Laguna Beach City Council
September 22, 2026

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
A Roadmap for Trusted Information
and Responsible AI
in Community Development
A Vendor-Neutral Framework for Municipal Planning, Records Governance, and AI Implementation
Peer Review Draft
Version 0.9
Prepared by
Karen Martin
Urban Planner
Founder & President, Pacific Planning Group, Inc.

Laguna Beach, California
Email: [insert preferred email]
August 2026
“Good decisions require trusted information.”
“Technology should strengthen public judgment, not replace it.” A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
Purpose of This Peer Review Draft
This paper is distributed for professional peer review and discussion. Its purpose is to invite constructive comments regarding the concepts, legal framework, governance principles, planning implications, records-management considerations, municipal workflows, technology assumptions, and implementation recommendations presented herein.

The objective of peer review is to improve the paper before formal publication and before adaptation into any city-specific policy, resolution, pilot charter, or implementation package.

Comments are especially welcome from professionals in planning, city management, municipal clerk administration, city attorney offices, building and safety, fire prevention, public works, information technology, records management, CEQA, housing, coastal planning, and municipal administration.

This document is offered as a vendor-neutral policy framework. It is not legal advice. It is not a proposal for consulting services.

Questions for Reviewers

1. Does the proposed roadmap accurately reflect current municipal practice?

2. Are there additional legal, governance, records-management, planning, technology, or administrative authorities that should be considered?

3. Are there practical implementation challenges that are not adequately addressed?

4. Would the proposed framework be useful to a city manager, planning director, city clerk, city attorney, IT director, building official, fire official, public works director, or elected body?

5. What revisions would make this framework more useful to local governments and professional organizations?
Acknowledgements
Acknowledgements will be completed after peer review. The author gratefully welcomes comments and suggested improvements. Any remaining errors are the responsibility of the author.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
Abstract
Community Development departments are information-dependent public agencies. Planning, building, fire, public works, code enforcement, the City Clerk, the City Attorney, appointed bodies, and elected decision-makers all depend on accurate records, current law, adopted policy, permit history, and complete application materials. Artificial intelligence can improve speed, consistency, and public service only if it is built on trusted information architecture. This white paper proposes a vendor-neutral roadmap for responsible AI in Community Development. It begins with records governance, metadata, source citations, legal and policy hierarchy, internal consistency, and human review. It then identifies practical AI use cases such as application completeness screening, property and permit history packets, interdepartmental review packets, staff report research support, code-collision detection, policy conflict matrices, and living inconsistency databases. The framework is designed for cities that want to use AI responsibly without allowing AI to make land use decisions, legal conclusions, CEQA findings, enforcement determinations, or permit approvals.

Core Thesis
Trusted information comes before trusted AI.

Community Development does not need an AI tool first. It needs a trusted information system that AI, staff, applicants, the City Clerk, the City Attorney, and decision-makers can all rely on. AI is not a substitute for professional judgment; it is a new consumer and organizer of public information.

How to Use This Paper
 As a policy framework for City Managers, Community Development Directors, City Clerks, City Attorneys, IT leaders, and elected officials.
 As a roadmap for a limited pilot before expanding any AI platform or procurement.  As a professional discussion paper for planners, municipal clerks, public-sector technologists, records managers, and local government attorneys.
 As a source of templates: use-case inventory, records-lifecycle checklist, completeness screen, property and permit history packet, code-collision report, and vendor-fit questions.

Important Notes
Not legal advice
This paper is a planning, records-governance, and implementation framework for discussion. It does not provide legal advice. Legal interpretations, conflict resolution, final staff recommendations, CEQA conclusions, permit determinations, enforcement determinations, public records determinations, and risk assessments should remain with City staff, the City Attorney, and authorized decision-makers.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
Vendor-neutral and not a services proposal
This paper does not recommend a specific AI vendor, records platform, permitting system, consultant, or firm. It is a vendor-neutral framework that a city can use to evaluate existing and future tools against municipal use cases, records obligations, governance controls, and professional responsibilities.

Professional-source orientation
The framework relies on official statutes, public-sector guidance, planning practice resources, records-management resources, municipal clerk professional materials, and recognized AI-risk frameworks. It intentionally avoids relying on individual advocacy, media interviews, media coverage, or vendor marketing as authority.

Table of Contents

1. Why Trusted Information Comes First

2. The Community Development Information Problem

3. Professional Source Framework

4. City Clerk and Records Governance

5. Planning Authority, Internal Consistency, and Code Collision

6. Information Architecture: Metadata, Source Citations, and Relationship Mapping

7. Responsible AI Governance Framework

8. Roadmap: Govern, Organize, Understand, Support, Improve

9. Primary Use Cases

10. Pilot Implementation Plan

11. Technology and Vendor Evaluation

12. Recommended Next Steps
Appendices
A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026

1. Why Trusted Information Comes First
Community Development departments are often described in terms of staffing, process, and technology. But their underlying function is information governance. Every application, staff report, CEQA analysis, plan check, inspection, public works comment, fire review, code enforcement action, appeal, public records response, and policy update depends on the quality of the record being used.

Artificial intelligence does not change that basic reality. It magnifies it. If a city gives AI incomplete records, disorganized scans, outdated policy text, unverified permit history, or internally inconsistent code, the system may produce faster answers while increasing the risk of error. The correct starting point is therefore not AI procurement. It is trusted information:
reliable records, defined source hierarchy, metadata, staff verification, public records controls, and a professional framework for human review.

Roadmap principle
AI should not become the fastest way to repeat a City's records problems. It should become a controlled way to reveal, organize, cite, and correct them.

For cities, the purpose of responsible AI should be practical: improve public service, reduce repetitive staff research, make application intake clearer, help departments work from the same record, strengthen staff reports, surface conflicts earlier, preserve institutional knowledge, and support better decision-making without delegating public judgment to software.

