Building Your AI Policy Garden

The Community Garden Model for Cutting-Edge AI Policy

A few weeks ago, I walked past a community garden in my neighborhood. It takes up half a city block, with overflowing raised beds, a fence with a gate that latches but doesn’t lock, cramed between the elementary school and the bike path. A city truck was there with someone tending the water lines while others tending their early spring flowers and prepping some vegetable plants.

I get asked by people I run into in meetings and at my daughters school, “everyone is saying I should be using AI but nothing fits into what we are doign day to day.” From the 10,000 foot view of their organization, AI looks like a great way to improve efficiency, automate boring tasks, and cut costs. It is applying that to the day to day operations that make it hard.

The perfect example of how to build an AI first organization was with us the whole time. that community garden is exactly what building and maintaining a cutting-edge school or small business AI policy should look like. Most AI policy advice tells you to build a castle, setting up walls to lock out the world. However, AI adoption is fundamentally about letting good ideas in safely. If we want to safely harness AI while keeping our classrooms and operations human-centered, we need a community garden.

Here is what a cutting-edge AI community garden looks like in practice, blending the best of modern governance with operational efficiency.

The Welcome Sign: Advanced Definitions

A cutting-edge policy establishes formal definitions for terms like Artificial Intelligence, Generative AI, Large Language Models (LLMs), Algorithms, and Natural Language Processing (NLP). Crucially, the community must distinguish between “Open-Source AI” (where inputs might be used to train public models) and “Closed-Source/Proprietary AI” (where underlying code and data inputs are obscured and contractually protected). As organizations evolve, definitions should also expand to include “Agentic AI,” which refers to systems that can take actions and operate autonomously across multi-step workflows.

The Fence: Strict Digital Safety and Data Privacy

In our AI garden, the fence represents strict data privacy and digital safety bans:

• Zero-Tolerance Threats: We explicitly fence out toxic weeds by strictly prohibiting the use of AI to generate false, misleading, or derogatory representations of others. This includes a zero-tolerance ban on creating or distributing deepfakes and Non-Consensual Intimate Imagery (NCII).
• Data and IP Protection: The fence protects student and customer Personally Identifiable Information (PII) from being ingested by public AI models. Districts and businesses must mandate “Closed AI” systems (like enterprise plans) that provide contractually-guaranteed data security. Additionally, boundaries must be set to ensure generated content does not infringe on Intellectual Property (IP) or copyrights.
• Protecting Community Trust: The fence also guards against losing the trust of your community. AI-generated outreach must be carefully managed so it isn’t marked as spam or perceived as inauthentic by parents or customers.

The Head Gardener & Stakeholder Committee: Proactive Governance

The “head gardener” (the Superintendent, Tech Director, or even a commitee) doesn’t dictate exactly what everyone grows, but maintains the conditions that make growing possible.

To support the gardener, cutting-edge organizations establish an ongoing Stakeholder AI Workgroup to regularly evaluate new tools, assess privacy risks, and keep the community informed. Before any new AI tool is adopted, this group conducts a comprehensive risk assessment to verify legal and ethical safeguards.

The Shared Shed: Approved Tools to Maximize Productivity

The good tools live in the shared shed so nobody has to figure out the security review on their own. These shared tools and resources contain a curated, public inventory of approved tools.

• Deep Integration: Rather than just relying on basic chat windows, the shared tools should offer AI deeply integrated into the tools your team already uses (like Google Workspace, Canva, or financial software) to seamlessly tackle busywork and administrative burdens.
• Socratic Systems: increasing engagement means decreasing screentime; the shed might feature “Socratic/RAG-Based Systems” restricted to curated content and designed to scaffold learning through questioning and interaction with peers, rather than simply giving immediate answers.

The Individual Plots: Human-Centered Ethical Use

Each gardener’s plot looks different because each job is different. AI is meant to empower and supplement human interaction, not replace human effort.

• Automating the Boring Stuff: Staff should use AI to automate “the boring stuff” that eats up hours every week. This includes extracting details from emails, organizing spreadsheets, repurposing long-form content into shorter snippets, and drafting personalized responses to repetitive questions.
• The “Human-in-the-Loop” Rule: While AI does the heavy lifting, employees must rigorously review, verify, and approve the accuracy of all AI-generated content before an email is sent, a grade is entered, or a lesson is taught.
• Student Plots (Ethical Academic Use): With express permission, staff can use AI for ethical purposes like research assistance, data analysis, writing assistance (grammar and structure feedback), and language translation/accessibility support.
• An Honest Harvest: Using AI to complete work is the equivalent of serving frozen veges at the dinner table. We need to be honest with ourselves and others about when we use AI and when we feel we need to do something ourselves. When we do use AI, we must properly cite the tool and explain how it was utilized.

The Rhythms: AI Literacy and Testing

A community garden runs on rhythm, and so does a working AI practice. Rather than relying on a top-down rollout, innovation compounds through “daily watering” and “weekly workdays.”

Schools and businesses must commit to developing “AI Literacy” for both students and staff, integrating safe AI usage into their normal rutines of doing work and professional development.

The Golden Rule for Adoption: Don’t try to automate everything at once. Pick ONE painful process and fix it properly. If an educator or employee has an idea for a new tool, they bring it to the head gardener, test it in a small, contained plot for a season, and if it works, it gets added to the shared shed for everyone to use.

Ultimately, this model provides what every organization needs from an AI policy: containment without paralysis. The fence keeps our data safe, the shed is stocked with deeply integrated, vetted tools, and the individual plots empower our teachers, students, and staff to automate the busywork and thoughtfully innovate. It should evolve as the organization and the environment it is in evolves.

If you need a starting point, here is a template to give you some structure to start with. Fill in the sections with what makes sense for your organization.

AI Policy Template

Post the document somewhere everyone can find it. See if it makes sense to everyone else. Does everyone know what to do if they think of a new AI app or prompt they want to try?

Remember to come back and update the document regularly. The tools available to your organization are changing all the time and the people in your organization need to know what is available to them and what is not yet safe to use.