HomeMy WebLinkAbout2026/08/19 - ADMIN - Agenda Packets - Community Technology Advisory Commission - Regular Community technology advisory commission meeting
August 19, 2026
6 p.m.
If you need special accommodations or have questions about the meeting, please call Jacque Smith at 952.924.2632 or the
administration department at 952.924.2505.
Community technology advisory commission
The community technology advisory commission is meeting in person at St. Louis Park City Hall,
5005 Minnetonka Blvd., in the Westwood Room on the third floor. Members of the public are
welcome to attend in person.
Visit www.stlouispark.org/government/boards-commissions to view the agenda and reports.
Agenda
1. Call to order – roll call (6 p.m.)
2. Approval of minutes – July 15, 2026
3. Presentations - none
4. Business
a. Review submitted signal items (10 minutes)
i. Next steps
ii. September signal assignments
b. Review AI whitepaper (20 minutes)
c. Prep for Sept. 14 annual meeting with city council (10 minutes)
5. Communications and announcements (5 minutes)
a. Staff
b. Commission members
6. Adjournment
Future meeting/event dates:
Sept. 14, 6 p.m., CTAC annual meeting with council
Sept. 16, 6 p.m., regular meeting
Oct. 21, 6 p.m., regular meeting
Nov. 18, 6 p.m., regular meeting
Dec. 16, 6 p.m., regular meeting
CTAC Purpose Statement: The Community Technology Advisory Commission advises the city council and
other city bodies on how technology services can improve resident services and enhance quality of life. By
researching emerging technologies, assessing their potential impact, and recommending solutions
grounded in clearly defined problems, the commission helps ensure that technology decisions support the
city’s long-term strategic goals.
Community technology advisory commission meeting
July 15, 2026
6 p.m.
Community technology advisory commission
Minutes
Members present: Elsa Anderson, Reid Anderson, Edward Bulliner, Rudyard Dyer, Lindsay Keogh,
Konnor Slaats, Kayla Stautz
Members absent: Ben Straus, Tom Marble
Staff liaison: Jacque Smith, communications and technology director
1. Call to order – roll call
The meeting was called to order at 6:04 p.m.
Members introduced themselves. Chair Slaats noted he has served on the commission since
2021 and works in product operations. Commission Member Stautz said she has been on the
commission for close to a year and works as general manager of the consumer business for a
technology company. Commission Member Reid Anderson noted he works as a software
engineer and has been on the commission for just under a year. Vice Chair Dyer noted he has
lived in St. Louis Park for over a decade, has served on the commission since 2022, and works in
due diligence at an alternative investment management company.
Commission Member Bulliner introduced himself as new to the commission and to St. Louis
Park, having moved here approximately a year and a half prior. He noted he works in geographic
information systems software, has a background in statistics and data science, and previously
worked in government. Commission Member Keogh introduced herself as new to the
commission and noted she works as a product manager, focused on search and digital
experiences. Commission Member Elsa Anderson introduced herself as a 16-year-old high school
junior.
2. Approval of minutes – April 15, 2026
It was moved by Commissioner Slaats, seconded by Commissioner Anderson, to approve the
April 15, 2026, minutes. The motion passed.
3. Presentations – none
4. Business
Review and approve purpose statement
Ms. Smith displayed the current purpose statement alongside a proposed draft for the
commission's review. The draft statement introduced language around city-provided public-
facing technology.
Chair Slaats expressed a preference for broader language, suggesting that a focus solely on
public-facing technology could exclude relevant matters such as technology by outside entities
which the city may have to respond to, for example, data centers. He proposed language
referencing how technology can improve resident services and enhance quality of life.
Commission Member Keogh agreed that framing the purpose around residents and quality of
life would appropriately capture the commission's advisory scope, noting that the draft reflected
the city's strategic direction and the increasingly interwoven role of technology across city
functions.
Vice Chair Dyer suggested the language could also acknowledge technology that could
negatively affect quality of life. He proposed morphing the original and draft statements to
reference how city-provided technology can affect residents.
Commission Member Bulliner noted the importance of a holistic perspective and acknowledged
that back-end systems often affect public-facing outcomes.
Commission Member Reid Anderson suggested simplified language referencing city-provided
technology without the public-facing qualifier.
After discussion the commission agreed to modify the draft to reference how technology can
improve resident services and enhance quality of life, removing the public-facing qualifier while
retaining the general structure of the draft statement. The revised statement was approved, as
follows:
“The Community Technology Advisory Commission advises the city council and other city bodies
on how technology services can improve resident services and enhance quality of life. By
researching emerging technologies, assessing their potential impact, and recommending
solutions grounded in clearly defined problems, the commission helps ensure that technology
decisions support the city’s long-term strategic goals.”
Work plan items: background, workgroups, progress
Chair Slaats provided background on the commission's work over the past year. He described
the commission's role as an advisory body to city council and noted that a meeting with council
in April 2026 helped clarify expectations, with a follow-up council meeting scheduled for
September 2026. He explained that the commission had split into two workgroups: one focused
on artificial intelligence and one focused on evaluating the city's digital presence, including the
city website and the My St. Louis Park app.
Chair Slaats then presented a proposed framework for how the commission might structure its
future work. He described a multi-stage process involving three phases: signals, exploration and
active work.