2. The Community Development Information Problem
Community Development is a cross-disciplinary function. Planning may be the front door for a discretionary application, but Building, Fire, Public Works, Code Enforcement, the City Clerk, the City Attorney, and decision-makers often need the same underlying information. The difficulty is that the information is rarely organized in one place or in one form.

Pain point Operational effect AI opportunity
Image-only or poorly OCR-scanned
files
Staff cannot reliably search or
extract key facts.

OCR with quality control;
document classification; source
citations.

Large PDFs spanning years or
decades
Applications, letters, approvals,
permits, plans, and inspections
may be mixed together.

Split records into document types;
create dated timelines; preserve
source-page references.

APN, address, and alias
inconsistencies
Staff, applicants, title records, and
historical permits may identify the
same property differently.

Create APN/address crosswalks,
alias tables, and property identity
metadata.

Planning, Building, Fire, Public
Works, and Code Enforcement silos
Departments may review from
different fragments of the record.

Create a shared interdepartmental
review packet with disciplinespecific flags.

Incomplete submittals Staff spends time chasing missing materials before substantive
Compare submittals against
entitlement-specific checklists and
A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
review can begin. produce source-cited missing-item
reports.

Code collision and internal policy
inconsistency
Staff may cite one provision for
approval and another for denial,
even within the same document.

Generate code-collision reports
and internal consistency matrices
for staff and legal review.

Cross-document conflicts General Plan, zoning, LCP, specific plans, design guidelines, and old
approvals may point in different
directions.

Flag legal hierarchy, potentially
relevant authority, and the need
for City Attorney review.

Institutional knowledge risk Experienced staff may leave with undocumented historical
understanding.

Capture staff-verified corrections,
interpretations, and source-cited
research as institutional
knowledge.

Local modernization context
Many cities are already trying to improve processing through code cleanups, application changes, administrative review pathways, and objective standards. A responsible AI roadmap should support those reforms rather than compete with them. AI is most useful when it helps make existing reform efforts continuous, source-cited, measurable, and easier for staff to maintain over time. City-specific examples can be added in a local appendix or staff report without changing the core framework.

3. Professional Source Framework
A city-facing AI roadmap should not be written around a technology product. It should be written around public-sector professional standards and legal responsibilities. The following framework organizes the major professional source categories that should inform municipal AI use in Community Development.

Source category Representative sources Implication for city AI roadmap AI risk management NIST AI Risk Management
Framework; NIST Generative AI
Profile
Use governance, mapping,
measurement, and risk
management before deployment.

California public-sector AI
governance
California Department of
Technology Technology Letters and
GenAI policies
Require use-case inventories, risk
assessment, responsible use,
procurement controls, training,
transparency, and lifecycle
management.

Local government management ICMA AI in Local Government materials; National League of Cities
AI resources; GovAI Coalition
templates
Start small, govern deliberately,
train staff, avoid shadow AI, and
measure outcomes.

Planning practice and ethics American Planning Association AI and ethics guidance
Define the problem first, protect
sensitive data, preserve
professional judgment, monitor
bias, and ensure transparency.

Records and City Clerk practice California Secretary of State records guidance; CalRIM;
IIMC/NAGARA; California
Treat AI as part of the information
lifecycle: create, store, retrieve,
use, retain, archive, dispose.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
Municipal Clerks Association
Planning law and consistency California Government Code, Coastal Act/LCP framework, ILG,
HCD, relevant case law
Flag internal consistency, zoning
consistency, completeness rules,
LCP requirements, and legal
framework questions for staff and
City Attorney review.

4. City Clerk and Records Governance
The City Clerk perspective is central, not incidental. AI prompts, outputs, completeness screens, property history packets, conflict matrices, routing summaries, staff corrections, and sourcecitation logs may all become records if they are prepared, owned, used, or retained in the conduct of public business. A city should not wait until an AI tool is embedded in daily workflow to decide how those records are managed.

The information lifecycle
Lifecycle stage City questions
Create Who may use AI? What prompts are permitted? What information may not be entered?
Store Where are prompts, outputs, source files, and staff corrections stored?
Retrieve Can the City Clerk, records staff, and authorized departments search AI-related records?
Use What AI outputs may be used in staff work? What
must be verified before use?
Retain What retention schedule applies to prompts, outputs, drafts, logs, and source-cited packets?
Archive Which AI-assisted records have historical,
administrative, or legal value?
Dispose How are non-retained drafts and obsolete outputs disposed of in accordance with adopted retention
rules?
Records governance should also address the status of draft outputs. A draft AI-generated completeness screen, for example, may be less authoritative than the final staff completeness letter, but it may still be subject to records-retention and disclosure analysis depending on whether the City uses or retains it. The same is true for an AI-generated conflict matrix, staffcorrected property history packet, or legal-framework flag.

City Clerk controls for AI pilots
 Define AI-related records before launch: prompts, outputs, logs, source-cited packets, staff corrections, and vendor-generated reports.
 Confirm where those records are stored and whether they can be searched, exported, and produced.
 Update or interpret retention schedules before routine use begins.  Preserve source citations and metadata so AI outputs do not become unsupported summaries.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
 Require vendors to address data ownership, exportability, retention, deletion, audit logs, and public records response support.
 Coordinate with the City Attorney before assuming any exemption applies to AI-generated drafts or deliberative work product.

5. Planning Authority, Internal Consistency, and Code Collision The planning side of responsible AI begins with legal and policy hierarchy. Not every document has equal authority. A current statute, certified LCP policy, adopted General Plan element, adopted zoning standard, final approval resolution, draft staff report, scanned letter, and AI summary should not be treated as equivalent sources. A responsible system must rank sources, cite them, identify conflicts, and route legal questions for professional review.

Internal General Plan consistency
California planning law expressly recognizes internal General Plan consistency. Government Code section 65300.5 provides that a General Plan and its elements and parts are intended to comprise an integrated, internally consistent and compatible statement of policies. Professional planning guidance describes this as horizontal consistency: consistency among elements, within elements, and between text and diagrams or maps. AI should therefore compare a General Plan against itself, not merely compare the General Plan to zoning.

