Government AI Playbook Permit intake and a first case that can travel. Six families of work, fourteen design tests, and a path that puts a person on every approval, every denial and every citizen-facing determination. Not a survey, not a certification, and not a close of any named regime. Edition: November 2026. 28 pages. Written for: Permit, planning, records, compliance and technology leaders in municipal and county agencies, and the sponsors who have to fund a first case. Canonical URL: https://asymilinkmeta.com/resources/government-ai-playbook Written for the search query: government AI playbook Key takeaways - This is the single government document behind the permit-compliance page — not a survey, not a certification, and not a close of any named regime. - Six families of work, each with an honest first case and a case that is too wide. Intake and completeness is an honest first case; a system that issues an approval, a denial or a citizen-facing determination without a person is not. - Fourteen design tests are applied to every case. A person sits on every approval, every denial and every citizen-facing determination. - Return is measured on a named process against a signed baseline, using the ROI guide. This playbook will not invent a government-sector result. What's in this playbook Agencies are not short of AI ideas. They are short of first cases that can travel: a named process, a signed baseline, an operator who is not only the original builder, a person who can stand behind an approval, a denial or a citizen-facing determination, and a way to stop. This playbook is the single government document behind the permit-and-compliance use-case page and the government playbook path. It is one playbook, on purpose, with permit intake and compliance review in depth. It is not a survey, not a certification under a named information-security or records framework, and it will not make a permitting determination in an agency's name. It maps six families of work. Each family has an honest first case and a case that is too wide. An intake assist that extracts fields, flags missing items, and prepares a file for a named intake officer is an honest first case in permitting, once the pilots already running are on an inventory and a proportionate governance path is written. A model that automatically approves or denies a "routine" permit is not. The same distinction is drawn for code and zoning review (a recommended reading, with evidence, for a named reviewer), routing and case assembly, applicant and public communications (a person on every official message), records and public-records requests (a named officer on every release or withholding), and an internal process already run: AI as an intake and evidence layer first; an unsupervised approval is not a first case this playbook will recommend. How the tests and the method are meant to be used Fourteen design tests are applied to every case. The answers will differ. The standard does not. A missing counsel answer, an undefined "review," or a pilot that is not on an inventory, is a reason to pause, not permission to proceed. Counsel and the records or compliance function decide what a public-records, privacy, due-process, equity or information-security rule, if it applies, requires, and when. This playbook will not make those determinations, will not certify alignment with a named framework, and will not claim a certified connector to a named government-technology vendor. Fairness is examined for this use, against a frozen set, by a person; this playbook will not publish a numerical threshold. Hosting is decided before a case is funded. Citizen data may not be able to leave. Where they cannot, AsymiLink AI runs on your own hardware, with an optional air gap. Built to SOC 2 standards is the exact wording we use; it is not a certification of the agency's system. The illustrative organization is a hypothetical mid-size municipal permitting agency, labeled as such and introduced in this playbook: overall 16, Emerging, Infrastructure a relative strength at 3, Governance the constraint at 1, two intake pilots neither of which is in production with an operator, and a next-step target of 24, Developing — a next step, not a timeline, and not a promise. Its honest first work is to write a proportionate governance path, put the two existing pilots on an inventory, then, if the tests pass, a bounded intake or completeness assist — not a third permitting pilot, and not an unsupervised approval or denial. The financial method is The Complete Guide to AI Automation ROI. This playbook will not invent a government-sector result. The 3.8x average first-year ROI across client engagements is an average, not a forecast or a commitment for any individual project. ROI is measured across client engagements against a baseline agreed before work starts. It is an average, not a forecast for any government project. Six families of work each with an honest first case, a case that is too wide, and a defined gate. Each item below is the shape of the map; the full wording, honest first cases and “too wide” warnings are in the PDF. F1. Permit intake and completeness F2. Code, zoning and compliance review F3. Routing, assignment and case assembly F4. Applicant and public communications F5. Records, retention and public-records requests F6. Internal process and a proved pattern The PDF gives, for each family, an honest first case, a case that is too wide, and a defined gate. The summaries above are the shape of the map, not a substitute for that wording. Fourteen design tests (excerpt) Fourteen design tests are applied to every case. The answers will differ. The standard does not. The full wording is in the PDF and the companion workbook. T1–early. inventory, including the pilots already run; proportionate consequence; a person on any approval, denial or citizen-facing determination, with review defined; counsel and the records function asked in time; explanations, data lineage and a frozen evaluation set. Mid-tests. hosting decided up front, including where citizen data cannot leave; a write that approves, denies or releases a record as a different design from a read; an audit trail that can reconstruct a day; a named operator and a switch-off; training and reuse written down. Later tests. fairness and equal treatment examined for this use, not scored against a threshold we invented; honest notice to the people who are affected; a bounded first case that raises the path, not a third permitting pilot; a signed baseline on the ROI method, and monitoring after launch that the organization that owns the system writes down. A completed test sheet is not a certificate, not a compliance score, and not a finding that a system is safe or aligned with a named regime. Frequently asked questions Q: Is this a certification under a named government information-security or records framework? A: No. This playbook will not certify alignment with a named framework, will not make a permitting determination in an agency's name, and will not claim a certified connector to a named government-technology vendor. Counsel and the records or compliance function decide what applies, and when. A completed test sheet is not a compliance score. Q: What is an honest first case? A: A named, bounded process that raises the path — an inventory of the pilots already running, a proportionate governance path, an operator who is not only the original builder, counsel asked, a defined gate, a trail, a stop, a signed baseline — and that the current foundations can carry. For the illustrative municipal permitting agency, that is the inventory and a governance path, then, if the tests pass, a bounded intake or completeness assist, not a third permitting pilot and not an unsupervised approval. Q: What does "review" have to mean? A: "Human in the loop" is too vague to be a gate. The design states, in writing, who the person is, which flags, recommended readings or packets they re-check, what they approve, deny, release, refuse or stop, and that the model cannot bypass the gate. A person sits on every approval, every denial and every citizen-facing determination. A "routine" permit is not a reason to skip the person. The flag is advice. The person decides. Q: Can an automatic approval or denial of a "routine" permit be a first case? A: No. This playbook will not recommend, as a first case, a model that approves or denies a permit without a person who can stand behind the determination, including for a "routine" permit. A bounded intake or completeness assist, with a named officer, can be an honest first case when the inventory and the governance path are already in view. A later auto-approval or records-release case takes the tests again. This playbook will not pre-clear it. Q: How should return be measured? A: On a named process, against a baseline the organization agrees before work starts, using the method in The Complete Guide to AI Automation ROI. This playbook will not invent a government-sector figure, an approval-time rate, or a backlog-reduction number. The 3.8x average first-year ROI across client engagements is an average, measured against a baseline agreed before work starts, not a forecast for a government project. Quantifying every engagement's return after launch is not a standard part of AsymiLink Meta's operations. Q: Where does this sit with the other library documents? A: The AI Readiness Assessment Framework uses the same eight-dimension model as Section 7. The offline-deployment guide is the companion where citizen data cannot leave. The ROI guide is the financial method. None of those documents, and not this one, is legal, records or regulatory advice, and none of them will make a permitting determination in an agency's name. This page is the public summary. The full PDF and companion workbook are available through the request form on the page. There is no public file URL that bypasses that form.