Overview

"We should be using AI" is a sentence heard in nonprofit board rooms and leadership meetings everywhere. But knowing you should do something and knowing how to start are very different things. Most AI strategy conversations stall because they begin in the wrong place, with tools instead of with problems.

This guide walks you through a seven-step process for building a grounded, achievable AI strategy for your nonprofit. It is designed for executive directors, program directors, and operations leads who are not technologists but who are responsible for making their organizations more effective. No prior AI experience is required.

The output of this process is a concrete, one-page AI strategy that identifies your top use cases, assigns responsibilities, establishes a timeline, and sets clear success metrics. You will have something you can share with your board, discuss with your team, and actually execute, not a document that sounds good but never gets implemented.

Before you start: clear your expectations about what AI strategy means. It does not mean predicting the future of AI or developing proprietary machine learning models. For small and medium nonprofits, an AI strategy is simply a clear plan for which specific AI tools your organization will use, for which specific tasks, by which specific people, starting when.

Identify Problems (Not Solutions)

The most common strategic mistake is starting with a tool, "let's use ChatGPT", rather than starting with a problem. Tools without problems are expensive hobbies.

Begin with a structured problem identification exercise. Gather your leadership team, program managers, and one or two frontline staff for a 90-minute working session. Ask each person to answer these three questions in writing before discussing as a group:

1. What are the three tasks in your role that take the most time relative to their importance?
   2. What decisions do you regularly make that feel harder than they should be, because you do not have the right information fast enough?
   3. Where does your work slow down or stall most often, and why?

Common answers from nonprofit teams include: fundraising appeals that take days to draft and revise; grant research that requires manually scanning dozens of funder websites; donor reports that require pulling data from three different systems; program impact summaries that get compiled from paper forms or scattered spreadsheets; volunteer communication that is never timely enough; and board meeting materials that take 20+ hours to compile.

List every problem that surfaces. Do not evaluate them yet. Aim for 15 to 25 specific, concrete problems. Vague problems like "communication is hard" are not useful, push for specificity: "writing the quarterly program narrative for our government funder takes my grants manager 12 hours and it always feels rushed."

Once your list is complete, each team member votes on their top three problems by impact. Tally the votes. The three to five problems that bubble to the top are your starting point.

Evaluate Which Problems AI Can Solve

Not all problems are AI problems. Applying AI to the wrong problems produces frustration, wasted money, and organizational skepticism about AI in general. This evaluation step is where many nonprofits skip ahead and pay for it later.

AI tools in their current state are genuinely strong at a specific set of capabilities: generating and editing written text (emails, reports, grant narratives, social posts), summarizing long documents into key points, categorizing and organizing unstructured information, identifying patterns in structured data, answering questions about documents you upload, and generating first drafts of structured content based on bullet points.

AI tools are not well-suited for: problems that are fundamentally about organizational structure or process (if your program data is collected inconsistently by three different staff members, AI cannot fix that inconsistency), decisions requiring deep contextual judgment about specific individuals (which major donor relationship strategy to pursue), tasks requiring real-time information not available in the AI's training data, and problems where the primary issue is human communication breakdown within your team.

Apply this framework to each problem on your list. For each one, ask two questions: (a) Is this problem primarily about generating, processing, or organizing information? If yes, AI may help. (b) Would this problem go away if we had a more organized process or better tool (not AI)? If yes, fix the process first.

Example evaluation: "Donor data is scattered across a spreadsheet, our email platform, and our event system". This is a data infrastructure problem. AI cannot fix scattered data; it can only operate on data you give it. The fix is consolidating your CRM first. AI can then help you analyze that consolidated data.

Example evaluation: "Writing our year-end fundraising appeal takes 6 hours of the development director's time". This is an information-processing and writing task. AI can draft a strong first version in 10 minutes from a brief. This is an AI problem.

For a team of 5-15 staff at a typical small nonprofit, you will usually find two to four problems where AI genuinely helps, and several where the real fix is a process or tool change.

Size the Opportunity

Before committing to any AI use case, estimate the real impact. This serves two purposes: it helps you prioritize which problems to tackle first, and it gives you the language to justify investment to your board and funders.

For each AI-solvable problem, answer three questions:

How much time does this currently take? Be specific. Not "a lot" but "approximately 4 hours per grant cycle, and we do 8 grant cycles per year, for a total of 32 hours per year."

How much time would it take with AI assistance? Be honest and conservative. AI rarely eliminates a task. It usually accelerates it. A realistic estimate is that AI draft generation takes 20-30% of the current time, but human review and editing adds back 30-50% of that savings. Net savings on writing tasks are typically 50-70% of current time.

What is that time worth? Use fully loaded staff cost (salary plus benefits). If your grants manager earns $52,000 per year with benefits, that is roughly $25 per hour. If AI saves 20 hours per year on grant writing tasks, that is $500 in recovered capacity, and more importantly, 20 hours redirected to other high-value work.