Code collision
A responsible AI roadmap should not assume that conflicts occur only between separate documents. In many cities, conflicts also exist inside the same adopted document. The zoning code may contain definitions, use tables, development standards, overlay provisions, procedures, or findings that do not align with each other. The General Plan may contain elements, policies, maps, or implementation programs that point in different directions. The Local Coastal Program may include conflicts within the Land Use Plan, within the Implementation Plan, or between the two. These internal conflicts can create the appearance of discretion, but in practice they may lead to inconsistent staff recommendations, unpredictable outcomes for applicants, weaker findings, and greater appeal or litigation risk. AI should be used to compare each major policy or regulatory document against itself, identify internal inconsistencies, cite the exact sections, and flag the applicable legal framework for staff and City Attorney review.

Core legal/process principle
The goal is not for AI to resolve the conflict. The goal is for AI to reveal the conflict, identify the governing consistency requirement, and prevent staff from treating a legal inconsistency as merely a discretionary choice.

Cross-document consistency
AI should also identify conflicts among documents. For example, zoning ordinances must be consistent with the General Plan under Government Code section 65860. In the coastal zone, certified Local Coastal Programs include a Land Use Plan and an Implementation Plan, and local A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
coastal decisions must be understood within the Coastal Act/LCP framework. Specific plans, design guidelines, overlay zones, objective standards, permit conditions, prior approvals, and current state law may all need to be mapped and compared.

Legal-framework flagging, not legal decision-making
When AI detects a conflict, it should not decide which interpretation wins. It should identify the conflict, cite the exact sections, explain the apparent source hierarchy, identify potentially relevant statutes, regulations, guidance, and case law for staff and City Attorney review, and recommend the type of professional review required. Case law such as Lesher Communications, deBottari, and City of Morgan Hill v. Bushey may be relevant in General Plan/zoning consistency scenarios, but the City Attorney should confirm current good-law status and application to the particular facts.

6. Information Architecture: Metadata, Source Citations, and Relationship Mapping
AI systems are often described as search tools, but Community Development needs more than search. It needs information architecture. A staff member should not simply ask, “What does the file say?” The system should be able to identify which document is authoritative, which version is current, which parcel or address is involved, which permit or resolution is referenced, which department owns the issue, and which legal framework may apply.

Planning metadata
Metadata category Examples
Property identity APN, address, prior address, lot, tract, legal description, related parcels, alias names.

Document type Application, plan sheet, staff report, resolution, condition, permit, inspection, CEQA document, letter,
notice, appeal, enforcement record.

Authority level Statute, adopted policy, adopted ordinance, certified LCP, final project approval, staff work product,
historical evidence, AI draft.

Department Planning, Building, Fire, Public Works, Code Enforcement, City Clerk, City Attorney, outside
consultant.

Topic tags Completeness, CEQA, coastal, fire access, grading, drainage, historic, slope, housing law, public notice,
appeal, variance, CUP, CDP.

Status Draft, adopted, certified, final, appealed, expired, superseded, unfinaled, pending, archived.

Source citation Document name, date, page, sheet, section, permit number, resolution number, condition number,
hyperlink/location.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
Relationship mapping
Relationship mapping is the next step beyond indexing. It connects statutes, policies, code sections, checklists, plan sheets, permit history, and departmental comments into a network. For example, a design review application may trigger zoning standards, General Plan policies, LCP policies, public notice requirements, fire access review, public works drainage review, CEQA analysis, historical permit conditions, and appeal procedures. A responsible AI tool should map those relationships and provide source citations for each connection.

7. Responsible AI Governance Framework
AI governance should begin before a pilot becomes routine workflow. NIST provides a useful structure through the functions Govern, Map, Measure, and Manage. California public-sector GenAI policies provide a model for inventory, risk assessment, procurement controls, training, privacy, security, and lifecycle management. Local governments can adapt these concepts even when they are not legally bound by state-entity requirements.

Governance function City application
Govern Adopt an interim AI governance statement; assign roles; define human-review requirements; prohibit
unsupported external issuance of AI outputs.

Map Inventory use cases, data sources, legal risks, records impacts, privacy issues, and affected departments.

Measure Evaluate accuracy, citation quality, time saved, false positives/negatives, public records impact, and
equity/fairness indicators.

Manage Correct outputs, retire failed use cases, update training, maintain vendor accountability, preserve
audit trails, and scale only what works.

Minimum governance controls
 Use-case inventory and risk classification before expansion.  Public records and retention review by the City Clerk and City Attorney.  Privacy and data minimization controls, especially for personal, enforcement, legal, and sensitive application materials.
 Access controls and least-privilege permissions for records, prompts, outputs, and logs.  Source-citation requirement for factual statements used in staff work.  Human review before any AI-assisted output is issued externally or relied upon in a decision.  Correction logs so staff-verified changes become institutional knowledge.  Vendor transparency: data use, model use, training/no-training commitments, exportability, retention, deletion, and audit support.
 Bias and equity review, especially for code enforcement routing, completeness screening, and discretionary review support.
 Pilot evaluation before expansion or procurement scaling.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026

8. Roadmap: Govern, Organize, Understand, Support, Improve The roadmap translates the professional framework into operational steps. It is not a software sequence. It is a management sequence that defines how Community Development should work in the AI era.

Phase Purpose Examples Human role

1. Govern Set rules before scaling. Adopt interim
governance; assign roles;
define records, privacy,
human review, and pilot
metrics.

City Manager, City Clerk,
City Attorney, IT,
Community
Development.

2. Organize Make records readable,
findable, and sourcecited.

OCR, metadata,
APN/address crosswalks,
document classification,
version control, retention
mapping.

Records staff, City Clerk,
department staff, IT.

