The Complete Guide to Using AI in the Government Industry in Billings in 2025

AI is now a state-level governance issue that Billings leaders must treat as infrastructure policy: Montana's SB 212, the “Right to Compute Act” sponsored by Billings Senator Daniel Zolnikov, was enacted to preserve residents' access to compute while requiring risk‑management plans for any “critical infrastructure” - from electric substations to wastewater plants - using AI, making city IT and public‑works teams responsible for documented controls and shutdown procedures (SB 212 (Right to Compute Act) – Montana Free Press); lawmakers framed this as part of a broader push to balance innovation and protection in state policy (Montana lawmakers look to build an AI framework).

That legal backdrop means practical, short‑course upskilling - for example, Nucamp's Nucamp AI Essentials for Work bootcamp (15 weeks) is a concrete next step for municipal staff who must implement audits, transparency practices, and vendor governance now.

“There's a new world. Let's open the door and then start restricting in a narrow, detailed way, not like other states that are basically trying to ban everything.” - Sen. Daniel Zolnikov

Montana now has a split-but-forceful AI rulebook that Billings leaders must treat as operational constraints: Senate Bill 212 (the “Right to Compute Act”) signed April 16, 2025, mandates documented risk controls and even ordered shutdown authority for AI‑controlled critical infrastructure - placing explicit responsibilities on IT and public‑works teams - and House Bill 178, signed May 5, 2025 and effective October 1, 2025, restricts government use of AI (including bans on certain profiling and public surveillance) and requires disclosure when AI‑generated material is published without human review; see the full SB 212 summary on Montana Free Press - Right to Compute Act (Montana Free Press SB 212 summary) (Montana Free Press SB 212 summary - Right to Compute Act) and the HB 178 overview on DataGuidance - Montana limits on government AI use (DataGuidance HB 178 overview) (DataGuidance overview of HB 178 - Montana limits on government AI use); on top of state action, a proposed federal “Big Beautiful Bill” moratorium could pause state enforcement for a decade, so local leaders must plan for two realities - comply now by building vendor contract clauses, published AI‑disclosure workflows, and an emergency shutdown playbook tied to operational runbooks, while tracking federal outcomes via coverage like the Daily Montanan - Big Beautiful Bill federal moratorium coverage (Daily Montanan coverage) (Daily Montanan coverage of Big Beautiful Bill federal moratorium) - a concrete “so what” is this: the city's procurement, communications, and utilities teams will need stamped, testable shutdown procedures and human‑review logs before HB 178's transparency rules take effect on October 1, 2025, or risk non‑compliance and service disruption.

“I'm scared. And I think Montanans should be scared about their data and the way they're affected by AI working in these different systems. It's scary.” - Rep. Jill Cohenour

Starting small means selecting pilots that map directly to Billings' existing workflows: begin with an AI‑assisted permit‑intake pilot inside Public Works Engineering that auto‑summarizes applications, checks required fields, and generates plain‑language, multilingual guidance so applicants know to submit complete packets at least 60 days before an event (see the City's Contractor & Permit Services for required permits and timing Billings Contractor & Permit Services - permits, fees, and 60-day filing guidance); run a parallel pilot to ingest Ecopia's high‑precision impervious‑surface vector layers into stormwater planning tools to improve fee modeling and target infrastructure inspections (Ecopia impervious surface data for Billings stormwater management); and digitize the Resource Library's forms and FAQs with AI‑powered plain‑language and multilingual outputs to reduce clerical back‑and‑forth and lower barriers for non‑English speakers (Multilingual accessibility and plain‑language translations for Billings government).

Each pilot ties to an existing data source or process, so success metrics are measurable: throughput of permit intake, time to verify stormwater impervious layers, and reduction in FAQ follow‑ups.

