AI for Nonprofits (4 of 10): Start With the Organization Not the AI Tool

October 22, 2026/ Mark Becker / ThursdAI

One way to make a poor AI decision is to begin with a product demonstration before agreeing on the problem.

The tool may be impressive, but the organization still has to answer a more basic question: what problem are we trying to solve?

In Article 3, we introduced the seven-part Cathexis AI Readiness & Enablement Process. This installment examines step 1: organizational discovery and AI readiness. The aim is to understand goals, existing practices, and concerns before evaluating tools. Later installments will address the technology landscape, data and governance, use-case prioritization, strategy, staff enablement, and implementation.

Establish leadership intent

Ask leadership what they expect from AI. Common answers include saving time, increasing revenue, improving communications, reducing administrative burden, analyzing information, and helping a small team do more.

Those goals are too broad to guide a purchase. Turn them into concrete questions. Which administrative burden? Whose time? How often does the work occur? What would improve if it took less time? What outcome matters enough to measure?

Leadership should also identify boundaries. Are there uses the organization will not consider? Which decisions must always remain with a person? How much risk and experimentation is acceptable?

For example, “save time on reporting” becomes more useful when the team identifies who prepares the monthly report, which sources they use, how long preparation and review take, and what decisions the report supports. Record that baseline before testing a new approach.

Find current and shadow AI use

Staff may already be using personal accounts to draft appeals, summarize documents, research funders, prepare meeting notes, write formulas, or analyze uploaded files. Shadow AI is AI use outside the organization’s approved tools or oversight. It can also include an unreviewed feature in software the organization already pays for.

Do not begin with an interrogation. A punitive approach encourages people to hide their use. Ask what tools staff use, which tasks they use them for, what information they provide, what has worked, and where they are uncertain.

This inventory often reveals useful experiments and immediate risks. It also shows where policy and approved organizational accounts are needed.

For an initial staff survey, ask: What tool do you use? For which task? What kinds of information do you provide? What has helped or gone wrong? What guidance or support would make the work easier? Ask about information categories; do not ask staff to submit confidential prompts or files to prove their use.

Understand the work

Ask staff to identify repetitive, frustrating, slow, or error-prone activities. Look for processes involving repeated drafting, summarization, classification, research, data movement, or searching across multiple sources.

Then examine the entire process. An AI-generated acknowledgment may save five minutes, but it will not fix a workflow that requires three spreadsheets, duplicate data entry, and manual reconciliation. Sometimes the right answer is better process design or automation rather than generative AI.

Include actions as well as outputs in the discussion. Does a tool only prepare a draft, or can it send an email, change a record, or trigger follow-up work? Record who reviews those actions. A task that affects a donor or beneficiary needs a clear owner even while the organization is still assessing readiness.

Assess staff readiness honestly

Readiness is not a single score. Staff will have different levels of experience, confidence, skepticism, and need. Some will experiment quickly. Others will need clear examples and reassurance. A few may fear that the project is designed to eliminate their positions.

Useful questions include:

• Do staff understand the basic strengths and limitations of generative AI?
• Do they know what information may and may not be entered?
• Can they recognize when output needs verification?
• Do managers allow time for responsible experimentation?
• Are there trusted staff members who can serve as early champions?
• Can employees raise concerns without being treated as resistant to innovation?

Include the people affected by AI

The Fundraising.AI Framework provides the responsible AI principles behind this discovery work. Our application of those principles starts with questions about people and consequences:

• Inclusiveness: Whose perspective is missing, and could the proposed process exclude or disadvantage someone?
• Accountability: Who owns the proposed use, reviews consequential actions, and addresses mistakes?
• Transparency and Explainability: Can we explain where AI contributes and how someone can question a result?
• Social Impact: Does the intended outcome serve the mission while protecting donor and beneficiary dignity?

If an AI-assisted process will affect donors, beneficiaries, volunteers, grant applicants, or community partners, include those perspectives in the assessment. Consider who benefits, who may be burdened, what biases could be introduced, and whether people should be told when AI is involved.

Privacy and Security also matters now. Put immediate boundaries in place when discovery reveals an unsafe practice; do not wait for a later governance exercise before addressing it.

The output of discovery

At the end of this step, the organization should have:

• Clear goals, intended outcomes, and boundaries
• An initial inventory of current AI use, including what tools can access and do
• Operational pain points, with task owners and baseline time or quality where available
• An initial view of staff readiness and concerns
• A small group accountable for the next steps
• Immediate actions needed to reduce obvious risk

This is a discovery record, not a completed AI strategy. It gives the team enough evidence to begin mapping the technology landscape. It also identifies questions for the later use-case, governance, and staff-enablement work.

A documented nonprofit example

Children’s Hospital of Philadelphia offers a fundraising example. In our 2023 interview with Jon Thompson and Nathan Chappell, the discussion began with how technology could advance philanthropy. The team found that wealth alone was not enough to indicate willingness to give. That illustrates a useful discovery question: what assumptions about our work should we test before deciding what information or technology we need?

Illustrative discovery scenarios

The following are hypothetical situations to help your team recognize questions worth exploring:

• Communications staff use personal chat accounts because no organizational tool has been approved.
• A finance employee wants help summarizing monthly variances, while leadership had assumed the main opportunity was donor communications.
• Program staff identify a useful knowledge-search idea but raise concerns about exposing confidential client information.

What you can do today

1. Hold a 30-minute leadership conversation about what AI should accomplish and what it should not do.
2. Send staff a short, nonpunitive survey about current tools, uses, successes, and concerns.
3. Identify three recurring tasks and document each process, its owner, current time or quality problems, and intended outcome.
4. Write down any immediate boundaries, such as prohibiting sensitive information in unapproved tools.

Responsible AI questions to consider

• Have employees, donors, beneficiaries, volunteers, and affected communities had a meaningful opportunity to shape the proposed use?
• Could it exclude people because of disability, language, access, culture, or biased data?
• Is leadership prepared to explain the intended social impact, limitations, and accountability?

Next ThursdAI: We will map where AI already exists across your nonprofit’s technology stack, what it can access and do, and the vendor questions that need answers.

Catch up: Article 1 on AI types | Article 2 on costs | Article 3 on the seven-part process

Sign up for ThursdAI updates

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.

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.