In the signals phase, commission members would bring forward brief summaries of technology
trends, peer city activities, policy developments or resident-relevant issues. These would be
submitted to Ms. Smith in advance of meetings for distribution with meeting materials and
discussed at the start of each meeting, with the commission deciding collectively whether to
watch, explore further or set aside each signal. Chair Slaats suggested that two members might
bring signals per meeting but that anyone could contribute.
Commission Member Bulliner expressed support for receiving advance materials so members
could formulate informed opinions prior to discussion.
Commission Member Reid Anderson noted the framework helpfully defines what substantive
commission work can look like for members who wish to contribute more.
Commission Member Stautz noted that technology evolves quickly and that the framework
would allow the commission to be more current and relevant. She also confirmed her
understanding that all three stages would operate simultaneously, with work moving through
the pipeline.
Commission Member Keogh noted the framework allows work to be taken in manageable
increments and raised the question of how resident voices might be incorporated into the
process, acknowledging that was likely a future consideration.
In the exploration phase, one or two members would draft a one-page framing of a potential
topic, working toward a clear problem statement. Chair Slaats offered an example question
related to the digital channels workgroup: what principles should the city use to determine if a
resident-facing digital service provides sufficient public value.
Vice Chair Dyer indicated he was open to reframing or restarting the digital channels workgroup
given the extended period of limited progress, much of which was due to waiting for the city's
website redesign. He noted the commission had previously reviewed website analytics and
encountered questions about the overlap and distinction between the city website and the
city’s app. Vice Chair Dyer suggested that a one-page exploration document could be a useful
way to reframe that work with new members.
In the active work phase, a lead commission member would outline the project and identify key
decisions, with two or three contributors drafting sections. The full commission would then
critique, debate and approve a final version for submission to council.
Commission Member Stautz asked whether it would be worth attempting the signals
component before the next meeting as a trial. Chair Slaats agreed it would. The commission
decided to ask members to submit signal summaries to Ms. Smith by Aug. 12, 2026, ahead of
the Aug. 19, 2026 meeting.
Ms. Smith noted the timing is well suited to the upcoming September council meeting and
suggested that the next meeting agenda include discussion of how to present the commission's
new direction to council.
Chair Slaats noted the draft AI paper compiled by Commission Member Stautz would be
distributed with the next meeting packet for review and discussion, along with the framework
document he had emailed the prior evening.
Regarding the AI workgroup, Vice Chair Dyer noted progress had been set back when a former
member departed. Chair Slaats indicated the workgroup was approaching completion and that
distributing the draft ahead of the next meeting would allow the full commission to review and
provide feedback.
Ms. Smith confirmed she would distribute all relevant materials to members in advance of the
Aug. 19, 2026 meeting. She also noted she would resend calendar invites for upcoming meetings
to all members.
Ms. Smith provided an update on the city's website redesign, noting a soft rollout is scheduled
for Aug. 17, 2026. She described improvements including enhanced accessibility, improved
search functionality and better mobile responsiveness.
5. Communications and announcements
There were no communications or announcements
6. Adjournment
It was moved by Chair Slaats, seconded by Commission Member Dyer, to adjourn. The motion
passed. The meeting adjourned at 6:47 p.m.
These minutes were created with the assistance of a generative AI transcript service, then edited and
finalized by a city staff person.
CTAC signals
A technology trend I have observed in St. Louis Park is the rapid growth of AI use,
specifically in schools. As a student, I spend most of my time surrounded by other
students and in the past year irresponsible use of AI emerged as a big discussion in
schools. Many students utilized the increasing availability and quality of AI for school
work and as a result many assignments had to be changed and even some curriculum.
Many assignments had to be completed on paper and only in class which I, along with
many other students found reduced the quality of the work. I believe that this will
become an even bigger issue in the upcoming school year as AI continues to improve
and the school board and city council will have to find a solution to ensure students'
quality of education remains strong. CTAC could be useful in providing resources for
teachers on how to avoid the misuse of AI while keeping the quality of education strong.
䅵杵獴
The growing importance of AI- and data-driven technology in city technology means that
responsible data stewardship is becoming core to good, transparent governance.City
government plays a critical role in re-establishing trust in government overall. If residents do not
trust how their data is handled by the city, it leads to more distrust in government and overall
dissatisfaction.
Common solutions used by cities, especially in the surveillance/policing space, gather and
share more data than ever with outside parties. Once data is collected by these sources, how
the data is used & accessed can be a black hole or pose a risk to the safety of the community,
depending on who gets access to the data.
However, some cities take a community-centered, transparent approach to solutions that gather
& process resident data, which both mitigates risk of data misuse, unwarranted surveillance,
and improves community trust.
Trends
● Privacy/security concerns
○ Flock: West St Paul, Burnsville, Duluth, Isanti county all have ended or chose to
not enter into an agreement with Flock, citing privacy concerns & spuriousness
about Flock’s behavior as a vendor partner
○ Unwarranted/rampant usage of data by ICE during the federal invasion to
intimidate observers and community members - sourced from tools such as
Accurint
○ Backlash against Skydio drones in minneapolis
● High-profile problems lead to distrust of AI-driven systems, wasting community
resources and underscoring the need to build trust through other mechanisms
○ 1.4 billion bad notifications a month lead to unnecessary traffic stops or effort
○ Flock: LA saw massive accuracy issues, leading them to end their pilot
○ Plymouth man pulled over by four cops due to a plate read error
○ Tennessee woman wrongfully jailed for six months based on AI facial recognition
Potential policy question: How should St Louis Park evolve its data usage & sourcing practices
to achieve its strategic vision, particularly around good governance & creating a welcoming,
safe community?