Sample opportunity sizing for a mid-sized nonprofit:

• Email and appeal writing: 45 hours/year, saves 30 hours, $750 value + higher quality
  - Grant narrative drafting: 80 hours/year, saves 40 hours, $1,000 value + faster turnaround
  - Program impact summaries: 24 hours/year, saves 12 hours, $300 value
  - Total: ~82 hours saved per year, ~$2,050 in recovered staff capacity

At a ChatGPT Plus subscription cost of $240/year for one license, this is a clear ROI calculation. Present this framing to your board: AI tools are a capacity investment, not a technology expense.

Note that hours saved is not the only value. Reduced stress, higher output quality, faster turnaround on donor communication, and staff satisfaction are real benefits that do not appear in the calculation but matter significantly.

Build Your Use Cases

A use case is a one-paragraph description of exactly how AI will be applied to a specific problem, by whom, using which tool, in what workflow. Use cases are the core of your AI strategy. They turn abstract intentions into concrete plans.

For each priority problem, write a use case following this template:

Problem: (Specific description of the current pain point)
AI Solution: (What the AI tool will do, specifically)
Tool: (Which platform: ChatGPT, Claude, Canva AI, etc.)
Workflow: (Step-by-step description of how a staff member will use the tool)
Time savings: (Estimated hours saved per week, month, or year)
Who is responsible: (Named person who owns this use case)
Success metrics: (How you will know it is working)
Risks: (What could go wrong, and how you will mitigate it)

Example use case: Development Director, Appeal Writing:

Problem: Writing fundraising appeals takes the development director 4-6 hours per appeal, often resulting in delays or rushed copy. The organization sends 10 appeals per year.

AI Solution: ChatGPT or Claude will generate a complete first draft of each appeal, including subject line, opening story, impact statistics paragraph, call to action, and closing. The development director will edit the draft for accuracy, voice, and specific donor relationship context.

Tool: ChatGPT Plus ($20/month, shared account).

Workflow: (1) Development director writes a brief in a standard template: donor segment, key impact story, ask amount, deadline. (2) Brief is pasted into a saved ChatGPT prompt template. (3) AI generates draft in 2 minutes. (4) Development director edits for 45-60 minutes. (5) Final copy is loaded into email platform.

Time savings: 3-4 hours per appeal, 30-40 hours per year.

Success metrics: Appeals completed at least 5 business days before send date (vs. current average of 1 day). Development director reports satisfaction with AI draft quality. No decrease in open rates or donation conversion rates.

Risks: AI draft may include inaccurate impact statistics. Mitigation: staff reviews all statistics against source data before editing is complete. AI may produce generic tone. Mitigation: maintain a brand voice guide that is included in every prompt.

Draft use cases for each of your top three to five AI-solvable problems. These documents become the core of your strategy.

Prioritize

With two to five use cases drafted, you need to sequence them. Trying to implement everything at once is how AI strategies fail: staff get overwhelmed, training is scattered, and nothing gets adopted deeply.

Prioritize use cases by plotting them on a simple 2x2 matrix with Impact on one axis and Implementation Difficulty on the other. High-impact, low-difficulty use cases go first. Low-impact, high-difficulty ones may not be worth pursuing at all.

Impact factors: hours saved per year, dollar value of recovered capacity, quality improvement (e.g., better grant writing leading to higher award rates), speed of delivery (faster donor response).

Implementation difficulty factors: technical complexity of the workflow, how much staff behavior change is required, data quality requirements, training time needed.

Typical prioritization result for a small nonprofit:

Priority 1 - Email and appeal writing: High impact (30+ hours/year), low difficulty (staff just need to learn a new draft process). Start here.

Priority 2 - Grant narrative drafting: High impact (40+ hours/year), medium difficulty (requires creating good prompt templates and reviewing AI output against funder requirements). Start after email writing is working well.

Priority 3 - Impact report compilation: Medium impact (20 hours/year), medium difficulty (requires reliable data sources). Start in quarter two or three.

Priority 4 - Donor segmentation analysis: Medium impact, higher difficulty (requires clean CRM data, more technical setup). Defer until data quality is addressed.

With your priority sequence set, assign each use case a target start date. Space them at least six to eight weeks apart to allow genuine adoption of each before introducing the next. An AI strategy that adds one well-adopted use case every two months will deliver far more value than one that adds five poorly-adopted use cases in the first month.

Pilot the First Use Case

Strategy without execution is just documentation. The pivot from strategy to reality happens when you run your first AI pilot, a structured, time-limited test of your highest-priority use case.

A nonprofit AI pilot has four components:

Scope: One use case, one team or department, 60-90 days. Deliberately narrow. Resist the temptation to expand scope mid-pilot. If your first pilot is AI-assisted email writing for the development team, it is only AI-assisted email writing for the development team during that period.

Training: Before the pilot launches, run a hands-on training session of 60-90 minutes. Cover: what the tool does and does not do, how to write effective prompts for this specific use case, how to edit AI output rather than accepting it uncritically, and privacy guidelines for what information should and should not be input into the tool. Create a written reference card staff can consult during the pilot.

Measurement: Establish baseline metrics before the pilot begins. If you are measuring time savings on email writing, log how long the last three appeals took to draft. If you are measuring quality, collect the last three appeal open rates. Then track these same metrics during the pilot.

