Top 10 AI Prompts and Use Cases and in the Government Industry in Boulder

Boulder city and county teams can accelerate service improvements by following Colorado's living Guide to Artificial Intelligence, which pairs a statewide GenAI policy and OIT intake for risk assessments with an AI Community of Practice to share best practices (Colorado Guide to Artificial Intelligence - OIT AI Guide); a 90‑day Google Gemini pilot - 150 participants across 18 agencies - reported concrete productivity gains (74% of participants) while underscoring the need for training, attestations and NIST‑aligned risk reviews (Google Gemini pilot case study - Colorado OIT).

For Boulder staff and contractors, pairing these governance guardrails with practical reskilling - such as Nucamp's AI Essentials for Work - offers a clear path to capture GenAI benefits while managing bias, privacy and security risks (AI Essentials for Work syllabus - Nucamp).

Program	Length	Early Bird Cost	Syllabus
AI Essentials for Work	15 Weeks	$3,582	AI Essentials for Work syllabus - Nucamp

"Gemini has saved me so much time that I was spending in my workday, doing tasks that were not using my skills."

Selection prioritized prompts and use cases that Colorado's rollout and national best-practice training show deliver measurable, responsible value: each candidate had to map to a clear government workflow (policy drafting, communications, case triage, data analysis, training, procurement, etc.), align with NIST-informed ethical criteria and attestation/training requirements, and be feasible with existing tools and procurement paths; evidence from Colorado's 90‑day Google Gemini pilot (150 participants across 18 agencies, with reported productivity and quality gains) and InnovateUS's practical GenAI courses informed what “workable” looked like in practice (Colorado Gemini pilot case study - Colorado Office of Information Technology, InnovateUS GenAI workshop series for the public sector).

Prompts were ranked by (1) demonstrable impact or pilot signal, (2) low risk of exposing sensitive data with mitigations available, (3) accessibility and inclusivity benefits, and (4) ease of staff upskilling via short, repeatable training - so the final Top 10 is tightly tethered to both Colorado's empirical pilot results and ready operational controls for Boulder teams to adopt.

Methodology Criterion	Example from Sources
Pilot evidence & metrics	90‑day Gemini pilot - 150 participants, 18 agencies (Colorado OIT Gemini pilot case study)
Training & attestations	Required responsible‑AI training and attestations (InnovateUS GenAI public sector courses)

"Gemini has saved me so much time that I was spending in my workday, doing tasks that were not using my skills."

Use the prompt "Summarize and flag risks" to turn dense policy drafts, contracts or proposed ordinances into a focused risk memo that highlights renewal dates, indemnity and confidentiality clauses, inconsistent obligations, and likely privacy or bias concerns - capabilities already demonstrated by AI contract‑review tools (AI contract review use cases - Texas Bar Practice).

In Boulder, route any GenAI‑assisted summary through Colorado's required controls: run the OIT/NIST‑aligned risk assessment, avoid entering non‑public or sensitive data into models, and record human verification steps so outputs meet state GenAI prohibitions and guardrails (GenAI Risks & Considerations - Colorado OIT).

When a policy or procedure touches “consequential decisions,” retain documentation and impact assessments to satisfy Colorado's legal duties under the CAIA (disclosures, appeals and recordkeeping), and always treat AI results as draft findings that require human editing before release (Colorado Artificial Intelligence Act - NAAG deep dive).

The practical payoff: AI can surface contract red flags (e.g., indemnities or auto‑renewals) in seconds, but without OIT intake and documented human review those gains risk noncompliance.

Prompt	Typical AI Output	Required Colorado Safeguard
"Summarize and flag risks"	One‑page risk memo with flagged clauses (renewals, indemnities, privacy)	OIT NIST‑aligned risk assessment; no non‑public data; human verification and retention of impact assessment

“If we didn't come forth with a product, people are going to be using it anyway. And there's danger in people actually using applications that are not part of your enterprise.”