3. Understand Help staff see what the
record says and where
conflicts exist.

Timelines, prior
approvals, completeness
comparisons, department
triggers, code collision,
cross-document conflicts.

Planning, Building, Fire,
Public Works, Code
Enforcement, City
Attorney.

4. Support Prepare staff-verifiable
packets.

Completeness screens,
property history packets,
interdepartmental review
packets, conflict matrices,
staff report research.

Professional staff and
decision-makers.

5. Improve Turn findings into
institutional knowledge
and policy maintenance.

Inconsistency database,
code update issue list,
corrected metadata,
training feedback, vendor
evaluation.

Department leadership,
City Attorney, City Clerk,
IT, legislative body.

9. Primary Use Cases
Use Case 1: AI-assisted application completeness check
The City loads entitlement-specific checklists and department-specific submittal requirements.

The applicant uploads forms, plan sets, reports, photos, owner authorization, technical studies, and other materials. AI pre-screens the package and identifies what appears submitted, missing, unclear, or inconsistent. This supports, but does not replace, the formal completeness determination. This use case aligns with Government Code section 65943 because incomplete letters must identify items actually required on the agency's checklist.

Use Case 2: Property and permit history packet
For each pilot property, AI assembles a source-cited packet summarizing APN/address history, prior planning approvals, building permits, final inspections, conditions of approval, CDPs, CEQA A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
documents, public works requirements, fire conditions, code enforcement history, and missing or conflicting records. Staff verifies the packet before relying on it.

Use Case 3: Interdepartmental review packet
Planning, Building, Public Works, Fire, and Code Enforcement should not begin review from different fragments of information. AI can prepare a shared packet that routes the same sourcecited record to each department, with issues organized by discipline.

Use Case 4: Staff report and CEQA research support
AI can help locate similar prior staff reports, identify applicable policies, assemble site history, flag outdated templates, and prepare draft research sections. It should not make findings, CEQA determinations, or recommendations.

Use Case 5: Code collision and internal policy consistency review AI compares each adopted document against itself: General Plan element against element, map against text, zoning definition against use table, development standard against finding, overlay against base zone, LUP policy against LUP policy, and Implementation Plan provision against another provision. The output is a code-collision report, not a decision.

Use Case 6: Cross-document consistency review
AI compares General Plan, zoning, LCP, specific plans, design guidelines, objective standards, historical approvals, and current state law. It identifies source-cited conflicts and flags legal frameworks for staff and City Attorney review.

Use Case 7: Living inconsistency database
Each staff-verified conflict becomes an entry in a living inconsistency database. Over time, the database supports code updates, General Plan maintenance, LCP amendments, staff training, and consistent application review.

Use Case 8: Public records and knowledge retrieval support AI can help records staff identify likely responsive materials, locate records by metadata, summarize document inventories, and preserve source references. It should not determine what is disclosable, privileged, exempt, or retained; those determinations remain with the City Clerk and City Attorney.

10. Pilot Implementation Plan
The pilot should use real municipal work. Sample documents are not enough. The purpose is to determine whether AI improves actual workflows and whether staff can verify, correct, and trust the outputs.

Suggested pilot file set
 Five active planning applications, including at least two with interdepartmental routing issues.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
 Five recently completed applications, including at least two that experienced completeness or routing problems.
 Five building plan check examples.
 Three code enforcement cases.
 Three Public Works-heavy review examples.
 Three Fire review examples.
 Five messy historical parcel files with significant permit or entitlement history.  Known internal policy/code-collision examples and cross-document conflict examples.

Pilot outputs
 Completeness pre-screen report for each test application.  Property and permit history packet for each test property.  Interdepartmental routing summary.
 Code-collision report for same-document conflicts.
 Cross-document policy conflict matrix.
 Legal-framework flags identifying potentially relevant statutes, regulations, guidance, and case law for staff and City Attorney review.
 Living inconsistency database entries.
 Pilot evaluation report with accuracy, time savings, staff usefulness, public records implications, risks, budget/staffing lessons, and next-step recommendations.

Success metrics
Metric Measurement approach
Accuracy Compare AI outputs to staff-reviewed results; track false positives and false negatives.

Citation quality Evaluate whether staff can quickly verify every material factual statement.

Completeness improvement Measure whether missing checklist items are identified earlier and more consistently.

Time saved Measure staff time for intake, property history, department routing, report research, and plan review
support.

Conflict identification Track whether AI identifies known internal and crossdocument conflicts and whether staff finds the flags useful.

Records performance Assess whether prompts, outputs, corrections, citations, and logs can be searched, exported,
retained, and produced.

Equity/fairness Audit whether flags and routing patterns differ without legitimate code-based reasons across
geography or applicant type.

Scalability Determine what can be scaled, what requires records cleanup first, and what should not be scaled.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026

11. Technology and Vendor Evaluation
The city does not need one magic vendor. It needs the right mix of records governance, planning workflow expertise, public-sector technology, legal review, and staff capacity. Technology should be evaluated against the roadmap, not the other way around.

Category Best fit Questions to ask
Existing AI plan-review /
permitting tools
Application screening, preliminary
plan review, corrections support,
customer service support.

Can the tool handle local checklists,
source citations, historical parcel
files, code collision, department
routing, retention, and
exportability?
Records-management / EDMS
vendors
OCR, indexing, retention, metadata,
public records workflows, trusted
records.

Can it process old image-only PDFs
and preserve records retention and
authenticity requirements?
Permit-system integrators Live applications, plan check workflow, department routing,
status tracking, permit history.

Can it connect current applications
to historical records and
department review packets?
Public-sector cloud / search /
document intelligence integrators
Secure document environment, AI
search, permissions, document
intelligence, staff-facing assistants.

What does the city already own,
and can records be indexed with
permissions and citations?
AI document-processing specialists OCR, classification, extraction from mixed files, source-cited search.