Billings should embed AI talent where mission accountability lives: place AI practitioners inside department Integrated Product Teams (IPTs) in Public Works, Utilities, Permits, and Communications so technical staff report to the program owners who own outcomes, not to a detached shared pool; back those IPTs with a Central AI Technical Resource that supplies secure development environments, model-monitoring tools, and procurement templates; and form an Integrated Agency Team (IAT) with legal, CISO, procurement, and budget reps to clear data rights and approvals quickly.

This hybrid model - embedded practitioners + centralized toolset + cross‑agency IAT - follows the federal AI Guide's recommendations for organizing and managing AI and prevents the common trap of centralizing talent away from mission needs (GSA AI Guide for Government: organizing and managing AI in public agencies).

Make governance concrete: require each IPT to deliver a published human‑review log and a testable shutdown playbook tied to operational runbooks so procurement, ops, and communications teams can demonstrate compliance with Montana requirements and transparency rules; pair that mandate with an actionable AI adoption checklist and short, role‑focused training for embedded staff to recruit and retain people who want mission work, not just technical perks (Actionable AI adoption checklist for Billings government teams).

Billings' AI programs must pair infrastructure capacity with rigorous data stewardship: establish a living data dictionary, enforce metadata standards and master‑data management, and bake automated data‑quality checks into every pipeline so models only consume validated, documented fields before they touch permit, utilities, or emergency datasets; the practical payoff is clear - auditable definitions and access controls turn one‑off pilots into repeatable services rather than hidden failure modes.

Operationalize this by assigning domain data stewards who run scheduled quality validations, maintain lineage and business glossaries, and coordinate with procurement on data‑sharing terms and access controls; these are the same core tasks called out in enterprise governance playbooks like Amgen's Director, Data Strategy & Governance guidance (Amgen Director Data Strategy and Governance job listing) and by data‑quality leads who define standards and tooling.

Start small: require published dictionaries and testable access rules for each pilot, and use an actionable AI adoption checklist to tie stewardship milestones to procurement and go/no‑go decisions (Billings government AI adoption checklist and procurement guide).

Billings must treat AI security, privacy, and responsibility as operational necessities - not optional features - by combining practical controls (RBAC, token‑based authentication, key/secret management) with model governance practices (versioned model registries, reproducible metadata, and continuous monitoring) so models are auditable and recoverable if they fail or are attacked; the MLOps model‑governance guidance explains these components in detail and why logging, reproducibility, and endpoint protection are non‑negotiable (MLOps model governance: security, logging, and reproducibility guidance).

Invest in local expertise and shared tooling: attend regional security‑and‑AI sessions - like INTERFACE Montana's “AI and Social Engineering,” Zero Trust, and secure M365 Copilot briefings - to build staff competency and vendor selection criteria with demonstrable security controls (INTERFACE Montana AI and Zero Trust event: AI and Social Engineering sessions).

One concrete, memorable requirement: require a testable emergency shutdown playbook tied to operational runbooks plus human‑review logs and model versioning before production rollout - this single control both preserves critical services during incidents and produces the audit trail regulators and procurement teams will demand.

When Billings evaluates buy‑versus‑build for AI, procurement must be the risk‑management engine - use the State Procurement Manual's templates (RFP, Sole‑Source Justification, Software License Agreement, and CEP/Tier II checklists) to document requirements from need‑identification through contract closeout (Montana State Procurement Manual RFPs, Forms, and Procurement Guide).

Layer federal best practices from the OMB guidance into contracts: require vendors to disclose embedded AI, notify the city of new features or model retraining, prohibit using city data to train external models unless explicitly permitted, and build post‑award monitoring and incident reporting into service levels (Federal guidance for responsible AI acquisition by local governments).

Remember the procurement loopholes: donated tools, small purchases, and feature rollouts can bypass oversight and introduce unchecked AI - so formalize review thresholds and a simple “human‑review + shutdown” clause in all agreements (TechPolicy's procurement analysis documents how non‑traditional pathways have hidden consequential AI deployments).