Prompt: "Create alt text and plain‑language summaries" turns dense agency content into accessible, usable communications for Boulder residents - generate two alt‑text variants (concise for lists, extended for screen readers), include scene context (who, what, location, action) and any policy‑relevant timestamps, and pair each image description with a 1–2 sentence plain‑language summary that answers “what this means for me” and a short list of next steps; Colorado's pilot shows these outputs improve inclusion in practice - about 20% of pilot participants identified as having accessibility needs and reported clearer communication and greater confidence - so always route AI suggestions through the state's training and attestation workflow and require human verification to avoid disclosing non‑public data (Colorado pilot report on responsible AI in government - InnovateUS) and align with local responsible‑AI guidance for Boulder teams (Nucamp AI Essentials for Work - responsible AI practices for Boulder governments); the practical payoff: a single verified alt‑text + plain‑language pair can turn a complex permit page into a clear call to action that reduces help‑desk calls and speeds constituent outcomes.

Metric	Value / Finding
Pilot participants with accessibility needs	~20%
Use cases analyzed in pilot	Over 2,000
Reported accessibility benefits	Improved communication & enhanced workplace confidence

“I almost felt like I could see it.”

Summarize reports and extract action items

turns long meeting transcripts, committee reports or multi‑page vendor briefs into a concise, verifiable to‑do list that staff can assign and track; Colorado teams can pair Zoom's AI Companion or Teams Copilot - both able to produce Quick Recaps, summaries and

next steps

or capture conversation points - with human review and the state's OIT/NIST intake to keep sensitive data protected (Approved AI transcription tools and best practices from CU Anschutz, Zoom AI Companion meeting summaries and action items - CU OIT).

Verify outputs before circulation - accuracy checks and participant confirmation are mandatory per university and legal best practices - because AI summaries can be fast but imperfect (one practitioner used an LLM to draft a hearing memo in under five minutes, then validated quotes and actions manually) (Example: LLM hearing memo by Marci Harris).

The practical payoff for Boulder: a vetted AI summary converts lengthy records into clear assignments and deadlines so teams spend less time chasing context and more time delivering constituent outcomes.

Prompt	Typical AI Output	Required Colorado Safeguard
Summarize reports and extract action items	One‑page summary with bullet action items, owners, and suggested deadlines	Use approved transcription tools; OIT/NIST risk intake; human verification before distribution

Prompt: "Draft responses and triage cases" can speed constituent-facing work by generating clear reply drafts, suggested next steps, and a triage score that prioritizes urgent housing, benefits, or emergency-related requests; but Colorado's playbook requires controls before hitting “send.” Any AI‑assisted reply that affects a resident's rights or access must follow the State OIT intake and NIST‑aligned risk assessment, avoid entering non‑public or sensitive data into external models, and include a consumer notice when an AI system is interacting with the public (Colorado OIT guide to artificial intelligence and public sector guidance).

Under SB24‑205 deployers must provide plain‑language disclosures, let individuals correct data, and - when AI is a substantial factor in a consequential or adverse decision - offer an appeal and human review; noncompliance exposes organizations to enforcement by the Colorado Attorney General (and potential penalties) so verify outputs, log human verification steps, and retain impact assessments before relying on AI in case handling (Colorado SB24‑205 (AI Act) full text, NAAG analysis of the Colorado AI Act).

The practical payoff: a verified AI draft plus a documented human review can cut response time while preserving residents' rights and the city's legal footing.

Prompt	Required Colorado Safeguard	Key Consumer Right
"Draft responses and triage cases"	OIT/NIST risk intake; no non‑public data in external models; human verification and retained impact assessment	Notice of AI use; correct personal data; appeal/human review for adverse consequential decisions

The act requires a developer of a high-risk artificial intelligence system (high-risk system) to use reasonable care to protect consumers.

Explain data trends and suggest hypotheses

Prompt: The quoted prompt above turns raw time‑series, permit volumes or meter reads into plain‑language trend summaries, ranked candidate explanations, and a short list of testable hypotheses that staff can validate with local datasets; pair those AI outputs with Boulder's practice of working across utilities to identify best practices and test cost assumptions so suggested hypotheses feed directly into operational pilots (Boulder utilities best‑practice coordination - Dec. 11, 2012 study session).

