AI for Nonprofits (3 of 10): A Practical Process for Getting Started
Nonprofits do not need an AI strategy because AI is new. They need one when AI can help solve important problems, reduce avoidable work, improve access to information, or increase the organization’s ability to advance its mission.
Buying licenses is not the same as becoming AI ready. An organization can provide AI tools to every employee and still lack clear goals, usable data, appropriate safeguards, staff confidence, and a way to determine whether the investment is producing value.
Cathexis Partners uses a seven-part AI Readiness & Enablement Process to help nonprofits move from general interest to responsible action. Each step below links to relevant Fundraising.AI tenets and a documented nonprofit example. These connections are our interpretation of how to apply the Framework; the examples illustrate individual practices, rather than endorsement of our process or proof of full Framework compliance.
1. Organizational discovery and AI readiness
Begin with the organization, not the product. What is leadership hoping AI will accomplish? Where are staff already using it? Which processes consume disproportionate time? What concerns do employees, donors, beneficiaries, and leaders have?
Ask staff about informal AI use without making the exercise punitive. Include people affected by the work, and establish a baseline: how long does the task take, where do errors occur, and what would a better result look like?
This step identifies the actual problems to solve and the organization’s capacity for change.
Fundraising.AI connection: Collaboration, Inclusiveness. Include diverse perspectives when defining priorities.
In practice: WaterAid formed a staff-led AI interest group with people from different levels and backgrounds, as described by a participant in Hart Square’s 2025 webinar recap. This illustrates discovery through staff involvement.
2. AI and technology landscape
Inventory the systems already in use and the AI capabilities already available. Review the Microsoft or Google environment, CRM, fundraising platforms, marketing tools, meeting products, reporting tools, integrations, and unofficial AI accounts.
Distinguish AI that predicts outcomes, generates content, retrieves organizational knowledge, and takes action through connected tools. For every proposed connection, check what information it can read, what it can change, and which permissions administrators can restrict.
This prevents redundant purchases and reveals whether the technology foundation can support the intended use cases.
Fundraising.AI connection: Transparency and Explainability, Privacy and Security. Understand capabilities, limitations, and access before enabling them.
In practice: British Heart Foundation tested Microsoft 365 Copilot within its existing technology environment for document work, communication, and search. Microsoft’s 2024 customer story illustrates evaluating AI in a stack the organization already uses.
3. Data security and governance readiness
AI can magnify existing data and permission problems. Review data quality, access controls, sensitive information, vendor practices, privacy requirements, acceptable use, human oversight, and accountability.
Check vendor data retention, training use, contractual protections, and deletion options. Review shared folders before connecting AI to them. Define approval points and logs for agents that send messages or change records, and a way to stop or reverse their actions. Test whether misleading instructions inside retrieved content can redirect the workflow.
Governance should not be designed to stop experimentation. It should make responsible experimentation possible.
Fundraising.AI connection: Privacy and Security, Data Ethics, Legal Compliance, Accountability. Set data rules and name responsible owners.
In practice: Parkinson’s UK describes its AI use in its public privacy notice, including processor agreements and treating AI as support rather than a decision maker. This is a concrete example of making data practices visible; its UK legal basis should not be copied automatically by a US nonprofit.
4. Use case and opportunity assessment
Start with work that matters, not impressive demonstrations. Identify opportunities across fundraising, communications, programs, finance, operations, executive work, and technology. Evaluate each opportunity based on potential value, effort, risk, and readiness.
Separate drafting from acting. Creating a donor email for staff review is a different risk from sending it autonomously. Consider accessibility, possible bias, and donor or beneficiary dignity alongside time savings. Test whether ordinary automation would solve the problem more reliably or at lower cost.
The result should be a short list of realistic priorities, not a catalog of everything AI might someday do.
Fundraising.AI connection: Social Impact, Inclusiveness. Prioritize useful outcomes and examine who could be harmed or excluded.
In practice: Khan Academy matched AI to a specific mission problem: students needing different kinds of learning support. OpenAI’s account of its initial Khanmigo pilot illustrates choosing a use case around a known need, rather than starting with a generic tool demonstration.
5. AI strategy and roadmap
Decide what to do now, next, later, and not at all. Determine when to use capabilities already in the stack, purchase a product, configure an automation, or build something more specialized. Define prerequisites, owners, costs, and measures of success.
Budget for setup, staff review, training, integration, and ongoing usage as well as licenses. Set decision points for expanding or stopping each pilot, and revisit the roadmap when capabilities or organizational needs change.
Fundraising.AI connection: Accountability, Sustainability. Assign owners and plan for ongoing cost and environmental impact.
