Top 10 AI Prompts and Use Cases and in the Government Industry in San Francisco

San Francisco's city services face rising resident expectations and tight budgets, so clear, well-crafted AI prompts matter: they turn bulky datasets and citizen requests into fast, actionable answers, streamline operations, and free staff for complex work.

Research shows local governments use AI to automate routine tasks, improve asset management, and boost resident engagement - tools that can, for example, shrink manual review times from hours to minutes and help predict infrastructure needs before problems cascade (see CivicPlus's overview of AI in local government).

Responsible prompt design also supports transparency and accountability - fact‑checking outputs, disclosing AI use, and protecting private data are now standard guidance from the National League of Cities.

For San Francisco teams ready to build practical skills, Nucamp's AI Essentials for Work bootcamp: learn workplace AI and prompt writing teaches prompt writing and workplace AI use so staff can safely deploy copilots, chatbots, and predictive models that actually improve services residents rely on.

Program	Length	Cost (early bird)	Registration
AI Essentials for Work	15 Weeks	$3,582	Register for the AI Essentials for Work bootcamp

Selection prioritized use cases that produce clear operational wins for California public teams, align with emerging legal and regulatory expectations, and can be adopted with existing staff training paths - for example, practical tools that support predictive traffic safety analytics or approved enterprise copilots.

Prompts were chosen for measurable payoff (work that reduces repetitive review), for legal defensibility (informed by sessions on governance and California's role at the UC Berkeley Law AI Institute programs on AI law and policy UC Berkeley Law AI Institute programs on AI law and policy), and for teachability (prompt engineering and risk mitigation are core modules in Berkeley's Berkeley Law Generative AI for the Legal Profession course (3 MCLE hours) Berkeley Law Generative AI for the Legal Profession course (3 MCLE hours), a self‑paced course that covers hallucination, confidentiality, and even offers 3 MCLE hours).

Accessibility and cost were also factors - programs offering government/nonprofit rates or short time commitments signal realistic upskilling pathways for city teams.

The result is a top‑10 list that balances practical impact, legal accountability, and a path to rapid staff adoption - yes, even earning MCLE credit in under five hours to jump‑start trustworthy deployment.

Criterion	Evidence
Operational impact	Predictive traffic safety analytics case study for San Francisco government
Legal & regulatory alignment	UC Berkeley Law AI Institute sessions on AI law and California policy
Trainability & credentials	Berkeley Law Generative AI for the Legal Profession course - self‑paced, under 5 hours, 3 MCLE credits

“If you've been thinking about how to apply generative AI into your work in a responsible way, Berkeley Law Executive Education's Generative AI for the Legal Profession course is the ideal first step.” - Miles Palley

San Francisco legal teams can get practical, time‑boxed training on bringing generative AI into day‑to‑day practice with Berkeley Law Executive Education's skills‑based

Generative AI for the Legal Profession

offering - a self‑paced, ~3‑hour course designed to teach what generative AI is and how to integrate it responsibly (perfect for busy government counsel juggling procurement reviews and privacy checks), and it sits alongside Berkeley's longer contract and ethics offerings that include dedicated modules on AI and commercial contracts.

For city departments in California that must balance innovation with disclosure, confidentiality, and defensible decision‑making, this short course is a clear upskilling route that complements deeper programs like the Commercial Contract Fundamentals course (which explicitly addresses AI & commercial contracts) and other Berkeley executive programs aimed at in‑house and public sector lawyers.

Explore the program details and scheduling on Berkeley Law's executive education site to match training windows with departmental learning plans and MCLE needs.

Program	Format	Start Date	Time Commitment
Berkeley Law Executive Education - Generative AI for the Legal Profession program (Self‑Paced)	Self‑Paced Online	Start Date: Feb 3, 2025	~3 hours

Riverbed's Riverbed IQ Ops brings AI-driven observability to life for California public IT teams - announced from Redwood City, Calif., the platform combines generative, predictive, and agentic AI to surface root causes, recommend fixes, and automate remediations so network blips don't turn into service‑wide outages; Riverbed reports observability bookings up 92% in H1 2025 and customers running over 64 million AI remediations annually, signaling real operational ROI. For San Francisco departments juggling hybrid cloud, remote endpoints, and Zero Trust requirements, IQ Ops' SaaS‑based Riverbed IQ (no extra infrastructure required) pairs Smart OTel precision telemetry with a Topology Viewer and role‑based workspaces to speed triage, integrate with ITSM like ServiceNow, and deliver low‑code automations that free staff for higher‑value work.