How accurate is it on actual city
records, and can staff verify every
output?
Internal or independent
use-case/policy review
Use-case definition, workflow
mapping, output evaluation,
records/legal issue spotting.

Does the solution solve Community
Development problems
independent of any single vendor
product?
Vendor-neutral procurement principles
 Require a real-document pilot before major expansion.  Do not accept unsupported claims about accuracy; require measurable outputs and staff verification.
 Require the ability to export city data, prompts, outputs, metadata, logs, and corrections.  Require clear commitments on data use, model training, confidentiality, retention, deletion, and cybersecurity.
 Require source citations to exact documents, pages, sheets, code sections, policy sections, permits, or resolutions.
 Require the tool to distinguish governing authority from historical evidence and draft AI research.  Require staff correction workflows so verified corrections become institutional knowledge.

12. Recommended Next Steps

6. Adopt an interim responsible AI governance statement for Community Development before expanding AI use beyond limited pilot activity.

7. Create a cross-functional AI governance group including Community Development, City Clerk, City Attorney, IT, Public Works, Fire, Building, and Code Enforcement.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026

8. Inventory current systems: permitting platform, document management, records storage, Microsoft/Google/cloud environment, GIS/APN data, code publishing platform, public records platform, and department file locations.

9. Create a records and metadata plan for pilot files before asking AI to summarize them.

10. Select pilot use cases that include application completeness, property history, interdepartmental review, code collision, and cross-document consistency.

11. Create standard output templates and staff verification protocols.

12. Run the pilot, measure results, correct outputs, document limitations, and decide what to scale only after the city sees what works.

13. Prepare a formal resolution or policy adopting the roadmap principles once the city confirms the governance structure and pilot scope.

Closing principle
Trusted records produce trusted staff reports. Trusted staff reports support trusted decisions. Trusted decisions build trusted government. AI should strengthen that chain, not bypass it.

Appendix A: Source Hierarchy for AI Review
Rank Source type Examples AI instruction Legal/operational consequence
1 State/federal law
and binding
standards
State statutes,
CEQA, Coastal Act,
building/fire
standards, housing
law
Use as governing
authority; verify
currency.

May preempt or
control local
requirements.
2 Certified/adopted
local law and policy
Certified LCP,
General Plan,
zoning code,
specific plans,
adopted standards
and maps
Use as adopted
authority; flag
internal and crossdocument conflicts.

Must be applied
consistently and
within legal
hierarchy.
3 Final projectspecific approvalsResolutions, CDPs,
final permits,
recorded
conditions, final
inspections
Treat as strong
project-specific
evidence; verify
final status,
expiration,
supersession, and
applicability.

May establish
rights, obligations,
nonconforming
status, or
conditions.
4 Staff work product
and historical
evidence
Staff reports, letters,
old applications,
notes, draft
materials, unfinaled
permits
Use as
context/evidence;
cite source; do not
treat as controlling
authority.

May inform facts
but requires
verification.
5 AI-generated
research
Summaries,
timelines, packets,
matrices, drafts
Treat as draft
research requiring
source citations and
human review.

No independent
legal or decisionmaking authority.

A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
Appendix B: Records Lifecycle Checklist for AI Pilots
 What records will the AI tool access?
 Which records are official, draft, historical, confidential, privileged, or sensitive?  Who may upload records or prompts?
 Where are prompts, outputs, logs, and corrections stored?  Who can search, retrieve, export, or delete AI-related records?  Which retention schedule applies to prompts, outputs, logs, drafts, and packets?  How are source citations preserved?
 How are staff corrections retained and reused?
 How will the city respond to public records requests involving AI inputs or outputs?  What records must never be entered into the tool without legal or privacy review? Appendix C: Standard Output Templates
Completeness Pre-Screen
 Checklist item
 Submitted / possibly submitted / missing / unclear / not applicable  Source document and page/sheet
 Applicant correction needed
 Staff verification needed
 Department reviewer
Property and Permit History Packet
 APN/address cross-reference
 Timeline of applications, approvals, permits, inspections, appeals, and enforcement actions  Conditions of approval
 Department-specific issues
 Missing or conflicting records
 Source citations
 Open questions for staff or City Attorney
Code Collision Report
 Topic affected
 Document reviewed
 Provision A
 Provision B
 Type of collision: definition/use table/standard/finding/procedure/map/program  Potential legal framework
 Operational consequence
 Recommended review: Planning / City Attorney / other department  Code cleanup priority
A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026
Policy Conflict Matrix
Issue Source 1 Source 2 Potential legal
framework
Recommended
review
General Plan
internal
inconsistency
Element/policy/map Element/policy/map Gov. Code § 65300.5; ILG/HCD guidance
Planning + City
Attorney
Zoning code
collision
Definition/use
table/standard
Finding/
procedure/overlay
Code interpretation;
General Plan
consistency; due
process/fair
application
concerns
Planning + City
Attorney
General Plan vs.
zoning
GP policy/map Zoning section Gov. Code § 65860;
potentially relevant
case law
Planning + City
Attorney
Zoning vs. LCP Zoning section LUP/IP
policy/standard
Coastal Act /
certified LCP
framework
Planning + City
Attorney + coastal
expertise
Permit history vs.
current code
Prior
approval/condition
Current
code/LCP/GP
Final status, vested
rights/nonconformi
ng analysis as
applicable
Planning + City
Attorney
Appendix D: Vendor-Fit Questions
 Can the tool process old image-only scanned PDFs and mixed files?  Can it cite exact documents, pages, code sections, policy sections, plan sheets, permit numbers, and resolution numbers?
 Can it distinguish adopted authority from historical evidence and draft research?  Can it compare applicant submittals against city checklists?  Can it produce a property and permit history packet?
 Can it identify same-document code collision and internal General Plan inconsistency?  Can it flag cross-document General Plan/zoning/LCP/specific plan conflicts?  Can it identify potentially relevant statutes, regulations, guidance, and case law without making legal conclusions?
 Can it work within city security, access control, retention, and public records requirements?  Can staff correct AI outputs and preserve corrections as institutional knowledge?  Can the city export all data and outputs if it changes vendors?  What is included in the contract price versus billed separately for onboarding, data cleanup, training, and ongoing evaluation?
Appendix E: Selected Professional and Official Sources

