How AI Is Helping Government Companies in Fresno Cut Costs and Improve Efficiency

Fresno's move toward AI-driven government efficiency is rooted in local capacity-building and a new state regulatory environment: Fresno State's Artificial Intelligence Initiative commits to embedding AI across curriculum, research, workforce development and industry partnerships - creating a local talent pipeline that city and county agencies can partner with to pilot automation and reduce dependence on outside consultants (Fresno State Artificial Intelligence Initiative); at the same time California's Judicial Council is setting statewide guardrails for generative AI, requiring written policies by Dec.

15, 2025, which means Fresno agencies must pair innovation with clear accountability (California Judicial Council generative AI rules and requirements); finally, planners should weigh infrastructure limits - CalMatters reports AI's steep energy demand is already pushing policy debates about new power sources - so workforce training that focuses on practical, efficient AI use (for example the 15‑week AI Essentials for Work syllabus) becomes a strategic priority for fiscally responsible adoption (AI Essentials for Work syllabus - Nucamp 15-week professional bootcamp).

Bootcamp	Length	Early bird cost
AI Essentials for Work	15 Weeks	$3,582

California's September 2023 Executive Order N‑12‑23 set a statewide playbook that directly shapes how Fresno agencies adopt generative AI: agencies must inventory high‑risk GenAI uses, designate senior managers for ongoing oversight, and route any GenAI contracts through the California Department of Technology for review, while CDT must provide controlled pilot environments (“sandboxes”) so governments can test tools before full procurement - steps designed to surface risks and document benefits before large-scale rollout (California Executive Order N‑12‑23 GenAI statewide playbook).

The March 2024 purchasing guidance built on that order, requiring continuous monitoring, pre‑use risk assessments, and staff training; for Fresno this means appointing an AI compliance lead, using state sandboxes to validate workflows, and applying equitable‑impact criteria when choosing vendors to avoid deployments that could harm vulnerable residents (California March 2024 AI purchasing guidelines for government).

Those procedural requirements convert experimentation into accountable pilots that agencies must report and revise as governance evolves - so Fresno's next step is operational: match local pilots to the state's inventory and training timelines to stay compliant and evidence‑driven.

“This is a potentially transformative technology – comparable to the advent of the internet – and we're only scratching the surface of understanding what GenAI is capable of,” Newsom said.

Caltrans' GenAI pilots are a practical blueprint Fresno agencies can borrow: the Traffic Mobility Insights work is testing models that ingest data streams, images and video from thousands of sensors across California's 52,000+ lane miles to predict bottlenecks, detect incidents faster, and guide targeted safety upgrades (Caltrans GenAI Traffic Mobility Insights project details).

State contracts using Azure OpenAI and other Generative AI tools aim to turn historical and real‑time data into actionable operational changes - predicting where congestion will form and prioritizing interventions for vulnerable roadway users - so Fresno transit, public works and regional operators can pilot similar analytics in controlled sandboxes and focus limited capital on the highest‑impact locations rather than broad, costly rebuilds (Governor Newsom's GenAI highway congestion initiative announcement).

The bottom line: applying these models locally helps triage high‑risk intersections and speed incident response, which reduces cascading delays and improves reliability without immediate large infrastructure spend.

“Using GenAI through smart, responsible implementation will be a game changer in developing solutions to ease traffic gridlock and reduce deaths and serious injuries on our roadways.” - Caltrans Director Tony Tavares

CDTFA's real‑world GenAI call‑center pilot offers Fresno a low‑risk template for faster, more accurate customer service: over a 10‑month test the department used a vendor tool to search more than 16,000 pages of internal guidance and generate draft replies for agents - shortening average handle time across a center that fields hundreds of thousands of contacts annually (2022–23: ~660,000 calls, 17,000 emails, 127,000 chats) and relieves peak‑season pressure that previously required reassigning as many as 280 staffers; agents always review AI suggestions before sending, which preserves human judgment while speeding responses and lowering abandon rates.

Key operational lessons for Fresno: run pilots in a state sandbox, require agent review and monitoring for factual accuracy, compare multiple vendors in side‑by‑side tests, and measure time‑saved as the primary ROI so saved hours can be redeployed to higher‑value work.

Read CDTFA's announcement on the pilot and the state's Call Center Productivity project for procurement and governance details.

