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

Tacoma is quietly turning experiments into practical savings: with EPA grant funding the city is testing Prairie Robotics' truck‑mounted cameras that scan curbside bins for contamination to tailor outreach and cut processing costs, while across Washington officials have been using ChatGPT to draft emails, mayoral letters and policy documents - a mix of time‑savers that raises questions about accuracy and transparency.

Research on municipal AI shows the same levers Tacoma's pilots tap - automating repetitive work, speeding document processing, and freeing staff for higher‑value tasks - so city teams can balance efficiency gains with oversight and public trust.

For local staff looking to use AI responsibly on the job, structured training like Nucamp's AI Essentials for Work (15 weeks, prompt writing and practical AI skills) can help turn pilots into reliable, auditable services.

Learn more about Tacoma's curbside camera trial at SmartCities Dive, read reporting on Washington officials' ChatGPT use at KNKX, or explore the AI Essentials for Work course.

Attribute	Information
Description	Gain practical AI skills for any workplace; learn tools, prompt writing, and apply AI across business functions.
Length	15 Weeks
Cost	$3,582 (early bird); $3,942 afterwards. 18 monthly payments available; first payment due at registration.
Syllabus	AI Essentials for Work syllabus - Nucamp
Registration	Register for AI Essentials for Work - Nucamp registration

“AI is becoming everywhere all the time.” - Bellingham Mayor Kim Lund

Tacoma is turning a common curbside headache into a data‑driven fix: a phased, two‑year pilot - backed by a $1.8 million EPA Recycling Education and Outreach Grant - has mounted Prairie Robotics' smart cameras on recycling trucks to spot contamination in real time, send personalized postcards to households, and target outreach where it will cut processing costs most.

Early results are striking: one truck that ran an audit of roughly 600 households flagged 90 mistakes (about a 22% contamination rate), underscoring why the city is starting with plastic bags and bagged recyclables before widening the scope and hopes to outfit seven trucks by year‑end.

The system is designed to protect privacy (faces and plates are blurred, images stored in the U.S., and no fines during the pilot) while giving staff actionable hotspots on a dashboard.

For a clear summary of the program and technical goals see SmartCities Dive coverage of the Tacoma recycling pilot or Waste Today roll‑out details for the program.

Attribute	Detail
Grant	$1.8 million (EPA Recycling Education & Outreach)
Duration	Two‑year pilot through mid‑2027
Vendor	Prairie Robotics
Initial focus	Plastic bags and bagged recyclables
Current / goal trucks	1 in operation; goal of 7 by year‑end
Contamination snapshot	~22% (90 mistakes out of 600 households)
Resident outreach	Postcards with images of contaminants (no penalties)
Privacy	Blurring of faces/license plates; U.S. data storage; images not sold

“Contamination impacts how we can deliver services and the cost of those services for all residents.” - Lewis Griffith, Tacoma Solid Waste Management Division Manager

Grant County's Multi Agency Communications Center (MACC) has quietly shown how conversational AI can unclog 911 operations: since launching its AVA assistant in May 2024 to handle non‑emergency lines, the system processes short, bilingual (English/Spanish) reports in minutes so telecommunicators can prioritize true emergencies and critical radio traffic, cutting wait times and operator burden (one early audit estimated about 1,500 hours saved).

AVA acts more like a virtual colleague than a menu tree - callers can correct answers, share locations or stream video, and staff can review transcripts or intervene in near‑real time - while clear guidance reminds people to dial 911 for emergencies.

Local reporting tracks rapid adoption (tens of thousands of calls in the first months and roughly 70,000 non‑emergency contacts by AVA's one‑year mark), and stories range from a caller reporting a found, “small, fluffy, black and white” dog to a Memorial Day case where AVA rerouted a head‑injury report to a dispatcher during a peak period.

Read Firefighter Close Calls' summary of the rollout or the ONE News piece on AVA's first year for more on how the system is reshaping call flow in Washington.

Metric	Value
Launch	May 2024
Calls handled (one year)	~70,000
Average non‑emergency call time	~3 minutes
Estimated operator time saved	~1,500 hours
Languages	English and Spanish

“AVA solves several primary issues for MACC dispatchers. Prior to AVA, dispatchers frequently placed non-emergency callers on hold to answer 911 calls or to respond to critical radio traffic… AVA allows the dispatchers to prioritize the workload by processing requests on the non-emergency line within minutes of the call to AVA.”