CTAC could:
● Propose a framework for gathering resident input to study the topic further
● Investigate strategies, principles, and risks for current & future resident-facing
technology and data collection
Sources:
https://www.govtech.com/transportation/for-cities-collecting-visual-data-brings-privacy-concerns
䅵杵獴
https://www.brookings.edu/articles/how-citizens-and-local-governments-can-advance-data-drive
n-policymaking-together/
https://mcity.umich.edu/wp-content/uploads/2023/03/Privacy-Frameworks-for-Smart-Cities_Whit
e-Paper_2023.pdf
August 2026 Signal
https://www.lmc.org/news-publications/magazine/july-august-2026/letter-of-the-law-july-2026/
What’s happening?
The League of Minnesota Cities ’July–August 2026 guidance says information entered into AI
systems, AI-generated outputs and associated system data may qualify as government data
under the Minnesota Government Data Practices Act.
It warns that putting nonpublic information into third-party AI tools could constitute a data breach
and that prompts or outputs may need to be retained, searched and disclosed.
Why might this matter to council?
SLP already has a GenAI use policy that prohibits use of medium-to-high risk data in GenAI
tools (March 18 packet). SLP also already urges compliance with the Minnesota Government
Data Practices Act, which the League article is referencing.
Is the League guidance going further by saying all inputs/outputs may be government data and
need to be tracked? Is that something SLP and other cities are tracking today, or are they just
tracking final outputs?
What policy or strategic question could this eventually create for Council?
Does the city’s current approach account for prompts, input histories, intermediate outputs, and
other AI system data, or primarily final documents?
If the city’s current policy covers these, this signal may not require further action.
If a gap exists, Council could eventually consider whether the city needs clearer standards for
retaining, retrieving, and classifying AI-related records. Any policy change would likely be
narrow.
Why might CTAC be useful?
CTAC could help if staff identifies an unresolved policy question. Potential work could include
comparing how peer cities treat prompts, intermediate outputs, and AI metadata under public-
records requirements.
Otherwise, the League guidance could validate St. Louis Park’s existing approach rather than
justify additional CTAC work.
Future of AI for Cities
A Policy White Paper
Community Technology Advisory Commission | City of St. Louis Park, Minnesota
2026
Executive Summary
Artificial intelligence is no longer an emerging technology. Across the United States and
internationally, cities of every size are deploying AI tools to detect potholes, process permit
applications, answer resident questions, translate public meetings, and manage aging
infrastructure. The question for St. Louis Park is not whether AI will affect city government, but
how the city will govern, evaluate, and integrate it responsibly.
This white paper, prepared by the Community Technology Advisory Commission, explores how
cities can adapt their policies, processes, and governance frameworks for an AI-enabled future.
It draws on examples from peer cities, national guidance from the National League of Cities
(NLC), and the city's own existing technology posture.
The paper is organized around five topic areas: (1) the cross-cutting risks and governance
principles that apply to all AI use; (2) internal city operations, including back-office tools and
workflow acceleration; (3) public-facing services such as resident chatbots and meeting
transparency tools; (4) public works and infrastructure operations; (5) planning, zoning, and
permitting; and a forward-looking look at agentic AI and what may come next.
Across all areas, the Commission reaches a common conclusion: the value of AI depends on
the quality of underlying data, the clarity of governance structures, the preservation of human
accountability, and a sustained commitment to equity. St. Louis Park is better positioned than
many peer cities because it has already adopted internal generative AI guidelines and has
existing resident-service channels like MySLP. The next step is to extend that framework
thoughtfully as new use cases emerge.
Section 1: Risks, Governance, and Guiding Principles
Before evaluating individual AI applications, the Commission believes it is important to establish
a shared governance vocabulary and identify the risks that apply across all municipal AI use
cases. This section frames the core issues that should inform every subsequent decision the
city makes about AI.
Cross-Cutting Risks
AI Slop and Hallucination. Generative AI systems can produce confident, fluent, and entirely
incorrect answers. A prominent municipal example: in 2024, New York City's MyCity chatbot
advised business owners that they could legally reject Section 8 housing vouchers and that
serving cheese bitten by a rat was permissible. Deploying AI tools that interact directly with the
public without rigorous accuracy review creates real reputational and legal exposure.
Resident Data and Privacy. Public AI models may retain user inputs for training. Staff using
unapproved tools can inadvertently expose resident financial records, health information, or
other protected data. In early 2026, the acting director of the U.S. Cybersecurity and
Infrastructure Security Agency reportedly triggered internal security warnings after uploading
sensitive government documents into a public ChatGPT instance. Third-party contractors may
also process city data through unauthorized AI systems without the city's knowledge.
Surveillance Creep. Camera networks, mounted sensors, and detection tools deployed for
narrow purposes such as pothole detection can be tempting to repurpose for enforcement,
public safety monitoring, or behavioral surveillance. Governance frameworks should establish
clear use-case boundaries and require council review before any expansion.