Pilot metric	Value
Pilot duration	10 months
Reference materials searchable	16,000+ pages
2022–23 contact volume	~660,000 calls; 17,000 emails; 127,000 chats
Peak reassigned staff (pre‑AI)	~280

“Integrating GenAI into our operations complements the efforts of our teams. Helping agents find the right answer is just one advantage of this new technology.” - Trista Gonzalez, CDTFA Director

Administrative automation offers Fresno a practical path to cut permit backlogs and speed grant processing by using state-backed tools and procurement pathways: California is providing an Archistar e-check that uses computer vision, machine learning, and automated rulesets to instantly check designs against zoning and building codes and allow property owners to pre-check plans before submission - this e-check is being offered free to local governments via a state‑philanthropy partnership and is available statewide through a contract any local government can access (California Archistar e-check building permit tool); concurrently, California's GenAI agreements and pilot toolkit (including the State Digital Assistance AI tool and Microsoft 365 Copilot pilots) plus the RFI2/sandbox procurement approach give Fresno a controlled way to test automations that summarize documents, draft form letters, and speed staff workflows without compromising oversight (California GenAI deployments and sandbox procurement overview).

For grant and housing repair programs, published program caps - loans up to $40,000, grants up to $10,000, combined assistance up to $50,000 - create clear rule sets that automated checks and eligibility screens can enforce, shortening manual review time and freeing staff for complex cases (USDA Single Family Housing Repair Loans & Grants (California) program details).

The practical payoff: tools already in the state's toolbox aim to turn multi‑week reviews into hours or days while preserving human review and accountability.

Program	Maximum assistance
504 Home Repair Loan	Loan up to $40,000
504 Home Repair Grant	Grant up to $10,000
Combined assistance	Up to $50,000

“Bringing AI into permitting will allow us to rebuild faster and safer, reducing costs and turning a process that can take weeks and months into one that can happen in hours or days.” - Steadfast LA Chairman Rick Caruso

As Fresno scales AI into government workflows, protecting frontline workers means coupling pilots with enforceable labor safeguards, clear oversight, and funded retraining so automation augments jobs instead of displacing them; Sacramento's Assembly Budget Subcommittee stressed the need for published POC lessons and a still‑outstanding review of “impacts on state workforce,” a transparency step Fresno should require before moving pilots to production (Assembly Budget Subcommittee GenAI hearing and workforce oversight transcript).

Local unions are already a practical partner: recent Unit 1 bargaining wins guaranteed no demotions or wage loss while creating Analyst III/IV career paths - an example Fresno can mirror by writing reclassification, redeployment, and apprenticeship commitments into contracts and by routing saved hours into defined upskilling cohorts promoted through community events like Fresno's “Preparing for Tomorrow” AI workshop (SEIU Local 1000 Fresno updates and bargaining actions, Preparing for Tomorrow: Fresno AI workforce workshop event details).

The practical payoff: when pilots mandate human review, transparency, and funded training, time‑savings from automation reliably convert into higher‑value public service rather than layoffs - one specific metric to track is the percentage of saved agent hours reallocated to trained roles within 12 months.

Safeguard	Concrete element
Union guarantees	No demotion or wage loss; promotion pathways (Unit 1 reclass agreement)
Legislative oversight	Publish POC outcomes and workforce impact review (Assembly hearing request)
Training pipeline	Local workshops and funded upskilling cohorts tied to redeployment targets

“The team is grateful for the many Unit 1 members who provided valuable feedback... We could not have achieved the important concessions that were won in this agreement without this kind of teamwork.” - Kevin Healy

Begin with a narrow, measurable pilot: designate an AI compliance lead, choose one high‑value workflow (call‑center replies, permit prechecks, or incident triage), and run a controlled test in the state “sandbox” while enforcing local safeguards such as Fresno Unified's Acceptable Use Policy (never input student or sensitive IDs into public models) to protect data and meet legal requirements (Fresno Unified AI Guidance for AI use in schools).

Use Fresno State's AI initiative and WITH‑Cyber partnership to recruit trained students and faculty for pilot evaluation, cybersecurity validation, and rapid workforce upskilling so saved hours become redeployment opportunities rather than layoffs (Fresno State Artificial Intelligence Initiative overview, Fresno State WITH‑Cyber pilot details and funding).

Operational rules: require human review of all AI outputs, run side‑by‑side vendor comparisons, capture primary ROI as staff hours saved, and record the percentage of those hours reallocated to trained roles within 12 months; use those metrics to decide whether to scale, pause, or reprocure under California's sandbox and procurement guidance.