Review: At the end of the pilot period, conduct a 60-minute review meeting. What worked? What did not? What would need to change to make this a permanent part of the workflow? What unexpected benefits or problems emerged? Capture the answers in writing.

Common pilot outcomes: Many organizations discover the AI tool works better than expected for some tasks and worse for others. The most common finding is that output quality is high when staff take time to write detailed prompts and edit output carefully, and low when they paste minimal input and expect finished copy. This insight, that the quality of what you put in directly determines the quality of what you get out, becomes the foundation for training the next team that adopts the use case.

Document Your Strategy

Once you have completed the first pilot and gathered real data, document your full AI strategy in a concise, shareable format. Aim for one to two pages, long enough to be substantive, short enough that your board and staff will actually read it.

A nonprofit AI strategy document typically includes:

Vision statement (two to three sentences): What does your organization want AI to enable? Example: "We will use AI tools to amplify the capacity of our team, enabling them to serve more constituents without burning out. AI will handle repetitive information tasks; our staff will focus on relationships and judgment."

Use cases (bulleted list, prioritized): Each use case with a one-sentence description, the tool to be used, the team responsible, and the target start date.

Timeline: A 12-month roadmap showing which use cases launch in which quarter, when evaluations happen, and when decisions about scaling or discontinuing will be made.

Budget: A realistic estimate of tool costs (most nonprofit AI budgets are $50-$500 per month for small organizations), plus staff time for training and implementation. Do not underestimate staff time, a realistic figure is 10-15 hours per staff member for initial training and workflow adjustment.

Governance: Who has authority to approve new AI tool subscriptions? Who reviews privacy and data security for new tools? Who is the internal AI champion who stays current on the landscape and troubleshoots problems?

Ethics and values commitments: Be explicit. "We will not use AI to make decisions about program eligibility or client services without human review." "We will not input personally identifiable client information into public AI platforms." "We will be transparent with donors and clients about our use of AI-generated content."

Communication plan: How will you communicate AI adoption to staff, board, and community? Proactive communication reduces anxiety and resistance.

Share the strategy document with your board at the next meeting. Frame it as an efficiency and capacity investment, not a technology experiment. Boards respond well to clear ROI framing: if AI tools save 80 hours of staff time per year and the average cost is $300/year in subscriptions, that is a 40:1 return on investment in recovered capacity.

Key Takeaway

The nonprofit organizations that get the most value from AI are not necessarily the ones that use the most sophisticated tools. They are the ones that were disciplined enough to start with real problems, honest enough to evaluate which problems AI actually solves, careful enough to pilot before scaling, and thorough enough to document what they learned.

A strategy built on this foundation, problems, evaluation, sizing, use cases, prioritization, piloting, documentation, is fundamentally different from one built on enthusiasm about a tool or pressure to "keep up with AI." It is slower to build, but dramatically more likely to deliver lasting value.

Start small and specific. Pick one writing task that currently costs your team four or more hours per week. Spend two weeks learning how to prompt an AI tool well for that specific task. Document the time savings. Share the result with your leadership team. That single successful use case is worth more than ten half-implemented ideas, and it builds the organizational confidence to expand AI adoption systematically over time.

Frequently Asked Questions

Should we involve staff in building the strategy?

Yes, and this is not optional. Staff who are excluded from AI strategy development become resistant to AI adoption. The people doing the work know the pain points better than leadership does. They will identify problems that executives are not aware of. More importantly, staff who participate in designing the strategy feel ownership over it. They are far more likely to adopt tools they helped choose than tools that were handed down to them. Include at least one representative from each major functional area (programs, development, operations) in the problem identification and use case development steps.

What if we're not sure which problems AI can actually solve?

Start with a simple test before committing to any strategy. Take one time-consuming writing task your team does regularly, a donor update email, a program summary, a meeting agenda, and spend 30 minutes trying to get AI to help with it. Use the free tier of ChatGPT or Claude. Give the AI a detailed description of the task and see what it produces. This direct experience is more informative than any amount of research. If the output is 60% usable with editing, AI can likely help with that problem. If it produces something that requires complete rewriting, the tool may not be worth it for that specific use case.

How do we know if an AI solution is actually working?

Define your success metrics before the pilot begins, not after. If you measure success only by positive feelings about the tool, confirmation bias will make any tool look successful. Measure concrete things: hours spent on specific tasks before and after, number of drafts completed per week, days between task start and completion, staff-reported confidence ratings on specific tasks. Set a specific threshold for success, for example, "the tool is worth continuing if it saves at least two hours per week and staff report it is not adding more work than it saves." Evaluate against that threshold, not against a vague sense of whether things feel better.

What is a realistic timeline for seeing results from an AI strategy?

Expect the first month to feel slower, not faster. Staff are learning new tools, adjusting workflows, and building prompt writing skills. Month two typically shows the first real time savings as habits form. By month three of a focused pilot, most teams report consistent savings and higher confidence. For a full organizational AI strategy, covering three to five use cases across multiple teams, expect 9 to 12 months from strategy development to all use cases operating smoothly. Do not set board expectations for dramatic efficiency gains in the first quarter; set them for steady, documented progress across a year.