Equity in Service Delivery. AI tools trained on complaint data may prioritize areas where
residents are more likely to report problems, systematically underserving neighborhoods with
lower reporting rates, language barriers, or lower institutional trust. Cities must actively monitor
whether AI improves or exacerbates service equity.
Unpredictable Costs. Operational costs for AI can fluctuate by orders of magnitude based on
small changes in how systems are prompted or how often they self-correct. Without hard
spending limits and performance boundaries, AI tools can quietly become fiscal liabilities.
Vendor Dependence and Data Lock-In. Many AI tools are embedded in proprietary platforms.
Cities should evaluate whether they can export their data, audit model performance, switch
vendors, and integrate tools with core systems before signing contracts.
Records and Legal Obligations. Images, videos, work orders, and resident reports may
become government records. Under the Minnesota Government Data Practices Act, data
created with AI assistance may need to be retained and produced in response to public data
requests.
A Governance Framework for Council Questions
The National Institute of Standards and Technology (NIST) AI Risk Management Framework
provides a useful foundation. The Commission has distilled its core functions into four questions
that council members and staff can apply to any proposed AI deployment:
NIST Function Council Question What to Look For
Govern Who is ultimately accountable if
this fails and responsible for
associated costs?
A specific human owner or
position, not just an IT vendor.
Map What specific problem does this
solve, what data does it touch,
and how is it paid for?
A clear use case with strict
boundaries on what resident
data is involved. Control
mechanisms to prevent runaway
costs.
Measure How will we prove this system is
unbiased, accurate, and
cost-constrained?
Pre-deployment testing plan,
ongoing performance audits,
and consistent cost
measurement.
Manage What is the human fallback plan
when the AI makes a mistake or
is unavailable?
A documented mitigation
strategy and a clear off-switch if
things go wrong.
These four questions do not resolve every governance challenge, but they ensure that each AI
deployment has a named human owner, a defined problem scope, a measurement plan, and a
documented fallback. The Commission recommends that St. Louis Park require documented
answers to these questions before approving any new AI system.
Principles for Responsible AI Use
Drawing on NLC guidance and peer city experience, the Commission recommends the following
principles for all AI activity in St. Louis Park:
• AI should support human decisions, not replace them. Staff, commissioners, and council
should retain accountability for all interpretations, recommendations, and approvals.
• AI should be attached to defined operational problems, not deployed speculatively.
• All AI tools should go through IT and legal review before deployment, including free and
pilot products.
• AI-generated content in public-facing channels should be disclosed and cite authoritative
sources.
• Performance should be measured against equity metrics, not just efficiency metrics.
• The city should maintain a public inventory of AI systems in use, consistent with best
practices from cities like San Jose and Lebanon, NH.
St. Louis Park's existing internal generative AI guidelines already reflect many of these
principles. The task ahead is extending that framework to cover operational, perceptive, and
predictive AI applications as they emerge.
Section 2: Internal City Operations
City employees across departments increasingly interact with AI tools in their daily work,
whether the city has sanctioned those tools or not. A proactive strategy for internal AI use
reduces risk, improves consistency, and creates opportunities to demonstrate AI's value to staff
and elected officials before expanding into higher-stakes resident-facing applications.
The Current Landscape
Cities are deploying AI for internal operations in three broad categories: productivity tools for
individual staff, workflow acceleration across departments, and data analysis and decision
support.
San Francisco has led among peer cities, adopting Microsoft Copilot for writing city reports,
analyzing data, and summarizing documents across departments. San Jose issued an RFP in
2025 for a city-wide generative AI platform for employee use. Seattle and Boston have each
published internal generative AI policies governing staff use of tools like ChatGPT, Gemini, and
Copilot.
According to a 2025 global study of 250 cities, 56 percent are actively piloting or using AI to
upgrade government operations and services, and 83 percent plan to do so within three years.
The trend is clear: cities that establish governance frameworks now will be better positioned to
adopt tools responsibly as they mature.
St. Louis Park has already taken a meaningful step by publishing internal generative AI
guidelines. The next phase is building on that foundation by clarifying which tools are approved,
how staff are trained, and how the city will evaluate and expand internal AI use over time.
Key Internal Use Cases
Staff Productivity and Writing Assistance
AI copilot tools integrated with productivity suites (Microsoft 365, Google Workspace) can help
staff draft reports, summarize meeting notes, respond to routine correspondence, and search
internal knowledge bases. Early research suggests these tools are especially helpful for newer
or less-experienced employees. The risk is over-reliance on AI-generated content that has not
been fact-checked or verified against authoritative sources.
Peer example — Chattanooga, TN: The city created a “Prompt Library” of pre-written prompts
to guide generative AI tools for specific staff tasks, standardizing outputs and reducing errors
across departments.
Grant Writing and Administrative Support
AI can assist staff with grant research, identifying eligible programs, drafting application
narratives, and checking compliance with grant requirements. This has particular value for
smaller cities where grant-writing capacity is limited.
Permitting and Workflow Acceleration
Several cities have integrated AI into permitting workflows to reduce review cycles, catch
incomplete applications early, and improve consistency across staff reviewers.