1. National Institute of Standards and Technology (NIST), AI Risk Management Framework and AI RMF Core. https://www.nist.gov/itl/ai-risk-management-framework
A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026

2. NIST, Generative AI Profile and AI RMF resources. https://airc.nist.gov/airmf-resources/airmf/5-seccore/

3. California Department of Technology, Technology Letter 24-01: Generative AI Risk Assessment and procurement requirements for state entities.
https://www.cdt.ca.gov/policy/technology-letters/technology-letter-24-01/

4. California Department of Technology, Technology Letter 25-01: responsible and secure use of GenAI within the State of California. https://www.cdt.ca.gov/policy/technology-letters/technology-letter-25-01/

5. California Secretary of State, Local Government Records Program. https://www.sos.ca.gov/archives/records-management-and-appraisal/local-gov-program

6. California Secretary of State, Records Management Handbook and Electronic Records guidance. https://www.sos.ca.gov/archives/records-management-and-appraisal/records-managementhandbook

7. CalRIM, Electronic Records Guidebook.
https://archives.cdn.sos.ca.gov/pdf/calrim_e_records_guidebook_full.pdf

8. International Institute of Municipal Clerks (IIMC) and NAGARA, Local Government Records Management Technical Publication Series. https://www.iimc.com/154/Publications

9. California Municipal Clerks Association, AI Deep Dive training for local government and municipal clerk offices. https://www.calclerks.org/events/EventDetails.aspx?group=&id=2046050

10. League of California Cities, The People's Business: A Guide to the California Public Records Act. https://truckeetahoeairport.com/documents/1052-the-people_s-business- _guide_to_the_california_public_records_act-pdf

11. American Planning Association, AI in Planning: Opportunities and Challenges and How to Prepare. https://www.planning.org/publications/document/9255930/

12. American Planning Association, Planning Ethics and Generative AI. https://www.planning.org/blog/9295637/planning-ethics-and-generative-ai/

13. International City/County Management Association (ICMA), Artificial Intelligence in Local Government survey and governance materials. https://icma.org/documents/icma-survey-research-artificialintelligence-local-government

14. National League of Cities, AI Demystified / AI in Action resources for municipalities. https://www.nlc.org/initiative/ai-in-action-empowering-local-governments/

15. GovAI Coalition / City of San José, AI policy templates and governance resources. https://www.sanjoseca.gov/your-government/departments-offices/information-technology/ artificial-intelligence-inventory/govai-coalition/templates-resources A Roadmap for Trusted Information and Responsible AI in Community Development Peer Review Draft v0.9 - Karen Martin - August 2026

16. California Government Code § 65300.5: General Plan internal consistency. https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml? lawCode=GOV&sectionNum=65300.5

17. California Government Code § 65860: zoning consistency with General Plan. https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml? lawCode=GOV&sectionNum=65860

18. California Government Code § 65943: development application completeness determinations. https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml? lawCode=GOV&sectionNum=65943

19. California Coastal Commission, Local Coastal Program information. https://www.coastal.ca.gov/lcp/lcpinfo/

20. Institute for Local Government, Planning Framework / General Plan consistency. https://www.ilgplanninghandbook.org/planning-framework

21. California HCD, Analysis of Consistency with General Plan and Coastal Element / Housing Element guidance. https://www.hcd.ca.gov/housing-element/building-blocks/requirements/general-plan

22. Lesher Communications, Inc. v. City of Walnut Creek (1990) 52 Cal.3d 531. https://law.justia.com/cases/california/supreme-court/3d/52/531.html

23. deBottari v. City Council (1985) 171 Cal.App.3d 1204. https://law.justia.com/cases/california/court-ofappeal/3d/171/1204.html

24. City of Morgan Hill v. Bushey (2018) 5 Cal.5th 1068. https://law.justia.com/cases/california/supremecourt/2018/s243042.html Public Record Letter on General Plan Records and AI Policy City Council Study Session September 22 2026
September 22, 2026
Mayor and Members of the Laguna Beach City Council
c/o City Clerk
505 Forest Avenue
Laguna Beach, California 92651
Re: Records first and required attachments before General Plan RFP release and AI policy implementation
Dear Mayor and Councilmembers:

Please do not release the Comprehensive General Plan Update RFP or move beyond limited AI pilot activity until the Council establishes the records, consistency, and accountability requirements in the attached documents. The City must make recordkeeping its first priority. It cannot reliably determine whether a plan, ordinance, project, permit, or consultant deliverable is consistent with adopted policy if the governing records are scattered, unsearchable, incomplete, or unsupported by source citations.

General Plan consistency is not established by citing one supportive policy in isolation. Government Code section 65300.5 requires the General Plan and its elements and parts to function as an integrated, internally consistent, and compatible statement of policy. The Forest Avenue Promenade illustrates the problem. The City relied on the Land Use Element, but did not provide the public with a project-specific crosswalk to the other General Plan elements, maps, and policies the project may implicate. A citation to one element does not establish overall General Plan consistency.

Many residents, including me, have lost faith that the current Council understands how to carry out this work without putting the City into another costly mess. Too often, the City begins spending money, adopts an ordinance, or launches a project before it defines the governing rules, record, budget, scope, and public decision points. That is reactive government, not comprehensive planning. The Council must set the framework now in public, rather than leave it to later staff interpretation, consultant scope changes, or an AI tool.

The Comprehensive Zoning Code Update shows why this matters. Its authorized contract ceiling increased from $321,449 to $709,749. That does not prove the full amount was paid, but it does require a public accounting of actual invoices, scope changes, work products, acceptance, remaining budget, and work still unfinished before the City enters another multiyear planning contract.