Seattle's Project Green Light shows how AI can shave wasted minutes and tailpipe pollution without ripping up streets: a public‑private pilot with Google Research has adjusted signal timing at three key junctions (15th & Market, Greenwood & 80th, and NW 53rd in Ballard) using anonymized driving trends to create “green waves,” cut unnecessary stops by as much as 30%, and slice intersection emissions roughly 10% - all delivered at no charge to the city and designed so engineers implement recommendations into existing controllers quickly; early results have encouraged plans to extend timing changes near event hubs like Lumen Field and to monitor impacts on multimodal safety and transit priority.

For a city juggling 1,125 traffic lights, these lightweight, data‑driven tweaks offer a tangible win: fewer red‑light idling moments, less local pollution, and faster commutes that can add up to real fuel and time savings.

Read the Seattle Department of Transportation overview of Project Green Light or the Scientific American analysis for additional context and caveats.

Metric	Value
Pilot intersections	3 (15th & Market; Greenwood & 80th; NW 53rd in Ballard)
Potential stop reduction	Up to 30%
Estimated emissions reduction	About 10% at treated intersections
Cost to Seattle	Project provided at no charge

“Solving urban traffic is not rocket science. It's more difficult.” - Aleksandar Stevanovic

Washington law enforcement is experimenting with AI to turn "thousands of hours" of body‑camera footage into usable training insights: Spokane County won a nearly $945,520 DOJ grant to pilot TrustStat, a Polis Solutions system that uses large language models to analyze speech and image‑processing to flag movements, facial expressions, and voice intonation tied to de‑escalation and use‑of‑force scenarios, then compare those signals to baseline readings captured during training so instructors can see what works in the field.

The three‑year effort is explicitly scoped for training‑practice review (not adjudicating force), aims to build empirical guidance for better coaching, and will include an oversight process for setting questions and actions - a useful next step for agencies juggling far more video than humans can review.

Read the Spokesman‑Review coverage of the Spokane TrustStat project or the KREM report on the DOJ grant and rollout for full details.

Attribute	Detail
Grant	$945,520 (DOJ Bureau of Justice Assistance)
Project	TrustStat
Vendor	Polis Solutions (Dallas)
Duration	3 years
Scope	Analyze body‑worn camera video for training effectiveness (de‑escalation, use of force)
Tech	LLMs for speech; image processing for movement/facial cues
Limit	Not for determining legality/appropriateness of force; focused on training review

“My interest has always been human performance. How do we make the most professional law enforcement officers we can, in every sense of the word? How do we get people who are high performers to perform even better?” - Sheriff John Nowels

Washington's DNR has turned a modern set of eyes on the state's fire-prone landscape: a five‑year pilot launched in 2023 now operates 21 Pano AI stations (with five more coming online) that deliver 360‑degree, high‑definition panoramas and AI smoke detection - capable of spotting ignitions within roughly a 15‑mile radius - and relay verified alerts to DNR dispatch so crews can be routed with precise longitude/latitude data; the live feeds were opened to the public this summer so residents can literally watch the system at work.

The program, funded through the Wildfire Response, Forest Restoration, and Community Resilience Account created by 2021's House Bill 1168, has already helped identify incidents on both sides of the state and pairs machine detection with human review to reduce false alarms.

For a state juggling shrinking budgets, this lightweight mix of cameras, cloud AI, and real‑time staffing offers a pragmatic speed boost to keep small fires small.

Attribute	Detail
Cameras live	21 (plus 5 being installed)
Pilot start	2023 (five‑year pilot)
Funding	Wildfire Response, Forest Restoration, and Community Resilience Account (HB 1168)
Field of view / tech	360° HD stations; AI plus human verification; 24/7 monitoring
Public feed	Washington DNR Wildfire Watch public feed

“Early detection is a key part of DNR's wildfire rapid response model, and now Washingtonians can peek behind the scenes at how part of that detection process works.” - George Geissler, DNR State Forester and Deputy Supervisor

Pierce County's social‑services intake has become a testbed for practical AI that cuts paperwork and speeds families into shelter: with a $1 million county grant to stand up a centralized Parkland intake hub and a separate $500,000 Google Workspace grant to build the Ash Nazg system, Family Promise of Pierce County runs a 24/7 hotline, GetBed.org access, and short intakes that can be completed in seven questions and 15 languages - shrinking an hour‑long interview to minutes while AI spits out an intake report, text/email referrals, e‑signature packets, and even a resume in under two minutes.