Peer example — Hamilton, ON: The city used AI to scan first-stage building permit
applications for compliance with city rules, building codes, and zoning requirements, achieving a
60 percent decrease in permit processing time and a 70 percent decrease in overall review time
from application receipt to conditional approval.
Peer example — Los Angeles Planning: The Planning Department uses AI to analyze land
use regulations and zoning data to proactively identify issues before formal permit application
review begins, reducing back-and-forth between applicants and staff.
Peer example — Honolulu: CivCheck, an AI-guided permit preparation tool, helps applicants
check for code compliance and required documents before submission, reducing review time by
catching problems upstream.
Policy Implications for St. Louis Park
For internal operations, the Commission recommends the following:
• Extend the existing generative AI guidelines to cover productivity tools, copilots, and any
AI embedded in core software platforms such as the city’s permitting or work-order
systems.
• Establish an approved tools list. Following Seattle’s model, require that all AI software,
including free and pilot products, go through IT and legal review before staff use.
• Invest in staff training. Partner with the League of Minnesota Cities, MNIT, or local
educational institutions to develop AI literacy programs for staff at all levels.
• Create a feedback channel for staff to surface AI use cases, identify problems, and
share what is working across departments.
• Set spending controls. Any AI tool with usage-based pricing should have a hard
spending cap and monitoring process in place from day one.
Section 3: Public-Facing AI and Resident Engagement
Public-facing AI applications carry higher reputational and governance stakes than internal tools
because errors affect residents directly and can undermine trust in city government. At the same
time, well-designed resident-facing tools can meaningfully improve access to services, reduce
barriers for non-English speakers, and extend the city’s reach outside business hours.
Chatbots and Self-Service Assistance
AI-powered chat interfaces are increasingly common in city government. When well-governed,
they can handle routine informational inquiries, help residents navigate city services, submit
service requests, and identify the right staff contact for complex issues. When poorly governed,
they produce incorrect guidance with high confidence, as New York City’s MyCity chatbot
demonstrated in 2024.
The key design principles for safe resident-facing chatbots are: ground responses in
authoritative city documents and ordinances; include clear disclaimers; provide a path to human
review for any consequential decision; and disclose that the user is interacting with an AI.
Ann Arbor, MI "Ask Ann" provides 24/7 chat assistance in 71 languages, helping residents
find city services, submit requests, and contact staff. Unresolved questions
are escalated to staff during business hours.
Dearborn, MI AI translation and virtual chatbot tools serve a community where more than
half of 110,000 residents speak a language other than English at home.
Atlanta, GA ATL311 chatbot handles routine service requests and common questions,
reducing call-center volume and improving after-hours responsiveness.
Midland, TX "Ask Jacky" is integrated with SeeClickFix to allow residents to report issues
and track status via an AI-powered chat interface.
Kelowna, BC Building permit chatbots had over 5,000 weekend conversations and 6,500
after-hours conversations in the first six months of 2024, reducing routine
staff inquiries.
Meeting Transparency and Accessibility
AI tools for meeting accessibility address two related challenges: making government
proceedings understandable to residents who cannot attend in person, and serving residents
who do not read or speak English fluently.
Meeting agenda explainers: Saratoga, CA deployed an AI tool called Hamlet that summarizes
council agendas, supporting documents, and meeting recordings in plain language. Government
Technology reports that several cities are using similar tools to improve transparency and
resident engagement.
AI translation and captioning: Sunnyvale, CA provides real-time translation and captioning at
city council meetings. Smart Cities Dive reports growing use of AI captioning tools like Wordly
for public hearings to serve non-English speaking residents. The NLC notes that Dearborn’s AI
translation services have significantly improved information access for its large Arabic-speaking
population.
Open data and civic intelligence: Washington, D.C. piloted DC Compass, a generative AI
assistant integrated with the city’s open data portal, allowing residents to ask questions about
datasets, generate visualizations, and understand neighborhood-level statistics without
specialized technical knowledge.
Equity and Participation
AI-assisted public engagement tools have genuine potential to improve equity by reaching
residents who cannot attend meetings, do not speak English fluently, or are unfamiliar with
government processes. But they can also replicate or amplify existing inequities if deployment is
not thoughtful.
Digital access remains a prerequisite for many of these tools. Residents without smartphones,
reliable internet, or digital literacy may be excluded from AI-mediated engagement channels.
The city should monitor whether AI tools improve participation among historically
underrepresented groups or consolidate engagement among those already most engaged.
Policy Implications for St. Louis Park
• Any resident-facing chatbot should be grounded in authoritative city documents, updated
regularly, and subject to accuracy review before launch and on an ongoing basis.
• AI-generated responses in public channels must disclose AI involvement and provide a
clear path to human review.
• Meeting accessibility tools should be evaluated for translation quality across the
languages most spoken by St. Louis Park residents.
• The city should monitor whether public-facing AI tools improve participation and access
for underserved residents, not only convenient metrics like total interactions.
• All resident-facing AI applications should be included in a public AI inventory.
Section 4: Public Works and Infrastructure Operations
Public works may be the most operationally mature area for municipal AI because the work is
data-rich, repetitive, and highly visible to residents. Streets, sidewalks, snow removal, sewer
systems, waste collection, traffic signals, and resident-reported issues all generate information
that can be used to detect problems earlier, prioritize work more effectively, and improve service
delivery consistency.
