Building an AI Playbook for Your Nonprofit

Why Individual AI Use Is Not Enough

Research consistently shows that 81% of nonprofits use AI on an ad hoc, individual basis. Staff members discover tools on their own, develop personal workflows, and generate value for themselves, but rarely share what they have learned with colleagues. This pattern creates several compounding problems that limit the impact AI can have on the organization as a whole.

Knowledge evaporates when staff leave. If your most AI-proficient employee departs, they take their prompts, workflows, and hard-won lessons with them. Organizations that rely on individual expertise rather than documented systems face this cycle repeatedly: a capable person joins, learns to use AI effectively, builds undocumented knowledge, and eventually leaves. Their replacement starts from scratch. A playbook breaks this cycle by embedding organizational knowledge in systems that outlast any individual.

Inconsistency erodes quality and trust. When different staff members use different approaches for the same task, outcomes vary widely. One donor thank-you email is warm and specific; another is generic and stiff. One grant narrative follows best practices; another misses the funder's priorities. Documented workflows create a floor of consistent quality that every staff member can meet, regardless of their individual AI proficiency.

Ungoverned AI use creates real risks. Organizations without clear policies about which data can enter AI tools, which tasks require human review, and who is accountable for AI-generated content face growing exposure. As AI tools proliferate, the question is not whether your staff will use them but whether they will use them safely and consistently. A playbook provides the governance framework that makes confident, responsible AI use possible across the organization. If you are just getting started with structured AI use, our nonprofit leaders guide to AI covers the foundational concepts in depth.

Building Your Standard Operating Procedures

Standard operating procedures for AI-assisted tasks follow a specific structure that differs from traditional SOPs. Where a traditional SOP describes steps a human takes, an AI SOP defines the human-AI collaboration: what information the human gathers, what they ask the AI, what the AI produces, and how the human reviews and finalizes the output. The handoffs between human and AI judgment are the most important things to document.

Each SOP needs ten elements: a title and purpose (what task does this govern), scope (who uses it and when), roles and responsibilities (who initiates, reviews, and approves), a trigger (what event starts this workflow), inputs (what information to gather before prompting), prompt instructions (the actual template with placeholders for variable content), AI actions (what the AI produces), quality checks (specific review criteria), output destination (where the final result goes), and version history.

The quality check step is the most frequently skipped and the most critical. Every AI SOP should include explicit criteria for what a good output looks like before it leaves the organization. For a grant narrative, the quality check might specify: verify all outcome figures against source data, confirm funder priorities are addressed by name, and have a program staff member review for factual accuracy. For donor communications, it might require a tone review to ensure warmth and specificity. These criteria make the review step concrete rather than vague.

Start by documenting your highest-frequency, lowest-risk workflows. Grant narrative drafting, donor thank-you emails, social media content, program impact summaries, and board meeting preparation are good starting points. These tasks happen regularly, have clear quality standards, and carry moderate stakes, making them ideal candidates for early AI workflow documentation. As your team builds confidence with these, you can expand to more complex or sensitive tasks.

Governance That Enables Rather Than Restricts

The most common governance failure is writing a restrictive policy out of fear, then watching staff ignore it and use AI anyway without any guidance. Effective AI governance in a nonprofit playbook builds trust rather than anxiety. It gives staff confidence that they are using AI appropriately, and it creates clear boundaries so people know when they need to ask for guidance.

A tiered risk framework is the most practical governance structure for nonprofits. Tier AI use cases by the sensitivity of the data involved and the potential impact of errors. Low-risk tasks, such as drafting internal documents, summarizing public information, or generating social media ideas, can proceed at staff discretion with basic policy guidelines. Medium-risk tasks, including external communications, donor-facing content, and program descriptions, require supervisor review before publication. High-risk tasks, those involving client data, financial figures, legal language, or board-level reporting, need a defined approval process and documentation.

The data handling rules are the most critical governance component. Public AI tools like ChatGPT and Claude receive data entered into their interfaces, which may be used for training or reviewed by safety teams. Your playbook must clearly specify what can never enter a public AI tool: client names and personal information, donor financial details, employee performance data, unreleased financial figures, and proprietary program data. These rules protect your organization and the people you serve. Staff need to understand not just the rule but the reason, so they can make good judgments in situations the playbook did not anticipate.

Equally important is defining what AI should not do, regardless of technical capability. Your playbook should explicitly prohibit using AI to make or recommend eligibility decisions for services, draft crisis communications involving trauma or grief, create performance or disciplinary documentation, or produce any content presented as coming from a real person without human authorship. These boundaries reflect your organization's values, not just its risk tolerance. For a comprehensive approach to AI governance frameworks, see our guide on building AI champions who drive responsible adoption.

From Individual Capability to Organizational Intelligence

The organizations closing the gap between AI adoption and AI impact are not those with the most sophisticated tools or the largest technology budgets. They are the ones that have done the unglamorous work of documenting what they know, creating systems that share that knowledge, and building the governance that makes confident use possible at scale.

An AI playbook is the infrastructure that makes this possible. It captures the best practices your most capable staff have developed and makes them available to everyone. It reduces the cognitive overhead of starting every AI task from scratch. It gives new staff a foundation to build on rather than a blank slate to navigate alone. And it creates a feedback loop that helps the organization learn and improve over time, rather than cycling through the same discoveries repeatedly.

Starting small is not just acceptable, it is the right strategy. Document two workflows. Build a prompt library for one department. Write a one-page data handling policy. Then iterate. The organizations achieving meaningful AI impact did not begin with comprehensive playbooks. They began with intentional, documented practice and built from there. For comprehensive support on developing your organization's AI capabilities, our guide on AI for board communications and our overview of AI for nonprofit leaders provide practical starting points for different parts of the organization.