Early results show rapid throughput (thousands of contacts and hundreds of placements per month) and a clear “so what?” - people in crisis get matched to shelter or resources the same day instead of waiting in a phone tree.

For a detailed account of Ash Nazg and the hub's rollout see The News Tribune's coverage of Family Promise's work or read Family Promise's programs & services summary to learn how prevention, diversion, and case management tie into the tech.

Metric	Detail
County grant	$1,000,000 (Shelter Access Hub Grant)
Google grant	$500,000 (Workspace grant for Ash Nazg)
Hub location	Parkland intake hub (operations also reported from Spanaway portable site)
Short intake	7 questions; available in 15 languages
Deeper case data	~240 data points for in‑depth management
Outcomes	1,452 individuals referred into shelter; avg. ~121 placements/month

“People can go to GetBed.org, they can call us, they can text us, they can show up in-person with an appointment, and we can do the intake in 15 different languages and connect them to emergency resources.” - Steve Decker, CEO, Family Promise of Pierce County

Washington counties are using anonymized location intelligence to turn guesswork about tourism and retail into concrete, budget-ready actions: tools like Placer.ai let planners count cell‑phone entries and exits (Grant County used it at the county fair in 2023), measure dwell time and visitor origin, and even estimate spending and average income to guide everything from police staffing to lodging‑tax marketing campaigns; Snohomish County and Monroe used the same platform to learn that Monroe drew about 235,600 visitors in November, a striking, concrete number that helps officials plan staffing, targeted ads, and event ROI instead of relying on gut instinct.

The data's practical payoff is clear - cities can geo‑fence a site to test a campaign, track visitors year‑over‑year, and fine‑tune proposals for new businesses - while vendors emphasize privacy by stripping personal identities.

For an overview of AI pilots in Washington see the Municipal Research and Services Center summary, read Placer.ai's product page for platform capabilities, or review local reporting from Snohomish County for hands‑on examples.

Use case	Washington example / metric
Event attendance	Grant County fair counts (Placer.ai, 2023)
Tourism & marketing ROI	Monroe: ~235,600 visitors in November (Snohomish County reporting)
Public safety & staffing	Snohomish increased police staffing at big events based on foot‑traffic data

“Instead of a gut feeling, it gives us a good, solid guess at what's happening.” - Melody Dazey, Sky Valley Chamber of Commerce

MRSC summary of AI pilots in Washington | Placer.ai product page describing location-intelligence platform capabilities | Snohomish County local reporting and examples of location-intelligence use

Washington is turning cloud modernization from a one-off project into a statewide service strategy that actually moves the needle on speed, security, and cost: WaTech's Enterprise Cloud Computing Program (ECCP) lays out a vision for a secure, scalable statewide cloud ecosystem while finalizing both FinOps and an AI roadmap to make migrations predictable and auditable, and the Cloud Government Network (CGN) is already live with Azure Domain Controller as a Service - meaning agencies can adopt shared cloud services instead of maintaining costly datacenters.

Training and change management matter too: an enterprise Pluralsight rollout now reaches nearly 50 agencies and ~2,700 IT staff, and practical case studies show the payoff - L&I migrated more than 300 servers to Azure in under nine months, reporting measurable cost savings and lower emissions - proof that shifting from upfront capital-heavy systems to an operational, cloud-first model can speed service delivery for residents.

For teams in Tacoma, the takeaway is straightforward: pair WaTech's ECCP guidance with pragmatic migration playbooks like the L&I example to cut technical debt, gain budget predictability, and make government services feel faster and more reliable for Washingtonians.

Metric	Detail
WaTech program	WaTech ECCP statewide cloud roadmap with FinOps and AI roadmaps
Cloud Government Network	Live in production (Azure DCaaS; AWS testing complete)
Training reach	Pluralsight: ~50 agencies, ~2,700 IT employees enrolled
Agency case study	Washington L&I Azure VMware migration case study - 300+ servers migrated to Azure in 9 months

“We want to be the Amazon Prime of government. That's our goal.” - Bill Kehoe

Procurement and governance are becoming the backbone of Washington's practical AI rollout: Gov. Jay Inslee's Executive Order 24‑01 pushed agencies to stop treating generative AI as a novelty and start writing the rules - WaTech now leads a slate of deliverables from an initial Generative AI report to procurement guidance, risk assessments for high‑risk systems, and deployment checklists so purchases aren't just cheap tools but auditable services with guardrails.