Attached requests for Council direction
• Attachment A requires a reliable public record, a complete General Plan consistency method, a public deliverables ledger, and an interim process for policy conflicts. • Attachment B identifies the specific revisions that should be inserted into the General Plan RFP and consultant agreement before procurement begins.
• Attachment C sets the records, human-review, public-transparency, and pilot controls that must accompany any AI policy.

This request does not ask the City to stop comprehensive planning or responsible technology use. It asks the City to use the correct sequence: establish the record, identify the governing policy, make conflicts visible, and then decide what work to fund. Residents and taxpayers deserve a public record that lets them see what the City contracted for, what it received, and why it acted.

I am copying Stu News and the Laguna Beach Independent because the public should be able to see these commitments before the City issues a new RFP or expands AI use.

Respectfully,
Karen Martin
Laguna Beach resident, urban planner, and candidate for City Council cc: Stu News Laguna; Laguna Beach Independent
References: General Plan Update staff report and RFP attachments; Government Code sections 65300.5, 65454, 65860, 65943, and 7920.530; Public Resources Code section 30604.

Attachment A
Records and General Plan Consistency FIRST
Required direction before General Plan RFP release and AI policy implementation Records are the threshold requirement. The City cannot demonstrate that a proposed action is consistent with the General Plan, certified Local Coastal Program, zoning regulations, specific plans, prior approvals, or state law unless it can identify and cite the actual governing records. A consultant and an AI tool can organize information, but neither can repair an incomplete or unverified record after the City has already made a decision.

1. Establish and publish an authoritative planning and project record.

Before issuing the RFP, staff shall publish a source inventory for the General Plan and its elements, maps, and implementation programs; certified LCP; zoning regulations; specific plans; objective standards; active ordinances; final project approvals; technical studies; and all prior contracts and work products relevant to the update. Each item must identify its official source, adoption or effective date, current status, accessible location, searchability, and any known gap. The City must distinguish a record that has not been found from a record that does not exist.

2. Use a complete General Plan consistency method.

For every City-initiated planning action and project that requires a General Plan consistency finding, the public record must identify all relevant elements, policies, maps, diagrams, and implementation programs, not only a single favorable provision. It must also identify whether zoning, the certified LCP, a specific plan, a final approval, or state law raises a conflict or additional requirement.

Government Code section 65300.5 requires an integrated, internally consistent, and compatible General Plan; Government Code section 65860 separately requires zoning consistency with that plan.

The Forest Avenue Promenade is a current case in point. The City cited the Land Use Element, but that alone does not show how the project relates to the other relevant General Plan elements and policies.

The public record should include a source-cited consistency crosswalk, explain any policy tensions, and state the decision path for resolving them. A selective citation is not a comprehensive consistency analysis.

3. Create a public contract and deliverables ledger.

Before any General Plan Update invoice is paid, the City shall maintain one public dashboard showing each task, subtask, amendment, and optional study; approved scope; not-to-exceed amount; contingency; invoice and payment information; scheduled and actual milestone dates; nonconfidential work products; written staff acceptance; remaining budget; and explanation for each material delay, scope change, or cost change. The same dashboard should provide a retrospective status of the Comprehensive Zoning Code Update, including the original and amended scope, expenditures, completed and remaining work, transfer work, duplication, and gaps.

4. Adopt an interim Policy Conflict Record process.

When staff, an applicant, a member of the public, or an AI tool identifies a possible conflict among state law, the certified LCP, General Plan, zoning regulations, specific plans, objective standards, or prior approvals, staff shall prepare a source-cited Policy Conflict Record. It must state the exact texts or approvals involved, effective dates, affected matter, legal or policy framework, statutory deadline, responsible reviewer, and proposed resolution path. The process may not create a new substantive standard or an undocumented delay. Matters that require legal interpretation, a plan or code amendment, CEQA review, Coastal Commission review, or a noticed Planning Commission or Council action must be routed accordingly.

Required recordkeeping controls
• Assign responsibility for maintaining the authoritative inventory, records-gap register, public dashboard, and conflict records.
• Preserve source citations and version dates so a resident can verify the basis for a staff recommendation or Council action.
• Ensure that searches, exports, retention, and public-records review can be performed for material project files, consultant deliverables, and AI-assisted work product. • Use a written correction process when the City discovers that a record, policy citation, or project history is incomplete or incorrect.

Council direction at the study session
• Direct staff to publish the authoritative inventory and records-gap register before RFP release. • Direct staff to establish the public contract and deliverables ledger before consultant selection. • Direct staff to return with the interim Policy Conflict Record process and responsibility assignments.

Legal framework: Government Code sections 65300.5, 65454, 65860, 65943, and 7920.530; Public Resources Code section 30604. These requirements seek an accurate record and an identified review path; they do not direct a legal conclusion in a pending matter.

Attachment B
General Plan RFP Requirements
Required revisions before release of the Comprehensive General Plan Update RFP The RFP should begin with an accountable public baseline, not a consultant effort that must reconstruct missing records after procurement. The following requirements should be inserted into the RFP and consultant agreement before release. The consultant's work must be source-cited, publicly reviewable, and coordinated with the records and Policy Conflict Record process in Attachment A.

1. Make the authoritative records and consistency baseline the first notice to proceed.

The first deliverable shall verify the City's published source inventory and records-gap register. It shall establish the source hierarchy for state law, certified and adopted local policy, final approvals, historical evidence, and draft research. It shall identify incomplete, conflicting, image-only, or inaccessible records that affect the work plan and create a public plan for resolving those gaps.

2. Add a public handoff and duplication crosswalk to Task 14.

The consultant shall prepare an early crosswalk between the Comprehensive Zoning Code Update and the General Plan Update. It shall identify the original and amended scope; completed and remaining deliverables; expenditures and remaining budget; responsible parties; transfer work; duplication; and gaps. The crosswalk must be presented publicly to the full Council before the consultant begins policy drafting.

