For state, county, and municipal agencies

Voice AI for government services.

Unemployment claim status, DMV scheduling, property-tax intake, benefits eligibility pre-screening, 311 non-emergency intake. English and Mexican Spanish at parity from day one. Section 508 accessible. State records-act retention configurable per state.

By Addie Conner · Last reviewed 2026-06-05.

Kill-shot — government vertical is a niche, not a flagship

Govtech sales cycles are long. The wedge is state-AG-specific compliance, not the typical 311 use case.

The government vertical is the smallest lane in Sage Voice's portfolio and we do not pretend otherwise. Govtech competitors (Granicus, Tyler Technologies, CivicPlus, Bloomberg Government adjacent) own incumbent relationships at the state and municipal level. Salesforce Public Sector and ServiceNow have larger account managers. We are not displacing any of them at the platform layer.

Where Sage Voice has a defensible niche:

  • State-AG-specific AI disclosure compliance. California CPRA, Colorado SB 24-205 AI Consumer Protection Act, plus the patchwork of state algorithmic-decision rules — these are the same gates Sage Voice already resolves for the insurance vertical, and most govtech incumbents have not retrofitted them.
  • Bilingual EN + Mexican Spanish at parity. Border-state and Hispanic-majority municipalities (TX, CA, NM, AZ, FL) need bilingual voice on day one. Granicus and Tyler can ship English-first; ES-MX parity is harder for them.
  • State records-act retention configurability. CA PRA, TX PIA, CO Open Records Act, NY FOIL retention horizons configured per tenant state. Cleaner than building the same logic into a govtech monolith.
  • NIST SP 800-53 Rev 5 path. Sage Voice supplies the SSP boilerplate and control-mapping artifacts. We are not FedRAMP-authorized today and we do not pretend to be.

Kill condition. If a major govtech incumbent (Granicus, Tyler) ships a bilingual voice-AI module with state-AG AI-disclosure coverage, the government wedge closes. Realistic outcome for the government vertical: niche revenue, not a flagship line — likely 5-15% of total ARR at exit. If it is more than that, something has gone right that we did not predict.

Scope discipline

Non-emergency intake. Not 911. Not adjudication.

In scope: non-emergency intake, scheduling, status lookups, eligibility pre-screening, 311 non-emergency calls.

Out of scope without additional review: 911 and emergency dispatch, court-of-record proceedings, law-enforcement investigative interviews, immigration adjudication, court ordered evictions, and any benefits termination decision.

Compliance posture

Federal frameworks plus state records and privacy overlays.

Privacy Act of 1974

5 USC § 552a. For federal agency tenants: SORN reference + routine-use disclosure injected into the script. Audit log captures the SORN ID applied to each call.

NIST SP 800-53 Rev 5 / FedRAMP

Tenant ATO process supported via SSP boilerplate + control-mapping artifacts aligned to the FedRAMP Moderate baseline. Sage Voice has not completed FedRAMP authorization as of 2026-06-05.

NIST SP 800-171

Controlled unclassified information handling. CUI marking propagated into transcript R2 metadata and Durable Object retention policy.

Section 508

29 USC § 794d. TTY/relay friendly. Transcripts available on request. Opt-out to human channel always one phrase away.

State records acts

CA Public Records Act (Gov. Code § 7920 et seq., 5y), TX Public Information Act (5y), CO Open Records Act (3y), NY FOIL (6y). Retention configured per tenant state.

State privacy laws

California CPRA, Colorado CPA, Virginia VCDPA, Connecticut CTDPA, Utah UCPA, Texas DPSA. Voice recording = personal information. Right-to-delete supported for non-statutorily-retained records.

Colorado SB 24-205

Colorado AI Consumer Protection Act. Algorithmic decision disclosure where applicable. Aligned to the NAIC Model Bulletin on AI Systems framework for consistency across the tenant's other regulated programs.

Sensitive PII handling

DTMF or transfer. Never voice.

Full SSN, driver's license number, passport number, and biometrics are never collected via voice. Capture happens via DTMF or via transfer to a tenant-operated secure channel. The audit log records the channel used, not the value.