The strongest AI applications in public works are not AI replacing staff. They are tools that help
staff convert messy, high-volume real-world inputs—photos, sensor readings, service requests,
inspection notes, and asset data—into structured information that supports better decisions. The
city still needs public works professionals to validate conditions, authorize work, understand field
constraints, and communicate tradeoffs.
Examples from Peer Cities
San José, CA Vehicle-mounted cameras identify potholes (97% accuracy) and trash/debris
(88% accuracy) in real time. The city is expanding the pilot to bike lanes and
sidewalks, and also captures parking violations and lived-in
vehicles—raising surveillance creep concerns.
Memphis, TN AI analyzes video from city vehicles to identify potholes and property blight,
combining footage with existing data sources to predict areas at risk of
urban decay before residents report them.
Eagan, MN A second-ring Twin Cities suburb with 244 centerline miles of streets uses
LiDAR and AI for pavement condition scoring, producing dashboard-ready
condition maps and data downloads for capital planning.
Cranberry Twp., PA A pavement sensor and forecasting system providing localized 72-hour
pavement temperature forecasts saved $160,000 in winter maintenance
costs by reducing unnecessary road salting.
Montreal, Canada AI analyzes dump truck images to verify snow levels and prevent contractor
fraud without capturing license plates or other identifying information.
Seattle, WA Participates in Google’s Project Green Light, using AI and Google Maps
data to identify inefficient signal timings. Early results show potential to
reduce stops by 30% and CO2 emissions by 10%.
Potential Applications for St. Louis Park
For St. Louis Park, the most relevant public works applications would likely be incremental and
workflow-driven rather than large “smart city” transformations. The city’s existing MySLP/Accela
platform, with 17,288 subscribers and over 6,000 service requests submitted since 2014,
provides a foundation that AI can build on rather than replace.
• Converting resident service requests, photos, and notes into standardized categories,
locations, urgency levels, and routing queues.
• Using camera or inspection data to identify potholes, pavement markings, sidewalk
obstructions, illegal dumping, waste overflow, or snow-related conditions.
• Supporting pavement, sewer, and asset-management planning by combining inspection
data, condition ratings, historical repairs, and GIS.
• Helping field staff retrieve internal knowledge, prior work orders, maintenance history,
and standard operating procedures while in the field.
• Summarizing public works trends for staff and council, such as recurring hotspots,
seasonal patterns, response times, and service-level performance.
Risks and Watch-Outs
Public works AI can appear operationally low-risk but still raises serious governance questions.
Surveillance creep is a particular concern: the same camera network used for pothole detection
can be repurposed for monitoring people, encampments, or behavior. Equity in service
prioritization is another: AI trained on complaint data may systematically underserve
neighborhoods with lower reporting rates. And AI outputs should always be treated as signals
for human review, not final determinations—false negatives in public safety contexts can miss
genuine hazards.
Policy Implications for St. Louis Park
• Start with defined operational problems. Attach AI to specific service goals—faster
pothole identification, improved snow response, better asset planning—rather than
deploying speculatively.
• Prioritize integration before automation. If resident reports, work-order systems, and GIS
are disconnected, AI will add complexity rather than reduce it.
• Require human review for any action. AI can flag, sort, summarize, or recommend; it
should not independently authorize repairs, enforcement, or billing decisions.
• Draw clear surveillance boundaries. Establish policy on what camera networks can and
cannot be used for, and require council approval before expanding use cases.
• Measure service equity. Evaluate AI tools by neighborhood-level response times,
complaint resolution rates, missed issues, and whether benefits are distributed fairly.
Section 5: Planning, Zoning, and Permitting
Planning, zoning, and permitting are among the most promising municipal AI use cases
because the work is complex, document-heavy, rules-based, and highly consequential.
Residents, builders, developers, board members, and elected officials must navigate zoning
ordinances, building codes, application requirements, site plans, variances, setbacks, design
standards, and public hearing processes. These systems are difficult for non-experts to
understand, and delays can affect housing production, economic development, and resident
trust.
AI can help by searching large bodies of rules, identifying missing documents, summarizing
applications, comparing submissions against requirements, and providing clearer guidance to
applicants and staff. The strongest use case is not having AI approve development; it is using AI
to reduce avoidable back-and-forth, improve completeness at submission, and help staff focus
on judgment-heavy issues.
Examples from Peer Cities
Hamilton, ON AI-assisted site plan review for non-profit development projects produced a
60% decrease in review days and 70% decrease in overall processing time.
Kelowna, BC Building permit chatbots launched in 2023 guide homeowners and builders
through permit applications, reducing routine staff questions and extending
service to evenings and weekends.
Austin, TX Beta-testing an AI Pre-Check tool for residential zoning review, allowing
applicants to identify zoning compliance issues before formal submission.
Honolulu, HI CivCheck helps applicants verify code compliance, permitting requirements,
and required documents before submission, reducing review cycles.
Sacramento, CA AI roadmap identifies electronic plan check as a near-term priority to
automate code compliance and design validation.
Los Angeles, CA Planning Department uses AI to analyze land use and zoning data to
proactively identify issues during the permit review process.