3. Strengthen Task 2 into a full internal and cross-document consistency audit.

The audit must compare the General Plan against itself, including elements, policies, maps, diagrams, and implementation programs, as well as compare the General Plan with the certified LCP, zoning regulations, specific plans, active ordinances, objective standards, final approvals, and relevant state law. For each inconsistency, identify the affected text, governing issue, recommended correction, required CEQA or Coastal process, cost, responsible department, and interim decision rule that follows Attachment A. Publish the audit and misalignments table before major policy alternatives are selected.

4. Make Task 12 capacity and fiscal analysis required base scope.

The base scope shall include reproducible analysis of water, sewer, storm drainage, roadway operations, emergency access and evacuation, seasonal visitor demand, public-service needs, capital facilities, lifecycle maintenance, operating costs, funding restrictions, and financing options. It must test existing conditions and future development assumptions, including RHNA and infill assumptions, and directly inform policy and implementation decisions.

5. Make Task 15 a funded implementation program.

Require a 20-year capital and fiscal implementation plan tied to each General Plan action. It shall identify the responsible department, time frame, dependency, capital cost, operating and maintenance cost, funding source and restrictions, required zoning or LCP action, performance measure, and public reporting schedule.

6. Require a public contract and deliverables ledger.

For every task, subtask, amendment, and optional study, the ledger must show the approved scope, not-to-exceed amount, contingency, Council approval date, invoice amount and payment date, milestone dates, nonconfidential work products, staff acceptance, remaining budget, and a plainlanguage explanation for each scope, cost, or schedule change. The dashboard must be updated at least monthly, and milestone payments should follow written acceptance and public posting of the related deliverable except where protected by law.

7. Set public cost and decision controls before consultant selection.

Before award, the Council should approve a task-level not-to-exceed budget, separately priced optional studies and environmental review, deliverable acceptance criteria, and a clear contingency rule. No task substitution, contingency draw, scope expansion, budget increase, or material policy change should occur without a noticed public staff report to the full Council that identifies the dollars, work added or removed, reason, and schedule effect.

8. Correct the schedule and preserve meaningful future Council review.

The RFP must include a realistic critical path for the Housing Element, Coastal Commission review, CEQA, public review, Planning Commission recommendation, and Council adoption. It should include public Council decision gates after the records and consistency audit and after the capacity and fiscal findings, so the Council seated in 2027 and later can make informed choices rather than inherit a fixed process without the evidence.

Agenda packet references: Staff Report, pp. 1 and 7-9; Attachment 2, Tasks 1, 2, 12, 14, and 15; Attachment 2 anticipated schedule, pp. 19-20.

Attachment C
AI Policy Requirements
Records governance human review and pilot controls for Community Development The City should not treat an AI policy as a general permission slip. Responsible AI begins with trustworthy information. Before the City expands AI use in planning, permits, code enforcement, public works review, or public-records support, it must establish what records the tool will use, where they are stored, how the City can verify them, and who remains accountable for each decision.

1. Put records governance before routine AI use.

The City shall identify the records an AI system may access and publish a records-lifecycle plan for material prompts, outputs, source-cited packets, corrections, logs, and vendor reports. The plan must identify storage, search, export, retention, deletion, access control, and public-records review. A record that is used or retained in public business may require retention and disclosure analysis; the City should not assume that an AI-generated draft is exempt or outside the records system.

2. Require a source hierarchy and exact citations.

AI outputs must distinguish governing authority from historical evidence and draft research. State and federal law and binding standards come first; then certified and adopted local policy; then final project approvals; then staff and historical records; and finally AI-generated research. Every material factual statement must link to the underlying document, page, sheet, code section, policy, permit, or resolution so staff and the public can verify it.

3. Define permitted and prohibited uses.

AI may help locate, organize, cite, summarize, and flag records or possible policy conflicts. It may not make a formal completeness determination, interpret law, select a controlling policy, impose a condition, make a CEQA finding, determine what is disclosable or privileged, make an enforcement determination, or approve or deny an application. Staff and authorized decision-makers must verify any output before relying on it.

4. Use the interim Policy Conflict Record process.

When AI identifies a possible conflict, it shall generate a source-cited flag for staff review, not a decision.

The City shall use the Attachment A process to identify the exact sources, effective dates, affected matter, applicable framework, deadline, reviewer, and resolution path. The Planning Director and City Attorney should determine whether the issue can be resolved under existing law or requires a noticed action, amendment, CEQA review, or Coastal Commission process.

5. Require vendor controls and portability.

Any contract must protect City data and prohibit training on City records by default unless expressly approved. It must address confidentiality, cybersecurity, data ownership, audit logs, retention, deletion, correction workflows, and the City's ability to export its data, prompts, outputs, metadata, and logs if it changes vendors. Vendors must demonstrate source citations and staff verification on actual City records before expansion.

6. Use a limited, transparent pilot before expansion.

The Council should authorize narrowly defined pilot uses such as records retrieval, property and permit history packets, interdepartmental review packets, and source-cited policy conflict matrices. The City shall publish the pilot purpose, data sources, responsible staff, vendor, safeguards, evaluation criteria, accuracy results, false-positive and false-negative findings, records performance, cost, and recommendation before expanding use.

7. Maintain public accountability.

The City should publish an inventory of active AI uses, identify the department and purpose for each, and provide a way for residents and applicants to understand when AI-assisted information materially informed a staff process. Human staff and elected decision-makers remain responsible for public explanations, legal compliance, and final action.

Policy foundation: Karen Martin, A Roadmap for Trusted Information and Responsible AI in Community Development, Peer Review Draft v0.9, August 2026. The roadmap is a vendor-neutral framework, not legal advice. Relevant legal framework includes Government Code sections 65300.5, 65860, 65943, and 7920.530.