To preserve transparency and legal compliance, route AI findings through responsible‑AI controls and human verification - practices that keep explanations accessible to nontechnical council members while preserving audit trails and safeguards (Nucamp AI Essentials for Work syllabus - Responsible AI practices for Boulder governments, Nucamp AI Essentials registration - Anticipating local AI trends and impacts in Boulder).

The practical payoff: AI narrows the hypothesis space so analysts spend less time hunting for signals and more time validating the few most promising explanations with partners and real cost data.

Prompt turns repetitive correspondence - permit acknowledgments, vendor onboarding steps, resident follow‑ups - into vetted, consistent templates and stepwise triage checklists that staff can copy, customize and store in a central hub; link those templates to low‑code automation (Microsoft Copilot Studio) or Google Workspace's Gemini to populate fields, route approvals and log human verifications while preserving audit trails.

Colorado's 90‑day Gemini pilot showed concrete productivity gains (74% of participants reported increased productivity and 31% freed time to upskill), so automating routine emails can realistically return hours to higher‑value policy and constituent work rather than create new risk.

Follow the state playbook: require responsible‑AI training and attestations, run OIT/NIST risk intake before rollout, avoid entering non‑public data into external models, and respect platform limits (Copilot Studio agents use a shared monthly quota and are not recommended for campus‑critical functions).

Use these controls to standardize communication, reduce help‑desk volume and keep legal compliance front and center (Google Gemini pilot case study and findings on productivity, Microsoft Copilot Studio low-code automation overview and documentation, Colorado Guide to Artificial Intelligence and responsible AI guidance).

Prompt	Typical Integration	Colorado Safeguard / Note
"Draft email templates and checklists"	Gemini in Google Workspace or Copilot Studio agents to populate templates and route approvals	Required OIT/NIST risk intake, training/attestations; no non‑public data; be mindful of Copilot Studio quotas

"Gemini has saved me so much time that I was spending in my workday, doing tasks that were not using my skills."

Prompt:

"Analyze vendor contracts and flag sensitive data"

Turns routine procurement review into a targeted compliance check for Colorado's SB24‑205 landscape - scan for missing developer disclosures about training data, absent commitments to provide the documentation and impact‑assessment materials that create a rebuttable presumption of “reasonable care,” weak reporting/notification timelines, and contract terms that permit sharing or retention of sensitive personal data.

With SB24‑205 effective Feb. 1, 2026 and enforcement reserved to the Colorado Attorney General, this prompt should surface clauses that would force a deployer to disclose consumer data or leave the city unable to obtain the developer documentation required for compliance (e.g., summaries of training data, bias‑mitigation steps, and evaluation results); vendors should be contractually required to notify known deployers and the AG within 90 days of discovering algorithmic discrimination and to support NIST‑aligned risk management.

A practical, memorable step: add minimum deliverables to every AI vendor SOW now - training‑data provenance, bias‑testing reports, and a timely‑notification commitment - so Boulder teams avoid last‑minute renegotiations and reduce exposure to enforcement or operational delay (Colorado SB24‑205 full text and compliance duties, NAAG analysis of Colorado's Artificial Intelligence Act).

Prompt	What to Flag	Contract Language to Require
"Analyze vendor contracts and flag sensitive data"	Missing training‑data summaries, no bias/testing results, broad data‑sharing/retention rights, no 90‑day notification clause	Deliverables: training‑data provenance; bias‑mitigation & evaluation reports; 90‑day AG/deployer notification; NIST‑aligned risk management cooperation

"The act requires a developer of a high-risk artificial intelligence system (high-risk system) to use reasonable care to protect consumers."

Prompt: "Create training modules and attestations" helps Boulder teams turn policy into practice by auto-generating a short, role‑based curriculum (two‑hour self‑paced core course plus modular virtual sessions), an integrated attestation workflow, and the surveys and office‑hours schedule needed to measure uptake and risk posture; Colorado's pilot used exactly this model - branded courses in the state LMS with a Governor intro video and an attestation, mandatory completion for pilot participants, and regular surveys that fed program metrics - 150 testers across 18 agencies in a 90‑day pilot that reported clear benefits (e.g., more staff focusing on higher‑priority work) (Responsible AI for Public Professionals - InnovateUS course, Colorado pilot report - InnovateUS).