Program	Detail
WITH‑Cyber total funding	$4,000,000
Fresno State allocation	$1,000,000
Funding period	Through June 2026

“Our graduate and undergraduate students and faculty are critical for tackling the constantly changing and real world cybersecurity demands.” - Dr. Joy Goto

Fresno's AI pilots bring measurable gains but also concrete risks that must be managed: accuracy failures and “hallucinations” in generative outputs can taint police reports or permit decisions, civil‑liberties groups warn that AI‑drafted reports could warp memories or obscure accountability, and new California rules now impose pre‑use notices, opt‑out rights, annual cybersecurity audits and detailed risk assessments that agencies must factor into procurement and timelines (Fresno Bee report on AI in police reports, Fisher Phillips summary of California AI privacy regulations).

Practical mitigations already used in state pilots include limiting GenAI to low‑sensitivity cases, forcing officer or agent review of every draft, retaining first drafts for audit, and never submitting student or protected identifiers to public models per local policy - steps reflected in Fresno Unified's Acceptable Use guidance that prohibit sharing PII with external AI systems (Fresno Unified AI and privacy Acceptable Use Policy).

The so‑what: a single unchecked AI error in a police report or eligibility decision can trigger legal challenges, erode trust, and negate time savings unless Fresno couples pilots with mandatory human oversight, documented risk assessments, and timely CPPA‑compliant notices.

Risk	Mitigation (documented in pilots/regulation)
Accuracy / hallucinations	Require human review of all outputs; retain first drafts for comparison
Bias & civil‑liberties	Limit use to low‑sensitivity cases; conduct equity impact reviews; preserve audit trails
Privacy & compliance	Follow CPPA pre‑use notices, opt‑outs, risk assessments and cybersecurity audit schedules

“The computer is trained on data generated by humans, so it will reflect our biases and, in some cases, magnify the biases.” - Paul D. Knothe

Concrete pilots elsewhere point to realistic, near‑term savings Fresno can target: employee carpool programs used by Fresno organizations and similar employers cut parking demand and commuting costs while improving retention - one municipal program reduced parking requirements by 22% and employees can save $1,000+ per year through shared commutes (Shyft Fresno carpool program case study: employee savings and parking reduction); AI‑driven predictive maintenance has repeatedly cut unplanned downtime by up to 50% and trimmed maintenance costs 10–40%, a direct operational win for Fresno's public works and transit fleets that preserves service hours and avoids costly emergency repairs (ProValet predictive maintenance case studies: downtime and cost reductions); and logistics teams using AI‑enabled TMS report fuel savings of 7–11% and an 18% reduction in delivery windows - figures Fresno's city fleet and contracted carriers can convert into lower fuel budgets and faster program delivery (AI-enabled TMS and freight optimization: fuel savings and delivery window improvements).

Together these examples show how modest pilots (carpool, predictive maintenance, route optimization) can free budgeted staff hours and reduce recurring costs within 12 months, with a single program often paying back implementation effort through avoided capital or overtime costs.

Pilot type	Observed savings / outcome
Employee carpool	Parking ↓22%; employee savings $1,000+ / year
Predictive maintenance	Unplanned downtime ↓ up to 50%; maintenance costs ↓ 10–40%
AI route/TMS	Fuel savings 7–11%; delivery windows ↓18%

Move from planning to action: designate an AI compliance lead, run one measurable pilot in the state “sandbox” under California's GenAI playbook, and require human review, side‑by‑side vendor comparison, and a metric for success such as the percentage of staff hours saved and reallocated to trained roles within 12 months; recruit students and faculty through Fresno State's Artificial Intelligence Initiative to staff evaluations and cybersecurity checks, and close the skills gap by enrolling front‑line teams in practical courses like Nucamp's AI Essentials for Work syllabus to ensure operators know how to prompt, validate, and audit outputs (Fresno State Artificial Intelligence Initiative, California GenAI executive order and sandbox guidance, AI Essentials for Work syllabus).

The clear next step for Fresno agencies: run narrow, low‑sensitivity pilots in the sandbox, document risk and workforce impacts, and scale only when audits show factual accuracy, human oversight, and concrete redeployment of saved hours.

Bootcamp	Length	Early bird cost
AI Essentials for Work	15 Weeks	$3,582

"My vision is to position our students, faculty, and staff at the forefront of technological innovation by integrating artificial intelligence (AI) across all aspects of our university equitably, ethically, and securely." - Saúl Jiménez‑Sandoval, Ph.D., President