In practice: Good Things Foundation includes AI literacy and responsible AI standards in its published 2026–2028 strategy, alongside digital inclusion and community evidence. This demonstrates placing AI within mission priorities; the document is an organizational strategy, not a detailed internal AI rollout plan.
6. Staff enablement and adoption
Training is part of enablement, but it is not the whole answer. Staff need role-specific examples, approved tools, practical guidance, time to experiment, reusable workflows, support, and clarity about where human judgment remains essential.
Create a shared space for approved examples, mission and brand guidance, reusable instructions, and lessons learned, with suitable access controls. Teach staff to check sources, identify uncertainty, and recognize when a workflow needs escalation. Provide practice with real tasks and make room for questions and feedback.
Fundraising.AI connection: Continuous Learning, Collaboration. Build practical skills and share lessons with colleagues and peers.
In practice: Fund for the City of New York created a three-part nonprofit AI bootcamp with OpenAI Academy, covering foundational skills and responsible use. This is an example of a nonprofit investing in sector enablement, with materials teams can use in their own learning.
7. Implementation measurement and continuous improvement
Pilot the highest-value opportunities, test them with real users, monitor risk and performance, and improve them before expanding. AI products and organizational needs will continue to change. Governance and measurement must continue as well.
Measure quality, time saved after human review, cost per completed task, adoption, and incidents against the baseline. For fundraising pilots, consider net revenue, donor response, complaints, and trust. Retest after major model, tool, or permission changes. Keep a human fallback and clear criteria for stopping a workflow.
Fundraising.AI connection: Continuous Learning, Accountability, Transparency and Explainability. Review results, document actions, and correct problems.
In practice: British Heart Foundation retail pilot moved an AI-assisted donated-stock application from development into selected stores before broader expansion, according to Microsoft’s July 2025 case study. It illustrates staged implementation. The additional sales and Gift Aid amounts in that report are projections, not verified realized returns.
What newer AI capabilities change
AI can increasingly complete connected work. ChatGPT Work, for example, supports multi-step research, analysis, finished documents, spreadsheets, and supported scheduled or event-triggered tasks. Availability depends on the plan, permissions, and connected apps. Evaluate an entire workflow, including its review and handoff, rather than judging only a single response.
The practical opportunity is to move from isolated prompts to repeatable work: assemble approved source material, draft an output, check it against defined criteria, and route it to a person who can approve the next action. More capable tools make this easier; they also make access controls, source verification, and clear action boundaries more important.
Fundraising.AI’s November 2025 revision addresses agentic AI through guidance on authorization, audit trails, disclosure, and donor dignity. Build those considerations into the first pilot.
Responsible AI runs through every step
This process supports the Fundraising.AI Framework for Responsible and Beneficial AI for Fundraising. The Framework establishes tenets for responsible and beneficial use; the Cathexis process turns those tenets into organizational questions, decisions, safeguards, pilots, and ongoing practices.
That is especially important when AI affects donors, prospective donors, beneficiaries, volunteers, or employees. The more consequential the decision or action, the stronger the need for transparency, appropriate human oversight, and a clear path for accountability.
Start putting the Fundraising.AI Framework into practice
The Fundraising.AI Framework is intended for organizations to adopt and use. Begin by reviewing it with a small cross-functional group, signing up to support it where appropriate, and identifying what each tenet means for your fundraising and organizational practices. Then work through the seven parts of the Cathexis process: document what is known, what is uncertain, where immediate guardrails are needed, and which two or three opportunities are worth testing.
Outside help becomes useful when the organization needs independent facilitation, a broader technology assessment, deeper governance expertise, vendor evaluation, or implementation support. The value is not a secret checklist. It is applying the questions honestly and turning the answers into decisions.
What you can do today
- Name an executive sponsor and a small cross-functional AI working group.
- Ask each department for one time-consuming process and one AI-related concern.
- Rate your organization informally across the seven areas and document where evidence is missing.
- Choose one low-risk question to investigate rather than purchasing a tool immediately.
Responsible AI questions to consider
- Who represents fundraising, programs, operations, technology, leadership, and the communities affected by these decisions?
- Which Fundraising.AI tenets are already reflected in policy and practice, and where is evidence missing?
- Who is accountable for turning principles into operating decisions and reviewing them as technology changes?
Next ThursdAI: Responsible AI starts with the organization, not the tool. We will examine leadership intent, current use, staff readiness, workflows, and the people affected by AI.
Sign up to be notified when the next ThursdAI article goes live!
—
Cathexis Partners helps nonprofits assess their technology environment, improve business processes and select platforms based on documented organizational needs. Contact us to discuss your next step.