Practical next steps: scaffold modules by role, bake attestations into the LMS video or onboarding flow, and capture pre/post metrics so training becomes an auditable control, not just a checkbox.

Training Element	Example from Colorado Pilot
Core course length	Two hours (self‑paced)
Attestation	Mandatory, integrated into introductory video in state LMS
Pilot scale	150 participants; 18 agencies; 90‑day pilot
Support	Virtual sessions + regular office hours + surveys

“If we didn't come forth with a product, people are going to be using it anyway. And there's danger in people actually using applications that are not part of your enterprise.”

Prompt:

Generate code snippets and test cases

turns a written feature spec into concrete, runnable artifacts - most simply, unit tests that codify acceptance criteria.

For example, an AI can be prompted to “Generate Python unit tests for the user login feature (requirements listed below),” a pattern shown in Safe and Scalable AI‑Assisted Development in Digital Health that accelerates test coverage without starting from scratch (Safe and Scalable AI‑Assisted Development in Digital Health - Firefly Innovations).

In Boulder projects, pair every AI-generated snippet or test suite with documented human verification and local execution, and fold the output into existing QA and change‑management workflows as part of responsible deployments - practices and training available for local teams help maintain accessibility, transparency and legal compliance (AI Essentials for Work bootcamp syllabus - Responsible AI practices for workplace teams).

The concrete payoff: a verified AI test suite becomes a reproducible artifact that proves a feature works and shortens the manual debugging loop for busy municipal developers and contractors.

Prompt: "Summarize pilot survey responses and propose rollout plans" converts the standing survey data into clear, operational decisions by distilling the Gemini pilot's >2,000 survey entries into signals (participation, weekly cadence, productivity and accessibility outcomes), recommending phased rollout triggers (e.g., confirmed training completion, attestations, and consistent productivity signals) and sizing staffing for support cohorts and communications.

Use survey cadence (participants agreed to report at least three times weekly), Community of Practice attendance and hub analytics to estimate ongoing support needs and ROI; Colorado's 90‑day Gemini pilot (150 participants across 18 agencies) provides concrete benchmarks - 74% reported increased productivity, 83% better work quality and 75% enhanced creativity - that Boulder teams can use as go/no‑go evidence while preserving OIT/NIST risk intake, mandatory training and attestations before expansion.

For practical templates and survey structure, consult the Colorado OIT Gemini pilot case study and InnovateUS's pilot report to map metrics to a phased communication and rollout toolkit.

Metric	Pilot Result
Participants	150 (18 agencies)
Survey submissions	Over 2,000 (standing survey, 3x/week)
Reported productivity increase	74%
Improved work quality	83%
Enhanced creativity	75%
Freed time to upskill	31%

"Gemini has saved me so much time that I was spending in my workday, doing tasks that were not using my skills."

Responsible AI adoption in Boulder starts with Colorado's playbook: use the State's living Guide to Artificial Intelligence as the governance baseline, require OIT/NIST‑aligned intake and mandatory training/attestations as used in the 90‑day Gemini pilot, and phase rollouts with measurable pilots and human verification so tools help staff instead of creating legal exposure; see Colorado Guide to Artificial Intelligence - OIT governance guide and the pilot report that tied training + attestations to real productivity gains: Colorado Responsible AI Pilot Report - InnovateUS.

Pair those controls with practical reskilling - e.g., a short, role‑based responsible‑AI course and attestation before tool access, plus cohort pilots - and operationalize outputs through documented human review; for upskilling, consider Nucamp's AI Essentials for Work to teach prompt design, risk-aware use, and prompt verification workflows: AI Essentials for Work: syllabus and registration - Nucamp.

The payoff: governed pilots turn informal GenAI use into auditable, productive services that reduce repetitive work while keeping Boulder compliant and accessible.

Program	Length	Early Bird Cost	Links
AI Essentials for Work	15 Weeks	$3,582	AI Essentials for Work syllabus and registration - Nucamp

“If we didn't come forth with a product, people are going to be using it anyway. And there's danger in people actually using applications that are not part of your enterprise.”