Potential Applications for St. Louis Park
The most relevant near-term applications fall into four categories:
1. Applicant self-service. AI can help residents, small businesses, and property owners
understand basic rules such as setbacks, fences, accessory dwelling units, signage, and
parking—and help them determine whether a permit is likely needed—outside of business
hours. Tools must cite authoritative ordinances and provide links to current sources.
2. Submission completeness checks. AI can review application packages for missing
documents, incomplete fields, conflicting information, or obvious inconsistencies before staff
begins formal review.
3. Staff copilots. AI can help staff summarize application materials, prior permits, property
history, relevant ordinances, variances, and public comments, reducing administrative time while
preserving staff judgment.
4. Policy and scenario analysis. AI can help staff, boards, and council summarize how peer
cities handle zoning questions, identify common ordinance language, or model high-level
implications of proposed zoning changes.
Risks and Watch-Outs
Planning and zoning AI carries higher governance risk than public works triage because
land-use decisions affect property rights, housing, and neighborhood change. Due process and
transparency require that residents and applicants understand how AI was used in any decision
affecting them. Incorrect legal interpretation by a chatbot or copilot can cause residents or
applicants to spend money, delay projects, or misunderstand their rights. And tools trained on
historical permitting data may perpetuate patterns of unequal access or enforcement.
A specific technical risk worth noting: applicants could embed hidden or misleading text in PDFs
or plan documents to attempt to manipulate AI summaries or reviews. This “prompt injection”
risk is a real procurement and security consideration for any document-review AI tool.
Policy Implications for St. Louis Park
• AI should not make binding land-use decisions. Staff, boards, commissions, and council
remain accountable for interpretations, recommendations, and approvals.
• AI-generated guidance should cite current ordinances, city webpages, application forms,
and relevant policies. Residents must be able to verify the source of any guidance they
receive.
• Residents and applicants should always have a path to human review. AI should make
the process easier, not create a barrier that residents cannot understand or challenge.
• The city should preserve records of AI use in planning and permitting decisions
consistent with public records requirements.
• Procurement should require vendors to explain what data the tool uses, how it interprets
city rules, how often it is updated, and how the city can test for errors.
• Performance metrics should include application completeness, number of correction
cycles, total review time, applicant satisfaction, and equity of access across applicant
types.
Section 6: Looking Ahead — Agentic AI and the
Evolving Landscape
The AI use cases described in this white paper are largely assistive: AI that flags, summarizes,
translates, or classifies information for human review. The next wave of AI technology—already
emerging in enterprise and government contexts—is agentic AI, in which multiple AI agents
work together to automate and coordinate multi-step tasks with minimal human intervention at
each step.
What Agentic AI Means for Cities
The National League of Cities describes agentic AI this way: “Agentic AI can coordinate the
actions of several business processes, from checking applications for errors or missing
information to validating plans and generating compliance reports and then notifying applicants
of approvals and collecting applicable fees.”
In practical terms, agentic AI could eventually handle workflows that today require multiple staff
handoffs: receiving a permit application, checking it for completeness, routing it to the
appropriate reviewer, generating a compliance checklist, sending the applicant a status update,
and processing payment—all without human intervention at each step.
This is not science fiction. Agentic AI capabilities are being built into enterprise platforms today,
and cities that have strong data foundations and well-defined workflows will be first in line to
benefit. But the governance implications are significantly greater than for assistive AI, because
agentic systems can take consequential actions on behalf of the city without a human reviewing
each one.
Implications for St. Louis Park
The Commission does not recommend that St. Louis Park deploy agentic AI in the near term.
However, the city should be aware of this trajectory and begin laying the governance and data
foundations that will make responsible agentic adoption possible when the time comes.
Specifically, this means:
• Establishing clear workflow documentation so that any AI system—agentic or
otherwise—can be configured against defined, auditable processes rather than informal
practices.
• Ensuring that AI governance frameworks explicitly address autonomous action: under
what circumstances can an AI system take an action without human review, and who
bears accountability when it errs?
• Maintaining a human-in-the-loop requirement for all decisions that affect resident rights,
benefits, or obligations until the city has sufficient experience, audit capability, and public
trust to consider exceptions.
• Watching peer cities that are piloting agentic workflows in lower-stakes
contexts—scheduling, document routing, status notifications—to learn from their
experience before adopting similar tools.
External Resources and Learning Networks
St. Louis Park does not need to navigate this landscape alone. The following networks and
resources provide valuable guidance, peer learning, and policy models:
National League of Cities (NLC). NLC’s AI in Cities Report and Toolkit, City AI Governance
Dashboard, and AI Policy Dashboard provide case studies, readiness assessment tools, and
examples of city AI policies from across the country.
League of Minnesota Cities (LMC). LMC has published guidance on AI use consistent with the
Minnesota Government Data Practices Act, and its frameworks for data classification and risk
management are already referenced in St. Louis Park’s internal guidelines.
MNIT (Minnesota IT Services). MNIT guidance on AI risk classification and data governance
provides a state-level framework that St. Louis Park can align with.
Bloomberg Philanthropies / City AI Connect. A global learning community for cities exploring
generative AI in public services, providing peer exchange, expert access, and a shared
resource hub.
