Topic hub
Artificial Intelligence in Government
Where AI helps agencies today, how to evaluate and buy it, how to secure it, and how to govern automated decisions.
Artificial intelligence reaches agencies in three ways at once: as a product feature in software already licensed, as staff quietly using consumer tools, and as a deliberate project. All three need the same things: an inventory of where it is used, a rule about what data may go into it, a human accountable for consequential outputs, and documentation a member of the public could read.
The material here is practical rather than speculative. It covers the tasks where these systems are useful now, the questions to ask a vendor claiming AI capability, the security problems that are specific to language model applications, and the contract terms that keep an agency's data out of someone else's training set.
HQT capabilities in this area
- Data & AnalyticsData platforms, reporting, GIS, open data, and machine learning aligned to public-sector AI guidance.
Services
- Machine Learning & AI
- Responsible AI Policy & Governance
- Data Platforms & Warehousing
- Records & Public Records Support
- Data Governance & Quality
- Dashboards & Reporting
- Open Data & Transparency Portals
- GIS & Spatial Analytics
- Digital Evidence & Records Workflows
- DevSecOps & Delivery Pipelines
- Cloud Architecture & Migration
- NIST 800-171 & CMMC Readiness
Solutions by agency type
- Data and Analytics for County GovernmentsCross-departmental reporting, outcome measurement, and data sharing that respects confidentiality rules.
- Data and Analytics for Cities and TownsCouncil reporting, open data, and operational dashboards built on definitions everyone agrees with.
- Data and Analytics for Water, Power, and UtilitiesTurning meter, sensor, and asset data into leak detection, condition assessment, and rate support.
- Data and Analytics for Public SafetyOperational reporting, state mandates, and public transparency built from dispatch and records data.
- Data and Analytics for School Districts and CollegesStudent data warehouses, state reporting, and early warning analytics with privacy built in.
- Data and Analytics for State AgenciesData platforms, automated federal program reporting, and governed dashboards for state departments answering to funders, auditors, and the legislature.
- Data and Analytics for Transit and Transportation AgenciesReconciling ridership, on-time performance, asset, and federal reporting data into one set of definitions everyone can use.
- Data and Analytics for Special Districts and Joint Powers AuthoritiesBoard reporting, mandated filings, and GIS for single-purpose agencies that have plenty of data and no analyst on staff.
- Data and Analytics for Corporations and Private CompaniesData platforms, governance, and model documentation built to the standard an auditor or a government customer will ask a company to prove.
Step-by-step guides
- How to evaluate an AI feature before you enable itA review process for the artificial intelligence features vendors are adding to software agencies already license.
Reference articles
- Responsible AI in the public sector: principles and practiceWhat responsible use of AI means for agencies, the frameworks that guide it, and concrete practices.
- AI security: prompt injection and other risks agencies should plan forThe new attack surface that language model applications create, and the controls that address it.
- Vendor AI disclosure: what to require about AI inside products you already ownHow agencies can require vendors to disclose AI features added to software they already own, and act on the answers.
- Generative AI in agency operations: where it helps todayThe agency tasks where large language models are already useful, the ones where they are not yet, and how to tell the difference.
- Public records readiness checklist for agency technology systemsA checklist for records officers and IT staff testing whether an agency system can retain, find, produce, and redact public records.
- Designing human review for AI outputs in agency workflowsHow to design the human review step in an AI workflow so it catches real errors: triage, evidence, workload, and override data.
- Using AI to sort and summarize public comments on environmental documentsHow planning staff can use language models to cluster, deduplicate, and route comments on an EIR or general plan update without losing the record.
- Retrieval-augmented generation: AI assistants that cite agency sourcesHow an assistant can answer questions from the municipal code, policies, and records without inventing answers.
- Intelligent document processing: turning forms and PDFs into dataHow agencies extract structured data from applications, invoices, and scanned records with modern document AI.
- Automating agency workflows with integration and AIWhere automation removes manual steps between systems, and where AI extends it to tasks that used to need a person to read.
- Building an AI inventory: tracking where AI is used in your agencyWhat belongs in an agency AI use case inventory, how to find AI already embedded in licensed products, and who keeps the record current.
- Writing a generative AI use policy for agency staffWhat a staff generative AI policy has to decide: approved tools, data limits, review obligations, disclosure, and personal accounts.
- Automating accounts payable: invoice capture, routing, and controlsHow agency finance teams automate invoice intake, coding, and approval routing while keeping vendor bank change verification human.
- Computer vision for code enforcement: imagery, privacy, and due processHow agencies use vehicle, drone, and aerial imagery to detect code violations, and the privacy, retention, and due process rules that govern it.
- Zero Trust architecture for public agenciesThe principles behind Zero Trust, the maturity model, and a realistic sequence for a small security team.
- AI-assisted redaction for public records requestsHow machine learning can find personal information, criminal justice data, and other exempt content faster, with a person making the final call.
- Where to run a language model: a decision guide for agenciesHow agencies choose among a commercial AI service, a provider model in their own cloud tenancy, and a self-hosted open model.
- Endpoint detection and response: EDR and managed detection for agenciesWhat EDR adds beyond antivirus, and how agencies decide whether to watch the alerts in house or buy managed detection.
- AI impact assessments: a practical process before a system goes liveWhat an AI impact assessment contains section by section, who signs it, when it is refreshed, and how much work it takes.
- AI prompts and chat logs as public records: retention and disclosureHow prompts, uploads, and AI chat histories interact with state public records law, and what to log, keep, and disclose.
- Automated decision systems: notice, explanation, and appeal rightsWhat makes a system an automated decision system, and the notice, explanation, human review, and documentation duties that follow.
- AI coding assistants on agency software projects: policy and reviewHow agencies and their contractors should govern AI coding assistants: data exposure, licensing, ownership, review, and disclosure.
From the blog
- Prompt injection is the new SQL injection, and we test for it
- Running language models inside the agency's cloud boundary
- Your vendor's new AI feature is a new system, not an update
- An acceptable use policy for generative AI should govern data, not tools
- Retrieval quality, not the model, decides whether an AI assistant is useful
- Zero Trust without buying a Zero Trust platform
- Every model we deploy for an agency ships with a model card
- What we learned building an assistant that answers questions from a municipal code
- Using AI to triage service requests without losing the human
- Small open models are good enough for many agency tasks
- Five terms we think every agency AI contract should include
- If you only do one security project this year, make it multi-factor authentication
- Low-code or custom code? A decision guide for agencies
- Cloud cost surprises and public budgets do not mix