Explore the product details on Riverbed's AIOps product page or read the Riverbed press release for the AI‑driven observability launch to see how a unified platform can move teams from firefighting to predictive assurance.

Capability	Why it matters for San Francisco teams
Riverbed IQ Assist product page (Generative AI for observability)	Instant, context‑rich root‑cause insights to reduce mean time to repair
Predictive AI	Flags anomalies from real‑time and historical telemetry before they impact services
Agentic AI (low/no‑code)	Orchestrates safe, auditable automations to accelerate remediation
Smart OTel & Topology Viewer	Delivers precise telemetry and map‑based correlation across cloud, edge, and on‑prem

“With today's launch, we're unveiling our next‑generation Riverbed xx90 systems… packaged and integrated with Riverbed IQ for AI‑driven insights.” - Dave Donatelli, CEO of Riverbed

Glean Assistant turns scattered city data into a single, permission‑aware knowledge layer so San Francisco teams can find trusted answers in seconds - whether it's procurement rules, vendor contracts, or legacy SOPs tucked away in drives and Slack.

Built on a retrieval‑augmented approach with over 100 connectors and admin controls, Glean surfaces citation‑backed responses, creates shareable Collections and memorable Go‑Links, and can embed custom AI agents to automate routine workflows; customers report big wins (Duolingo saves 500+ hours/month and new hires shave roughly 36 hours off ramp time).

For city IT and policy shops that must balance speed with auditability and privacy, Glean's enterprise features (real‑time permissions sync and SOC‑2 / GDPR‑grade controls) help reduce time wasted on searching while keeping sensitive records protected - explore Glean's product overview or read the enterprise knowledge management guide to see how a grounded assistant can move teams from digging through silos to decision‑ready answers.

Glean Capability	Why it matters for San Francisco teams
Glean answers and intelligent search for knowledge management	Delivers role‑aware, citation‑backed responses to speed casework and policy review
Collections & Go‑Links for curated resource access	Curates scheduling, permit, and procurement resources for consistent staff access
Retrieval‑augmented generation (RAG) with permissions	Grounds outputs in city data while enforcing access controls and auditability

City ordinance language that enables data sharing while staying CPRA‑compliant should be pragmatic, auditable, and resident‑facing: require clear “notice at collection” and updated privacy policies that list categories of personal and sensitive data, purposes, retention criteria, and conspicuous opt‑out links (think a public map that ties every dataset to a labeled “Do Not Sell or Share” switch).

Recent CPPA action - finalizing draft regulations on automated decision‑making, risk assessments, and cybersecurity on July 24, 2025 - means local drafters must bake in data mapping, periodic risk assessments, and mechanisms to honor opt‑outs and limit sensitive data uses (see Foley LLP summary of CPPA draft regulations).

Practical drafting steps include aligning municipal privacy notices with the five CPRA notice types (privacy policy, notice at collection, opt‑out, limit‑use, financial incentives), embedding two or more consumer request channels, and building contract clauses that bind vendors to deletion, notification, and processing limits (Practical Law guidance on CCPA/CPRA notices is a helpful template).

Time horizons matter: several regulatory provisions phase in over 2026–2027, so ordinances should set interim compliance checkpoints and a clear schedule for the city's first formal CPRA risk assessment and vendor audits.

Action	Key Date / Note
CPPA finalized draft regulations	Foley LLP summary of CPPA draft regulations (July 24, 2025)
Some provisions effective	As early as January 1, 2026
ADMT compliance start	Businesses using ADMT need not comply until January 1, 2027
First formal risk assessment due	By December 31, 2027 (for post‑effective processing)

For San Francisco Purchasing, an AI‑powered RFP summarizer can turn procurement from a triage nightmare into a predictable workflow: feed in a 100+‑page solicitation and get a concise, compliance‑mapped action sheet that highlights must‑meet requirements, deadlines, and follow‑up questions so reviewers focus on judgment, not skimming.

Simple interventions like the NASCIO / U.S. Digital Response recommendation to include summary sheets pair naturally with generative tools that extract key requirements and draft compliant responses (see the U.S. Digital Response guide on improving RFPs with summary sheets), while platforms such as Inventive AI demonstrate how automated requirement extraction and starter drafts speed proposal work without removing human oversight.

