensed inspector
Oversight
Monthly review: are high-priority properties actually higher risk?
Quarterly demographic analysis to prevent biased targeting
Annual review of inspection outcomes vs. system predictions
Transparency
Ranking factors publicly documented
Property owners can request their property's risk score
System logic published publicly
Assistive
Housing Assistance Eligibility
Summary
Verifies income and household size. Auto-approves clear cases. Housing specialist reviews complex situations. Automatic approval only when income is verified and household composition is clear. All edge cases get human review.
Service
Section 8 Housing Choice Voucher Eligibility
Agency
[City] Housing Authority
What This System Does
Determines initial eligibility for housing assistance.
AI Role
Verifies income through database checks (employment, tax records)
Confirms household size and composition
Auto-approves when: income clearly below threshold, household verified, no complications
Flags: income near threshold, household composition questions, special circumstances
Human Role
Housing specialist reviews all flagged applications
Specialist assesses special circumstances (medical expenses, childcare costs)
Specialist makes the final determination on complex cases
For Applicants
If auto-approved: Notification within 5 days, move to waitlist
If flagged: Specialist contacts you within 10 days for additional info
Contestability
Request manual review: [phone/email]
Provide additional documentation at any time
Appeal denial through formal process
Free legal assistance available at [organization]
Oversight
Sample auto-approvals for accuracy
Monitor denial rates by demographic group
Track how often manual review overturns auto-decisions