AI4Cities (European). A collaborative project involving Helsinki, Amsterdam, Copenhagen,
Paris, Stavanger, and Tallinn using pre-commercial procurement to develop AI solutions for
mobility and energy, offering useful models for collaborative AI procurement.
Global Observatory of Urban Artificial Intelligence (GOUAI). Supported by UN-Habitat,
GOUAI maintains the Atlas of Urban AI, a repository of ethical AI initiatives and projects from
cities worldwide.
Conclusion: Toward an AI-Compatible City
The question this white paper set out to answer is how cities need to adapt to become
AI-compatible—not just technically, but in terms of governance, culture, and policy. The
Commission’s answer, drawn from peer city experience and national guidance, is that AI
compatibility is less about adopting any particular technology and more about building the
underlying conditions that make AI trustworthy and useful.
Those conditions include: clean, integrated data that AI systems can act on reliably; clear
governance structures that assign human accountability for every AI system; staff who
understand both the capabilities and the limitations of the tools they use; procurement practices
that require vendors to be transparent and auditable; and a commitment to measuring whether
AI improves service equity, not only efficiency.
St. Louis Park is better positioned than many peer cities because it has already established an
internal AI use policy, operates an existing resident service platform in MySLP, and has a
Community Technology Advisory Commission actively engaged in these questions. The work
ahead is to extend and deepen that foundation—incrementally, measurably, and with residents
at the center.
The Commission recommends that the city:
• Formalize a public AI inventory and governance process, requiring council review for any
AI tool that affects resident rights or services.
• Conduct an AI readiness assessment covering data quality, workforce capability, system
integration, and cybersecurity posture before adopting new AI tools in any department.
• Prioritize internal productivity tools and low-risk pilot applications in the near term, using
those experiences to build institutional knowledge before moving to higher-stakes use
cases.
• Establish equity measurement as a non-negotiable component of any AI evaluation, with
regular reporting to council on whether AI tools are improving or worsening service
equity across neighborhoods.
• Stay connected to the NLC, LMC, and peer city networks to monitor how the
technology—and the governance approaches to it—continue to evolve.
AI will not transform city government on its own. But cities that invest now in the governance,
data, and human capacity to use AI responsibly will be better positioned to deliver more
efficient, equitable, and transparent services to their residents in the years ahead.
Appendix: Key Resources and References
National and State Guidance
• National League of Cities — AI in Cities: Report and Toolkit (2024)
• National League of Cities — City AI Governance Dashboard:
nlc.org/resource/city-ai-governance-dashboard/
• National League of Cities — Use AI to Transform City Operations (2025):
nlc.org/article/2025/07/31/use-ai-to-transform-city-operations/
• League of Minnesota Cities — AI guidance aligned with Minnesota Government Data
Practices Act
• MNIT — Minnesota AI risk classification guidance
• NIST AI Risk Management Framework: nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf
Peer City Policy Documents
• City of Seattle — Generative Artificial Intelligence Policy
• City of Boston — Interim Guidelines for Using Generative AI (2023)
• City of San Jose — AI Handbook / Digital Privacy Office (2024)
• City of Tempe — Ethical Artificial Intelligence Policy (2023)
• City of Lebanon, NH — AI Registry
• New York City — Artificial Intelligence Action Plan (2023)
• City of Chattanooga — AI Prompt Library
Selected Case Study Sources
• Sacramento 2025–2028 AI Strategy and Roadmap
• Hamilton, ON — Bloomberg Harvard City Program All4One Pilot
• APWA / Eagan, MN — AI for Pavement Assessment Case Study
• Google Research — Project Green Light: sites.research.google/greenlight/
• Bloomberg Philanthropies / Johns Hopkins — City AI Connect: cityaiconnect.jhu.edu
• Global Observatory of Urban Artificial Intelligence — Atlas of Urban AI:
gouai.cidob.org/atlas/
• Eurocities — Algorithmic Transparency Standard: algorithmregister.org
This white paper was prepared by the St. Louis Park Community Technology Advisory
Commission. It draws on publicly available municipal policy documents, the NLC’s AI in Cities
Report and Toolkit, the Sacramento and Hamilton AI strategy documents, and reporting from
Government Technology, Smart Cities Dive, and other municipal technology publications.
Community technology
advisory commission
Presented by: Konnor Slaats and Rudyard Dyer
Staff Liaison: Jacque Smith
Policy question
Does the council have any work direction for this commission?
Workplan summary
Completion
Timeline
Status Update Workplan Item
4Q 2026In process. Statistics and information shared by staff with commission members for their
assessment. Discussion continues. Some information was included in the AI whitepaper.
Assess the city’s customer
response management tool
July 2026Completed and shared with staff liaison for use in communications and technology
department
Produce a whitepaper on the
future of generative AI for cities
Sept 2025CTAC members participated in user experience workshops and provided comments on user
feedback, wireframe and design. Staff liaison has provided updates on redesign process.
Participate in city website
redesign review process
December 2025Commission members received information from the community engagement coordinator
throughout the Vision 4.0 process; including survey invites; a commission member hosted a
community meeting; commission members reviewed Dec. 8, 2025, council report to inform
CTAC workplan goals for 2026
Support citywide Vision 4.0
process
OngoingOngoingStay informed about pending
legislation affecting cable and
technology