The practical, so‑what: require a contract clause that forbids vendors from using Billings' data to train outside models and mandates notification of AI feature changes - this single clause preserves data rights, prevents vendor lock‑in, and creates the enforceable leverage city leaders need to meet Montana's new transparency and shutdown obligations.

Billings can build a practical AI workforce pipeline now by connecting local K–12 and career programs with regional college retraining: short, hands‑on courses like the MSU Billings “Cybercat Hacking Academy” (Code Girls United, grades 7–12) and the Billings Career Center drone certification class create youth-to‑talent pathways, while statewide capacity - Missoula College's Cyber Rapid Training, Cisco Networking Academy courses (including an Introduction to Modern AI), and Microsoft's MSSA - can upskill municipal IT, emergency response, and permit staff on a 3–12 month cadence; see the roundup of Montana school and training news for local program details (Montana public schools cybersecurity camps, drone classes, and teacher apprenticeship news) and the Missoula-region education & training gateway for workforce resources (Missoula-region education and training gateway for workforce resources).

Importantly, a $4M federal grant is launching Montana's first Teacher Apprentice Program (paid four‑year apprenticeships leading to K‑12 licensure with the first cohort slated across 40 districts), which addresses the state's immediate staffing gaps and gives Billings a tested model to pay, train, and retain people in public‑facing AI roles - so what: pairing short technical certificates with paid apprenticeships creates a low‑risk, locally visible pipeline that produces staff who can run, monitor, and document AI systems under Montana's new transparency and shutdown rules (Nucamp AI Essentials for Work - multilingual prompts and municipal use cases for Billings).

Measure impact by tying each pilot to a small set of operational KPIs, instrumenting models for continuous monitoring, and hard‑wiring escalation into published runbooks: track permit‑intake throughput and percent of applications auto‑completed, measure model drift as the share of cases flagged for human review over a rolling 30‑day window (a rising drift metric should automatically trigger rollback and the emergency shutdown playbook), and log mean time to recovery (MTTR) for any model incident so regulators and auditors see traceable remediation.

Require vendors and internal teams to publish monthly performance dashboards and human‑review logs as contract deliverables, and use a tenant/KPI selection framework - like the performance‑tracking criteria used in FUSE's infrastructure projects - to prioritize pilots that create measurable civic value and clear go/no‑go gates (performance tracking & tenant‑selection KPIs).

Pair those technical guardrails with an actionable adoption checklist and short, role‑focused trainings so IPTs can interpret dashboards, run tests, and scale only when stewardship, security, and human‑review procedures are proven (actionable AI adoption checklist for Billings) - the so‑what: this approach turns pilots into auditable services that meet Montana's transparency and shutdown obligations while limiting service disruption as usage grows.

Action now: translate Montana's new rules into three concrete city tasks that are testable before HB 178's transparency requirements take effect on October 1, 2025 - (1) run at least one mission‑aligned pilot with published human‑review logs and a testable emergency shutdown playbook owned by the department IPT and Central AI Technical Resource, (2) update procurement templates to forbid vendor use of city data for external model training and to require notification of AI feature changes, and (3) raise staff capability quickly by enrolling department leads in short, practical training (for example, the 15‑week Nucamp AI Essentials for Work bootcamp) and by using regional security‑and‑AI briefings like INTERFACE Montana to vet vendor security claims; see the legal trigger and transparency rules in Montana's HB 178 overview (Montana HB 178 overview - DataGuidance analysis of government AI limits), details on the INTERFACE Montana security and AI sessions (INTERFACE Montana 2025 security + AI program details), and the Nucamp AI Essentials for Work bootcamp for role‑focused upskilling (Nucamp AI Essentials for Work bootcamp - practical AI skills for any workplace).

The single memorable test: no system goes to production until a department can demonstrate a live human‑review log, a versioned model registry entry, and a successful shutdown drill - that auditable checklist is what preserves services and keeps the city compliant.

“There's a new world. Let's open the door and then start restricting in a narrow, detailed way, not like other states that are basically trying to ban everything.” - Sen. Daniel Zolnikov