That means Tacoma teams buying AI will see concrete requirements - contract clauses, ADS procurement guidance, and interim “purposeful and responsible” use rules that lean on NIST and equity review - so vendors must bring governance plans, not just slick demos.

The upside is clear: well‑crafted procurement can turn pilots into repeatable, monitored services that reduce vendor risk and protect vulnerable residents; the memorable test is simple - won't sign a contract until the vendor shows its AI governance playbook.

Read the Governor Jay Inslee Executive Order 24-01 summary or explore WaTech generative AI resources and procurement guidance for the specific procurement and monitoring deliverables.

Deliverable	Notes
State Generative AI Report	Catalog potential initiatives and agency use cases
Initial Procurement Guideline for GenAI	Contract language, vendor expectations, oversight
Risk assessments for high‑risk AI	Identify systems requiring extra review and monitoring
Guidelines for Deployment	Interim rules for purposeful and responsible use
Workforce & Equity Reports	Analyze impacts on jobs, education, and vulnerable populations

“Our goal is to help the state continue using generative AI in ways that help the public while putting up guardrails around uses that present a lot of risk.” - Katy Ruckle, Washington's chief privacy officer

Washington's local AI scene is a practical, fast-moving mix of boutique consultancies, plug-and-play vendors, and curious city halls - a neighborhood where startups and small firms can partner with outside developers to turn ideas into working tools.

Firms advertising end-to-end AI work, from machine learning and predictive analytics to computer vision, OCR, and generative-AI chatbots, make it realistic for a Tacoma business to prototype a custom automation without building a data science team in-house; see MMC AI consulting and development services for examples of those capabilities.

At the same time, public-sector demand is real: reporting shows Washington officials are already using ChatGPT to draft emails, policy memos, and other materials, while Tacoma's curbside camera trial signals appetite for computer-vision solutions in local services.

That combination creates opportunity - but also a reminder that procurement and policy are catching up, so small vendors should package clear governance, accuracy checks, and transparency into their proposals if they want municipal contracts; for context read the KNKX report on ChatGPT use in Washington and SmartCities Dive coverage of Tacoma's pilot.

For Tacoma teams just getting started, the smart first move is education plus guardrails: begin with accessible primers like Tacoma Community College's ongoing “Generative AI” guide and a short, practical course (Skillsoft's 29m 41s “Navigating AI Ethical Challenges and Risks”) to build common vocabulary and safe habits, then bake in the four pillars - fairness, transparency, accountability, and privacy - that protect residents and reputations.

Practical risks to watch for include biased outcomes, data‑privacy gaps, and fast‑moving fraud vectors (deepfakes and voice cloning are already real threats), so pair small, auditable pilots with human review, data minimization, and documented decision‑trails.

For a deeper look at systemic failure modes and governance approaches, consult broader safety work such as the AI Safety textbook; combined, these resources help convert experimentation into repeatable services that save money without trading away trust or legal exposure.

Tacoma's next moves should be pragmatic and measurable: finish the phased rollout to the goal of seven Prairie Robotics–equipped trucks, use the city dashboard and the vivid early snapshot (one truck flagged 90 mistakes out of 600 households, ~22% contamination) as a baseline, and set simple success metrics - contamination reduction, outreach response rates, and per‑ton processing cost savings - to decide whether to continue after the two‑year pilot (through May 2027, with a June 2027 program review).

Protecting trust means keeping the privacy and data‑storage promises in place (blurred plates/faces, U.S. storage) and codifying those practices in vendor agreements, while making resident education - postcards with clear photos and friendly guidance - the public face of the work.

Pair operational rollout with staff training so teams can interpret dashboard hotspots, audit detections, and manage vendor performance; for hands‑on, workplace‑focused instruction, Tacoma teams can review the AI Essentials for Work syllabus - Nucamp and local reporting from King5 or SmartCities Dive to mirror lessons learned.

Start small, measure quickly, and let the data (and resident feedback) drive the scale decision.

Metric	Detail
Federal grant	$1.8 million (EPA Recycling Education & Outreach)
Pilot duration	Two‑year pilot through May 2027
Current / goal trucks	1 in operation; goal of 7 by year‑end
Contamination snapshot	~22% (90 mistakes out of 600 households)
Next review	June 2027

“It can see in real time what materials are actually being dumped into the truck.” - Preston Peck, Tacoma Environmental Services

AI Essentials for Work syllabus - Nucamp | King5 Tacoma reporting on local waste and recycling | SmartCities Dive coverage of smart waste systems