For capture teams and smaller vendors who rely on quick signals, RFP‑Engine and similar tools produce one‑page GPT summaries and scoring to triage opportunities faster - a single readable sheet can be the difference between chasing the right contract and wasting weeks on a longshot - and it keeps the city's procurement process fair, auditable, and faster for everyone involved.

AI Feature	Why it helps San Francisco Purchasing
U.S. Digital Response guide to RFP summary sheets for government procurement	Make dense RFPs readable for smaller vendors and speed initial vetting
Inventive AI government proposal drafting and requirement extraction with AI	Automates compliance mapping and starter content so staff focus on strategy
RFP-Engine AI one-page RFP summaries and scoring for faster triage	Triages opportunities quickly and supports data‑driven bid/no‑bid decisions

This combination of clear summary sheets, automated requirement extraction, and concise AI-generated triage outputs helps San Francisco Purchasing improve accessibility for smaller vendors, speed internal review cycles, and maintain auditability across procurement decisions.

When 911 systems, traffic signal controllers, or remote CCTV feeds hiccup, incident runbook automation turns panic into a predictable, auditable response: platforms like PagerDuty Runbook Automation platform for automated incident runbooks let public safety teams codify playbooks that detect, triage, and execute pre‑approved remediation steps - from automated diagnostics to remote device restart and reconnect at the edge - so outages are contained in seconds rather than hours.

Best practices from incident response research emphasize starting small (triage, data collection, low‑risk remediations), building clear roles and playbooks, and measuring MTTR/MTTD improvements to guide expansion, all of which align with city priorities for reliability and staff wellbeing (see a practical incident response automation implementation guide and Atlassian's incident playbook template for running and training playbooks).

For San Francisco's constrained ops teams, the payoff is tangible: fewer false alarms, less on‑call burnout, and consistent, auditable action paths that keep residents safe while engineers focus on preventing the next incident rather than fighting the last one - imagine an automated runbook rebooting a streetside sensor and restoring data feeds before a dispatcher even picks up the phone.

Capability	Why it matters for San Francisco public safety
Event‑driven automation	Immediate, context‑rich responses reduce MTTR and limit public impact
Edge automation (device restart/reconnect)	Remediates remote sensors and cameras without on‑site crews
Runbook Runners & secure execution	Execute jobs behind firewalls with encrypted results and audit trails
AI‑generated runbook starter	Speeds creation of vetted automation jobs while preserving human oversight

A topology viewer that spans on‑prem, cloud, and edge systems turns scattered telemetry into a single, navigable map so public‑safety teams see service dependencies the instant something goes wrong - think a live diagram tying a cloud‑hosted CAD instance to its on‑prem database and the streetside location feeds that inform dispatch.

Tools like OpsRamp's Topology Explorer visualize dependency relationships across containers, hypervisors, networks and public clouds to speed root‑cause analysis, while LogicMonitor's layer‑2/3 mapping and protocol‑aware discovery lets operators trace exactly which link or device is affecting traffic flows.

For 9‑1‑1 centers and EOCs, purpose‑built mapping (GeoComm Maps) adds public‑safety location intelligence and indoor maps, delivering the right location data to the right user without workstation installs.

Combined, these capabilities reduce mean time to repair, tighten security and compliance by reducing blind spots, and turn frantic incident triage into an auditable, repeatable workflow - imagine a map that immediately highlights the cascading dependency so engineers can isolate the fault before an outage spreads to other emergency services.

Capability	Benefit for San Francisco public safety
OpsRamp hybrid cloud topology maps for real-time dependency visualization	Faster root‑cause analysis across hybrid apps and infrastructure
LogicMonitor layer-2 and layer-3 topology mapping with protocol-aware discovery	Precise network path and device mapping to troubleshoot connectivity
GeoComm public-safety maps and Database-Backed Home Location (DBHL)	Location intelligence and indoor maps tuned for ECCs and EOCs

A Board of Supervisors meeting summarizer turns dense proceedings into an actionable, auditable record so San Francisco departments can move from discussion to delivery: capture the meeting title, date, time and location, check off attendees and roles, note quorum and votes, and distill each agenda item into decisions, assigned action items and deadlines - then distribute the draft within the recommended 24–48 hours to speed follow‑up and corrections (see iBabs meeting minutes best practices and GBQ guidance on the appropriate level of detail for minutes).

Grounded summaries should exclude off‑record chatter, flag sensitive items for closed session handling, and attach or reference supporting reports so auditors and the public can verify outcomes.

AI helpers that transcribe and generate topic‑based notes (for example, Jamie's assistant) can save the minute taker hours and create searchable archives that answer “what did we agree?” in seconds, turning a long meeting into a concise action list that gets work started before the week is out.

“It is a time saver. I am able to do minutes during the meeting from the online agenda packet.”

San Francisco agencies drowning in constituent email can scale a smart routing layer with agentic workflows that read an incoming message, fetch context from city records, and either draft a reply or push a ticket into the CRM - automatically tagging urgency and the right office so staff handle judgment calls instead of triage.

Built examples like Glean's prospect outreach agent show how agents trigger on new leads or inputs, gather account and document context, identify stakeholders, and produce tailored messages, and the broader Glean Agents platform makes it practical to wire those steps into enterprise systems with permissions-aware search and connectors for Gmail, Outlook, and Drive; the result is faster, auditable routing and fewer missed deadlines.

For San Francisco teams this means routing urgent constituent requests to the correct caseworker before the first shift starts, reducing backlog and preserving a clear audit trail for transparency and oversight - automation that hands staff the work that needs human judgment, not the inbox sifting.

Learn more about the agent workflow in Glean's prospect outreach agent and Glean's agent platform for enterprise deployments.

Capability	How it helps San Francisco government
Glean prospect outreach agent trigger activation for email and CRM	Automatically starts routing when a new message or lead arrives to eliminate manual triage
Connectors (Gmail, Outlook, Drive)	Keeps routing and context in sync with existing email and records systems
Contextual routing & RAG grounding	Uses document and account context to assign the right department and draft citation-backed responses
Permissions & security (Glean Protect)	Maintains access controls and audit logs so sensitive constituent data stays protected

“Glean helps you get work done, rather than just find information. The moment we launched Glean, there was so much positivity.”

A public‑facing FAQ for San Francisco's permitting portals should turn technical governance into plain language: explain when an AI component reviews an application, what data it uses, how residents can consent or request human review, and what steps the city takes to check for bias and protect privacy - transparency that builds trust instead of confusion.

Ground the FAQ in clear explainability, accountability, and data‑minimization principles (the ISO guidance on responsible AI is a practical blueprint), and include concrete links to documentation, audit logs, and appeal paths so outcomes are verifiable.

Describe bias‑mitigation and testing routines, cite lifecycle and post‑deployment monitoring practices used for regulated systems (see FDA recommendations on transparency and bias for AI-enabled functions), and offer an accessible one‑line summary on each page plus a “learn more” link for technical readers.

In practice, a short, visible line on the permit status page - e.g., “This decision used an AI assistant; request a human review” - can defuse concerns and speed resolution, turning opaque automation into a transparent civic service that residents can understand and challenge if needed (see Zendesk's practical guide to AI transparency for customer‑facing systems).

San Francisco teams can turn the Top 10 use cases in this guide into reliable, auditable services by pairing practical guardrails with focused upskilling: follow the City's July 2025 Generative AI Guidelines - use only approved enterprise tools (Copilot Chat is approved), never drop sensitive data into consumer models, document public‑facing AI in the 22J inventory, and always fact‑check outputs (San Francisco Generative AI Guidelines for municipal AI use).

Start with low‑risk pilots that free time for higher‑value judgment (pilot evidence at scale shows measurable hourly savings), then layer governance and procurement alignment as federal policy and funding shift under the new national AI Action Plan - coordinate local rollout with emerging federal procurement and infrastructure directives to keep legal and operational risks in check (White House AI Action Plan overview and executive orders summary).

Practical next steps: map priority processes for automation, require human review on any decision that affects residents, and build a training pipeline so staff know how to write and vet prompts - Nucamp's AI Essentials for Work bootcamp offers a 15‑week, job‑focused path to prompt writing and safe tool use for nontechnical staff (Nucamp AI Essentials for Work registration and syllabus).

The combination of clear rules, small wins, and targeted training turns responsible experimentation into sustained service improvements - imagine an approved prompt routing an urgent constituent request to the right team before the next shift starts.

Next Step	Why it matters	Source
Start small with low‑risk pilots	Proves value and limits exposure	Scaling AI pilots evidence
Governance & disclosure (22J)	Maintains public trust and auditability	San Francisco Generative AI Guidelines for municipal AI use
Train staff in prompt writing	Ensures safe, effective deployments	Nucamp AI Essentials for Work registration and syllabus